C. GORDON BELL with JOHN E.

McNAMARA
-

HIGH-TECH VENTURES
THE GUIDE FOR ENTREPRENEURIAL SUCCESS

A
vv
ADDISON-WESLEY PUBLISHING COMPANY, INC.
Reading, Massachusetts Menlo Park, California New York
Don Mills, Ontario Wokinghan, England Amsterdam Bonn
Sydney Singapore Tokyo Madrid San Juan Paris
Seoul Milan Mexico City Taipei
Many of the designations used by manufacturers and sellers to distinguish their products
are claimed as trademarks. Where those designations appear in this book and Addison-
Wesley was aware of a trademark claim, the designations have been printed using initial
caps.

Library of Congress Cataloging-in-Publication Data

Bell, C. Gordon
High-Tech ventures : the guide for entrepreneurial success /
C. Gordon Bell with John E. McNamara
p. cm.
ISBN 0-201-56321-5 (hardback)
1. High technology industries-Management. 2. New business
enterprises-Management. 3. Entrepreneurship. 4.Computer
industry-Management. 5. Computer software industry-Management.
I. McNamara, John E. 11. Title
HC79.HS3B45 1991 90-25717
620.0064~20 CIP

Copyright 0 1991 by Addison-Wesley Publishing Company, Inc.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form, or by any means, electronic, mechanical, photocopying,
recording, or otherwise, without prior written consent of the publisher. Printed in the United
States of America. Published simultaneously in Canada.

Cover design by Hannus Design Associates

Cover photographs by Kurt R. Hulteen, Lisa LaForge, and John W. Wait
Text design by Kenneth J. Wilson
Set in 10 point Palatino

ISBN 0-201-56321-5
Text printed on recycled and acid-free paper.
6 7 8 9 10-MA-0099989796
Sixth printing, July 1996
CONTENTS

Preface v
Chapter 1 Introduction: The Formation of High-Tech Companies 1
Chapter 2 The People 10
Chapter 3 The Business Plan: A Road Map and a Scorecard for the Future 34
Chapter 4 Cash, Financeability, and Control 59
Chapter 5 Technology and Engineering 85
Chapter 6 The Technology Balance Sheet 114
Chapter 7 Manufacturing 140
Chapter 8 The Product 152
Chapter 9 Marketing and Sales 201
Chapter 10 The Bell-Mason Diagnostic 251
Chapter 11 Case Studies 272
Chapter 12 Technical Workstations: Heuristics from History 316
Chapter 13 TheFuture 339
Bibliography 368
Index 374
Acknowledgments 387
PREFACE

High-Tech Ventures is written primarily for those who are creating the future
high-tech world by designing, building, and marketing innovative products. It is
equally useful for board members, investors, attorneys, accountants, consultants, and
others who are intimatelyinvolved in new ventures.These readers, plus students of the
start-up process and other voyeurs, will find here a quantitative evaluation method,
which includes a set of rules for examining a companyand comparingit with an ”ideal”
organization.
The analytical approachpresented in High-Tech Ventures differs considerably from
the anecdotes, testimonials, and confessionsso commonly found in today’s autobiog-
raphies and case studies.Case studies are by no means ignored in this book, however,
since the evaluation rules have been generated from actual cases and an entire chapter
is devoted to case studies. But the diagnostic method described here is much broader
than the approach employed in typical case studies, which focus on a particular
discipline--such as marketing, finance, or management-to enable readers to under-
stand a situation that aptly illustrates a flaw or an exemplary action. In contrast, the
diagnostic method presented in this book enables users to examine all the critical
dimensionsthat affect a new venture, which could very well reveal multiple flaws in the
company.Furthermore,the diagnosticprovidesa consistentmethodology for analyzing
and comparing cases.
AlthoughHigh-TechVentures is designed to serveembryonicentrepreneurialfirms,
the guidelinesgiven hereare equallyusefulforlarge-companyintrapreneurs(individuals
within big, established organizationswho are creating new businesses).These pioneers
must deal with the same technologies,products, markets, and general environmental

V
vi Preface

rules as entrepreneurs. All too often, however, intrapreneurs’ suggestionsare rejected

because the new products they propose either do not fit witlun the current business or
threaten the status quo. Thus, start-ups are the main arena for innovation in the world
market of informationprocessing, even though the basic technology and product ideas
often originate in a research laboratory or a large company.
As suggestedabove,most of t h s book outlinesa diagnosticmethod that canbe used
to assess the health of a high-tech venture.The idea of developing a tool for diagnosing
start-upsevolved over a periodof nearly ten years. The possibility first presented itself
when I was vice president of engineering at Digital Equipment Corporation (DEC).In
that position, I defined rules for testing the management, team, market(ing),product
position, and product development in an internal DEC guide formanagers, “Heuristics
for Building Great Products.”Since then, I have been involved with about twenty start-
ups, including Ardent Computer Corporation (now Stardent Computer, Inc.);Encore
Computer Corporation; MIPS Computer Systems, Inc.; Silicon Compiler Systems
Corporation (now Mentor Graphics Corporation); Silicon Graphics, Inc.; and Visix
Software, Inc.
My observations in these ventures-typically, from the vantage point of a tech-
nologist and product architect-have, over the course of a long collaboration, been
merged with the marketing, sales, and strategic insights of Heidi Mason, founder and,
until recently, chief executive officer of Acuity, a strategic marketing and public
relations firm in Silicon Valley. Our research has been encapsulated into a series of
guidelines, which take the form of questions or rules, known as the Bell-Mason
Diagnostic.
TheBell-MasonDiagnosticconstitutesa rule-based,human-applied,expert system.
Its more than six hundred rules are gleaned from experience and research in start-ups
and established organizations. Review and feedback from peers supplemented the
originalwork, and testing performed on high-tech companiesboth by ourselvesand by
Coopers and Lybrand provided a cross-check. Finally, the diagnostic was licensed to
Coopers and Lybrand for use in assessing high-tech start-ups.
As mentioned earlier, the rules reflect the diagnostic’sview of an ideal company,
with which an organization can compare itself. These rules take the form of objective
questions (e.g., regarding the existenceand content of certain plans and processes) that
can be used to gather verifiable information from key people in the company. Formal
assessment of a firm includes the following elements: a review of written plans, an
evaluation at the company’s site (tosecure more detailed documentation and simply
look at the organization and its progress), analysis of the information that has been
collected,preparation of a written assessment,and a feedbacksessionwith the principals
in the firm.Such a formal assessment should be conducted by a team of two or three
individuals who are familiar with the technology and the product and who have a
general knowledge of marketing and sales, organization,control, and finance.
 Preface vii

TheBell-MasonDiagnosticcanalsobeused informally.Anyone startinga high-tech

venture, or contemplating joining a start-up, can read the diagnostic and answer the
questions contained therein to test whether the company is healthy.
Although the diagnostic’s rules have been posited, reviewed, and tested, the user
is free to change or weight them to assess special situations,to perform a more detailed
analysis of a particular type of organization, or to emphasize certain dimensions. For
example, accounting and legal firms might add more rules in the cash, finance, and
control dimensions.
Futurists may be interested in reading High-Tech Ventures simply to learn more
about high informationtechnology,includingthe products, market, and trends. People
who suffer from ”technolophobia”and ”graphophobia”to such an extent that they are
frightenedby simple graphs and shudder at logarithrmcscalesare welcometo skipover
the graphs in this book and concentrate on the text.
Although the preceding paragraphs imply that the Bell-Mason Diagnostic was
developed through a simple evolutionary process, it wasn’t quite that simple. When
Heidi Mason and I first tried to apply a small set of rules to diagnose a start-up, we
discovered that the rules were inadequateand that, to be effective,they would have to
vary in detail with time, just as a new venture’s circumstancesvary with the passage of
time. We found that we had to precisely define the stages and substagesthrough which
a start-up passes in order to create detailed rules that would be appropriate for each
stage. In general, a start-up has two planning stages: the concept stage (stageI) and the
seed stage (11).These precede the heavily funded product development stage (111)and
the even more expensive market development stage (IV),when the company investsin
manufacturingand distribution.Finally, if all goes well, the firm enters the steady-state
stage (V).
In addition to the five time-dependent stages, there are twelve dimensions, or
aspects, that can be used to evaluate any company. The combination of five stages and
twelve dimensions raises the possibility of a sixty-chapter book! However, two sim-
plificationshave been made herein. Namely, only the first two stages,concept and seed,
are discussed in detail, since the viability of an entire venture is most often determined
during these initial, critical stages, and a number of the chapters cover more than one
dimension.Chapter 2, for example,deals with three dimensions:the CEO, the team, and
the board of directors. The business plan dimension is discussed in Chapter 3; cash,
financeability,and control,inchapter 4; technology,inchapters 5and 6; manufacturing,
inchapter 7; the product,inChapter 8;and marketingand sales,in Chapter 9. The entire
Bell-MasonDiagnosticis explained in Chapter 10, which includesa set of general, time-
independentquestionsthat can be used to evaluate a company on a cursorybasis at any
stage in its development.
No simple,one-lineformulasuch as “hiregood people”or “build an irresistibleand
unique product’’ will ensure the successof a high-tech venture. However,being able to
viii Preface

answer all the diagnostic’s questions affirmatively is a good start! Although in devel-
oping the questions,every effort was made to emphasizeeach of the company’stwelve
dimensionsequally and to give equal weight to the conceptand seed stages,a few of my
ownbiases have probablycreptin,especiallywith regard topeople, cash,manufacturing,
and marketing.
The following subsections provide a brief overview of some of the critical deter-
minants of a start-up’s health.

PEOPLE

Since a start-up is a fragile, embryonic organization, it is at the mercy of its chief

executiveofficer,the latest great American cowboy.Companiesfold or perform poorly
as a result of CEO failuremore often than they do from any other singlecause. Common
faihngs include the inability to manage, to hire and fire, and to make good and timely
judgments.Managers from large organizationsare especiallydangerous, sincethey are
often incapableof performing basic functions without the plethora of resources avail-
able in a big company. Various sins such as greed, sloth, and egoism can also kill an
otherwise potentially healthy venture.
I have seen too many CEOs with brittle egos who will do almost anything to be
considered right and who constantly place themselves ahead of their team members
and their organization.These individual-second CEOs, in many cases-are likely to
abandon a company that they themselves caused to fail. Ironically, their departure is
often cushioned by a “golden parachute.”

CASH

From study and experience, I have come to believe that too many entrepreneurs
have, in the past, had access to too much capitalto fund too many marginal-technology
companies marketing ”me-too” products. Furthermore, these entrepreneurs have
frequentlyfelt that they could afford to skip the seed stage and go directly to the market
developmentstage.Throughout the book,I will attemptto show why this is a very risky
strategy.
Early profitability is stressed in almost every chapter because, in a successful
venture,early profitabilitylessensor eliminatesthe need for further use of investorcash.
Profit is habit-forming-and so is loss. This bias comes from studying successfulstart-
ups. Each time the issue of profitability is raised in the book, I try to present the
underlying theory of its importance. For example, an unprofitable company cannot
become self-sustaining and will therefore eventually cease to exist.
 Preface ix

MANUFACTURING

It is essential for a start-up to determine whether another company is in a better

position to manufacture (or market) the start-up’sproduct. If so, the start-up must form
a key strategic partnership with this firm rather than trying to reinvent the wheel by
developing its own proprietary technologies, markets, and distribution channels. All
too often, however, time, greed, and ego prevent a start-up from seeking such rela-
tionships.

MARKETING

Most companies regard marketing as something of a black art, and none of the
organizationsanalyzedby the Bell-MasonDiagnosticcame closeto the ideal in t h s area.
Marketing is inherently the most difficult dimension to plan, execute, and measure,
because the long time span between an action and its result makes it hard to determine
the correlation between the two. Furthermore, there is the classic uncertainty over
whether a product’ssuccesscan be attributed to the quality of the product or the quality
of the marketing effort. Both a great product and great marketing are essential!

Companies that fail are, for the most part, those that are unable to deal with the
complexity of technology and the fast pace of technological change while simulta-
neously growing as organizations.Start-ups operate in an undefined, high-risk envi-
ronment, in an emergingmarket.Someunsuccessfulfirms cannot meet the challengeof
defininga product; others try to create customersin markets for which no market model
exists.
Successful start-ups, in contrast, manage to prevail over the inherent risks and the
challenge of technologicalcomplexity.They are able to increase their size, define their
product or products, and create their market.The successof these ventures would seem
to suggest that start-up experience is a prerequisite for establishing a viable organiza-
tion. There is an apparent paradox, however: many of the firms that created new
industries-Apple Computer,DEC, Microsoft,and SunMicrosystems-were begun by
entrepreneurswho had no previous experience in founding a company.
Startinga high-tech companyis more difficulttoday than ever before, because most
of the classes of computers and software, as segmented by price, have already been
developed and are being marketed successfully.Furthermore, in the early 1990s, less
venturecapitalis likelytobe availablefor new ventures.Thesefactors,whilediscouraging,
do not toll the death knell for the start-upprocess.They merely mean that entrepreneurs
and venture capitalists are going to be a lot more careful.
x Preface

Keeping an established high-tech company running successfully is also more

difficult today than ever before. A number of readershave argued, for example,that the
best-selling book In Search of Excellence (Peters and Waterman, 1982),which studied
someof our nation’smost successfulfirms,was flawed because severalof the companies
described therein later became less successful.Not so! The methodology presented in
High-Tech Ventures clearly demonstratesthat a growing firmin an emergingmarket can
easily get into trouble-or out of trouble-at any time and in any number of ways.
A case in point is the ”palace revolt” that occurred at a profitable, growing New
England start-up, prompting the board to replace the successful CEO, which, in turn,
eventually resulted in a poor merger. Likewise, a very large company can become
complacentand overlybureaucratic, or its healthcanbejeopardizedby aninappropriate
organizationalstructure.In 1990, DEC is sufferingboth from having held on to its single
VAX product vision for too long and from the displacement of minicomputers by
personalcomputers and workstations.Each of the twelvedimensionsof the Bell-Mason
Diagnostic includes at least one fatal flaw, which, if manifested at any stage of a
company’s life, can spell disaster!High-Tech Ventures also includes many case studies
and vignettes that illustrate this capacity for sudden failure or success.
This book does not present advice or step-by-stepdetails on what to do and how to
do it when founding a company.Instead, it posits a new way of examining embryonic
firms and emergingbusinessesvia the Bell-MasonDiagnostic,which constitutesa guide
and a checklist for the start-up process. (Checklists are powerful tools that every
engineer and marketing manager should learn to use.) The book also containsa wealth
of informationabout the computing industry.As an assessment tool, it can help readers
determine whether they are ready to start a company, or launch a new project within
their present company,and providesa model for doing so.Thus,High-Tech Venturestells
readers whether they understand the ”how-to” books, because it is, in effect, a
“should-be”book. (Thosewho are interested in the “how-to” books will find a number
of good ones listed in the annotated bibliography.)
In a way, founding a company or becoming involved with a start-up is like
participatingin an exciting sport: the risks are great, but so are the rewards.Carryingthe
analogy a step further, I might add that before engagingin any sport, ifs always a good
idea to learn as much as possible about it-by studying the winners and the losers, the
smartmovesand thefata1accidents.High-TechVentures will helpreadersdoexactlythat.
Good luck!

ACKNOWLEDGMENTS

I would like to thank the following reviewers for providing significant input: Paul
Baran, SteveBlank, SallyBrowning,Henry Burkhardt,Bob Chinn, SteveCoit,Elizabeth
 Preface xi

Corcoran, John Doerr, Alison Elliott, Mary Jane Forbes, Ted Johnson,Karen Mathews,
Pamela McCorduck, Sesha Pratap, David Salzman, Michael Schrage, John Shoch, Bert
Sutherland, and Bill Taylor.
Major contributionswere made by the followingindividuals:Ronna Alintuck,who
contributed to both the Gateway story and the marketingchapter;Adam Bosworthand
Eric Michelman, who provided the Analytica story; Mark Duncan, of Askmar, who
provided many of the figures in Chapter 5; Barry James Folsum and Bonnie Johnson,
who offered feedback on the book and applied the Bell-Mason Diagnostic.at Focus
Systems; Prabhu Goel, who contributed to the Gateway story; John Grdlos, who
contributedto the financingchapter;Lowell Hawkinsonand Bob Moore, who provided
theGensym story;Jeff Kalb, who provided theMasPar story; Bob Keeley,who tested the
use of the diagnosticon his classes in entrepreneurismat Stanford; Vinod Khosla, who
provided the Sun Microsystems business plan; Dave Nelson and Bill Poduska, who
provided the Apollo business plan; Matt Sanders, who contributed the basic ideas for
the manufacturing chapter; Arthur Spinner, who contributed to the board of directors
section; and Suhas Patil, who provided the Cirrus Logic story.
Gwen Bell made many suggestions, through discussion and detailed editing, and
was also a constant source of encouragement.
Coopers and Lybrand deserves special thanks. Bob Stavers of the San Jose office
helped create the Bell-Mason Diagnostic and make it work on real companies. Cheryl
Suchorsof the Coopersand Lybrand National High Tech Group in Boston worked with
Bob to champion the diagnostic and license it throughout Coopers and Lybrand.
The book could never have been written without Heidi Mason’s collaboration,
which resulted in the Bell-Mason Diagnostic.
1gratefully acknowledgethe followingfor their kind permission to reprint or adapt
some of the artwork that appears in this book: Askmar (Mark Duncan), The Gartner
Group, Harvard Center for Information Policy Research (Anthony G. Oettinger, John
F. McLaughlin, Anne Louise Antonoff),Hewlett-Packard (Frank Ura), Intel Corpora-
tion, International Data Corporation,and the IEEE ScientificSupercomputer Subcom-
mittee.
I would especially like to thank the staff at Addison-Wesley for their encourage-
ment and dedicated involvement in the book, specifically:Kate Habib, assistanteditor;
P e w McIntosh,production manager; Peggy McMahon,production coordinator;Leslie
Morgan, marketing manager; Linda OBrien, director of technical sales; John Wait,
editor-in-chief, for many ideas and much helpful editing; and Susan Cohan for
copyediting.
Chapter 1

INTRODUCTION:
THE FORlMATION OF
HIGH-TECH COMPANIES

As we move into the 1990s,entrepreneurial self-determinationcontinuesto releasevast

amounts of human energy and drive the formation of ”start-up” companiesbased on
a wide variety of technologies and applications. These emerging companies create a
variety of new products, ranging from relatively simple hardware and firmware
componentsto complete computer systems,which often offer as much as fifty times the
performance or performance/price advantage of products produced by long-
established firms. In bringing a new product to market, the engineering and overall
organizationalproductivity of the start-up can be ten to fLfty times greater than with a
large, existing firm.
High-tech start-ups follow the basic process outlined in Figure 1.1. (Readers
familiar with programming will immediately recognize that the process is formatted
just likea computer program,which seemed appropriatein the present context.Readers
unfamiliar with programming should have no trouble understanding the material
either, since it can simply be read as ordinary text.)
Let’s begin by considering the hrst few lines of this high-tech start-up ”program”:

Start a high-information-technology company

;f frustration is greater than reward
and greed is areater than fear of failure
and a new technology/product is possible then

1
2 Introduction

Start a high-information-technology company

if frustration is greater than reward
and greed is greater than fear of failure
and a new technology/product is possible then
beain
exit (job);
get (tools to write business plan);
write (business plan) ;
get (venture capital) ;
start (new company) ;
get (space, people, product development tools,
UNIX license);
sell (product idea) ; design (product);
market&sell&produce (product);
while new company is not profitable then
wait; get (more $ ) ;
sell (new company);
retire; wait; restart;
if entrepreneur wants to do it again then
start (another high-information-
technology company) else
start (new venture capital company);
&;

Figure 1-1. Simple "Program" for Starting a High-Information-Technology Company.

The foundersof the new companiesdiscussedlater in thisbook are typicallypeople

who, for a number of reasons, were frustratedwith their previousjob. On the one hand,
they felt compelledto keep the job because they needed the paycheckand wanted to stay
in touch with people and advancements in their field. On the other hand, the fact that
their frustration exceeded the rewards they expected to receive from their employer
tended to push them to leave the firm.
When it comes to starting a new company, this "push effect is necessary, but not
sufficient-at least, not by itself. Thousands of people are frustrated in their present job
and yet never strike out on their own. Two additional ingredients are required: having
a cash motivation that exceeds one's fear of failure and being able to envision a new
technologyor product that can serve as the basis for a viable enterprise.These latter two
ingredients are the "pull" to form a new company.
Thus, the "if" underlying the creation of a new firm is a combination of the "push
to leave an old company (involving such considerations as bureaucracy, lethargy,
politics, and frustration) and the "pull" to form a new company (involving such
considerations as a new product, superior technology, recognition, and financial
reward).Throughout each succeeding generation of technology, new companies have
 Introduction 3

formed from the interactionof these two forces,which appear to be of equal importance.
Furthermore, it is not uncommon for the same individual to experience the push and
pull effectsseveraltimes in his or her career,leadingto the formationof severalstart-ups.
For example,Gene Amdahl, SeymourCray, SteveJobs,Gordon Moore, Bob Noyce, and
Bill Poduska each founded more than one new company, most of which are discussed
in greater detail later in the book.
Now let’s continue with the next segment of the high-tech start-up ”program”
presented earlier in the chapter:

begin
exit (job);
g e t ( t o o l s t o w r i t e business p l a n ) ;
w r i t e (business p l a n ) ;

Once the decision has been made to start up a company, the first step is to write a
business plan. Occasionally, entrepreneurs do manage to write business plans while
they are still part of another organization. In the case of most successful companies,
however, a small core of foundersleave their jobs and write a detailed business plan on
a full-timebasis. This process typically takes from three to twelve months, depending
on the technology,market uncertainty,product complexity,and manufacturingprocess.
The standard planning tools used to create the business plan are a personal
computer and a spreadsheet program. The importance of the spreadsheet program is
that it enables the new company’s founders to develop a business plan that offers
investorslarge, but plausible returns. As with all powerful tools, though, spreadsheets
are subjectto misuse, and the market is litteredwith hundreds of businessplans that tout
completely unreasonable and unjustifiable financial numbers. As a result, it is not
surprisingthat high-technologybusiness plans are often perversely described as ”that
place in time and space where the rubber meets the blue sky.”
Because of the range of technologies a new company might utilize and the variety
of approaches it could take to structure its program, there are no hard-and-fast rules
governing the creation of a business plan. Still, experience shows that short plans are
better than long ones, not only because they are easier for investors to read and
understand but also because they force the entrepreneursto thnk in an orderly fashion.
Because new companies require significantcapitalization,the foundersmust now:

g e t (venture c a p i t a l ) ;

They do this by taking their business plan to venture capitalfirms, friends,and relatives
in an attempt to obtain funding. This process consists of rounds of courtship with
venture capitalists, during which the plan is refined, interviews are conducted to select
core personnel, and some ad hoc engineering conceptualizationis done to refine the
product design and marketing approach.Alliances to achieve additional funding may
4 Introduction

alsobe formed at this time.Despite all these efforts, of every hundred businessplans that
are submitted to a given venture capital company, only about thirty result in a first
meeting, ten receive a more detailed review, and less than one gets funding. In 1989,
roughly fifteen hundred companies were funded by venture capital firms.
If funding efforts are successful,the founder and his or her core personnel will:

start (new company) ;

This is also the point at which additionalpeople leave their jobs to form the nucleus of
the new firm.
In order for the fledgling organization to become fully operational, the founders
must now:

geL (space, people, product development

tools, UNIX license) ;

At this stage, the new company’s founders have to acquire the basic tools needed for
product development.In the early 1980s,systemscompaniespurchased a VAX,a copy
of UNIX, and a license to operate, develop, and sell it as the standard operating system.
In the 1990s, a collection of Sun workstations,Apple Macintoshes, or IBM-compatible
PCs is likely to substitute for the VAX.
Now that the new firm is in business, it proceeds to:

sell (product idea) ;design (product);

market&sell&produce ( p r o d u c t ) ;

The sequence of events-sell, design, and build-is very important. The product de-
velopment process starts with attempts to sell the product idea to potential customers.
Tlusprovidescriticalfeedbackforthe design (althoughitcanalsoproduceanunbuildable
product specification). The founding team typically spends between twelve and
twenty-fourmonthsdevelopinga product,oftenmakingsalesagreementswith traditional
companies that are unable to develop new products in a timely manner. The start-up
then begins to market, sell, and produce its product.
Marketing, sales, and production are extremely important, since these are the
activitiesthat determine the new company’s profitability,which is the subjectof the next
segment of our high-tech start-up ”program”:

while new company is not profitable then

wait; get (more $1 ;
sell (new company);

If the new venture does not at first achieve profitability, investors are asked to provide
 Introduction 5

additional funds, dilute the company, and wait patiently for success. If profitability is
still not achieved, the start-up must cease operations, merge with another firm, or be
acquired. Merger or acquisition (althoughon more favorable terms) are also possible
outcomes of the success scenario, which is discussed below.
With a hgh-growth (or simple) product and plan, the company may become
profitable after only a few quarters of sales. After the start-up has sustained its
profitability for several quarters (a “stair-step”pattern of revenue growth),it can either
remain in private hands, ”gopublic” with an initialpublic offering(IPO),or be acquired.
Although some firms do continue to be privately held for many years, this discussion
assumes that one of the start-up’s paramount goals is to gain substantial amounts of
additional operating capital-in which case, going public or being acquired is the
appropriate course of action.
Going public can wreak havoc with the company’soperations,sinceit demands the
full, ongoing attention of already-overworkedkey personnel. The focus of operations
temporarily shiftsfrom salesand serviceto the taskof auditingstrengthsand weaknesses.
Nevertheless, going public is financially beneficial, both for the firm and for its
employees and investors. The company gains capital with which to expand its opera-
tions, and the foundersmay realizesubstantialfinancialrewards. For example,if an idea
results in the founding of a hgh-tech business that subsequently becomes successful
and goes public (the odds of this happening are about six in a million), the dominant
founding entrepreneur can receive an average of $6.5 million (Nesheim,1988).
Although the rags-to-richesscenario of a start-up and its founders’prospering via
an IPO is attractive, this is not how the majority of new companies gain additional
capital. Instead, the most common method is for the start-up to be acquired by a
successfulfirm,usually in the same area of expertise,which enables the start-upto avoid
the trauma of going public and coming under public scrutiny.In 1989, for example, 149
companies(worth$2 billion) in the PC field were acquired,whereas only 18companies
(worth $300 million) went public via an IPO.
I strongly recommend staying private and independent as long as possible and
avoiding the inevitable urge to go public. Public investorsare rarely interested in a firm
and its technology, products, and market. Instead, most tend to be interested only in
stock appreciation.Public investors, in short, are not truly investingin a company; they
are merely renting one until a better opportunity comes along.
Now that the start-up has become successful, a number of possible scenarios
present themselves:

retire; wait; restart;

if entrepreneur wants to do it again thc.n
start (another high-information-
technology company) else
start (new venture capital company);
6 Introduction

How the start-up has obtained additional fmancing (i.e./by going public or by being
acquired) may affect what happens to the founder and/or to the company. In the case
of a firm that goes public, the founder may be exhausted from the effort that was
required to start the company and sustain it through the public offering, and very often
choosesto leave.In the case of a firm that is acquired,the founder may find it distressing
to no longer be his or her own boss and to (once again) be part of a large organization.
Furthermore, the new owners sometimes consider the founder’s job as CEO of the
acquired company to be redundant. Either way, his or her departure is likely.
A handful of founders have attempted to take extended vacations or retire at this
point in their lives, but they often find it impossible to wind down from the excitement
of the start-up process. They discover that they have become more enmeshed in the
creation effort than they originally thought. Such individualsmay reenter the start-up
arena as venture capitalists, ready to advise others, or they may choose to ”do it again,”
founding yet another start-up company.
Prospectiveinvestorsand employeesof firmsthat have recentlygone public or been
acquired should note that acompany’svalueoftenpeaksat this point.Although,ideally,
the enterprise has now achieved what might be termed steady-stateoperationsand has
reached the point of being able to developnew productsand sustain its profitability, the
departing founder(s) may have built an organization with little lasting value. The
company may be locked into a product architecturethat has no way of becoming self-
sustaining. A successful track record up to the point of IPO or acquisition, then, is no
guarantee of future success. Which brings us to the

Now that readers are familiar with the preceding simplified “program” for starting
a high-tech company, it is time to consider a more detailed scenario. The ”program”
shown in Figure 1-2not only expandsupon the simplified program but also divides the
creation of a start-up into five stages:

Stage I-Concept: The foundersdevelop an idea and createa plan for a companythat
willimp1ementthatidea.Theyeitherseekfundingforaseedstage to furthertestand
refine the idea, or they go directly to the product development stage.

Stage 11-Seed: The idea is refined, and a detailed plan for the company is created.

Stage 111-Product deuelopment: The product is developed and tested by users.

Stage IV-Market development: The product is sold, and the company becomes
profitable, thereby proving its viability.
 Introduction 7

Stage V-S teady state: Investorsand foundersachieveliquidity as the start-upeither

goes public, merges with another company, or continues its operations while
remaining in private hands.

The spark to start a new company

ii an idea for a technology/product/service company exists
and frustration is greater than reward in currentjob
pain
and greed Is greater than fear of start-up’s failure p i n
and financial and emotional support for start-up exist l&a

I. Concept stage:
exit (job); find (team);
get (spreadsheet tools); write (plan); find (investors);
-
if plan has low risk then
product development stage;

11. Seed stage:

exchange (stock in company for cash from friends and venture
capitalists);
recruit (superstar start-up team);
while company is on plan and money is in bank &
kgia verify (technology, product, market, business) ;
refine (plan);
-if seed stage is done and plan is still good then
product development stage;
end;
the following occurs if out of money or the plan is missed
if company is still viable and investors are willing Lhen
continue seed stage &
sell (company) or close (company);

111. Product development stage:

exchange (stock in company for cash from investors);
recruit (superstar development team);
get (space, development t o o l s , software licenses);”
while company is on plan and money is in bank
beain sell (product idea); specify (product);
develop (product); a market-driven product”
build (first product prototypes); via manufacturing
test (product, internally); alpha testing
manufacture (first products) ; manufacturing begins
test (product, externally); beta testing
announce (product);
market development stage;
end;
-
if company is still viable and investors are willing
continue product development stage
sell (company) or close (company);
(continued)
8 Introduction

(cont i m e d )
TV. Market development stage:
exchange (stock in company for cash from investors);
recruit (s-aperstarsales team);
while money is in bank &
sell (product); produce (product);
deliver (product);the business cycle
if company is profitable for six quarters L.La
exit and cash in; steady state
end;
if
- company is still viable and investors are willing L&XI
contin= market development stage eke
sell (company) or close (company);

V. Steady-state stage:
Exit and cash in: Company is sold to achieve liquidity.
sell (company for sales revenue X 20, to public or to
another company);
continue (company);
retire and return (entrepreneurs);
if entrepreneurs are not tired
start (next company) else
venture capital company;

end; completiori of start-up program

"In the early 1980s, the development tool was VAX, and the system plan was to design a product to
"beat VAX" with a UNIX-based system. In the late 1980s, the development tool was a Sun worksta-
tion, a Macintosh, or an IBM-compatible PC. The software license depended on the product: UNIX,
Macintosh, MS-DOS, or OS/2. In 1990, it's DOS with Windows and UNIX.
bFinisConner, CEO of Conner Peripherals, characterizes his market-driven product planning and
development philosophy as "sell, design, and build."

Figure 1-2. Detailed "Program" for Starting a High-Information-TechnologyCompany.

As the company grows, it proceeds according to a plan, which allows it to move

from one stage to the next by obtaining funds from its investors in exchange for stock.
During the course of each stage, the firm does one of two things:

Achieves its goals for the current stage and advances to the next stage

Misses its planned goals or runs out of money

In the second case, the firm becomes subject to the "if" statementat the end of each stage
of the start-up "program." If the investors believe the company is still viable and are
willing to proceed, then the company continueswith the stage.If not, it is sold or closed.
 Conclusion 9

This book focuses primarily on the first two stages, concept and seed, since these
form the foundationof any new enterprise.In both of these stages,the start-up’sviability
is determined by the interaction of as many as twelve different dimensions:

People-i.e., the CEO, team and culture, and board of directors (discussed in
Chapter 2)

Business plan (Chapter 3)

Cash, financeability,and control (Chapter 4)

Technology (Chapters 5 and 6), manufacturing (Chapter 7), and product

(Chapter 8)

Marketing and sales (Chapter 9)

In the chapters indicatedin the precedinglist, a set of rules (inthe form of questions)
is presented for each of the twelve dimensions.These rules can be used to test the new
company’s readiness to leave the concept stage and/or seed stage. In Chapter 10, the
rules for each dimension are summarized,and readerswill learn how to apply the Bell-
Mason Diagnostic to test a start-up’s ability to meet the requirements for long-range
success with respect to each dimension. Although, in the following chapters, the
application of the diagnosticis discussed in detail for only the concept and seed stages
(i.e., the hrst two, most critical stages), the diagnostic can, of course, be applied to an
organization throughout all five stages of its growth.

CONCLUSION

This chapter has introduced the two basic concepts involved in examining a start-
up: that start-ups are sequential in nature (i.e., a new firm passes through five
discrete stages) and that all aspects of a company are important and must be
considered in assessing a new venture’s health. The relationship between these two
basic concepts was illustrated in ”computer-program” format in Figure 1-2, which
presented a series of statements, grouped according to stage, with each statement
involving one or more of the twelve dimensions.
The first of the twelve dimensions,“people,” is also the most important, as readers
are about to learn in Chapter 2.
Chapter 2

THE PEOPLE

People, people, people.

-Arthur Rock

It is often said that the three most important factors in real estate are “location,
location, and location.” Likewise, the three most important factors in the formation
of start-up companies are ”people,people, and people,” because it is the people who
lead the firm and have ultimate responsibility for its success. The key personnel are
the chief executiveofficer (CEO)and those immediately adjacent to him or her in the
reporting structure-i.e., the board of directors above the CEO and the team of direct
reports below him or her. Although the board of directors has the ultimate fiduciary
responsibilityfor the company, it is the CEO who is responsible for leading the firm,
since the CEO leads the team members, who, in turn, lead the vital functions of
engineering, manufacturing, marketing, and sales.
The requirements for the board, the CEO, and the team change somewhat as a
company matures, and a person or group of people who may have been right for one
stage of a firm’s development may not be right for another stage. Each of the
following sections starts by presenting the time-independent general requirements
for a given position-beginning with the most important of these positions, that of
CEO-and then discusses possible flaws and more specific requirements, including
how these requirements may change between the concept stage and the seed stage.

10
 TheCEO 11

THE CEO: LEADER, COACH, MANAGER,

AND ”STANDARDS SETTER

The CEO sets all the standards for the company, including coaching, decision
making, delegation, effort, egalitarian behavior, energy, ethics, hiring, honesty,
leadership, management style, quality, thoroughness, and working style-Le., the
complete A-to-Z range of attributes that form ”the corporate culture.” The CEO, in
short, is the firm’s heart or ”clock,” which drives every event.
Academics, biographers, and autobiographers have written a great deal about
the personal characteristicsrequired to start a company and become its first CEO.
Silver (1985)believes that the typical entrepreneur is a happy, creative, insightful,
guilt-laden twenty-seven- to thirty-three-year-old who is a good communicator,
comes from a middle-class home with an absent father, had a deprived childhood,
is married or divorced, and can focus intensively for long periods of time.
It should be noted that wealth was not among the characteristicsjust specified.
Not only does Silver not require it, but both White (1977) and I believe that ”being
wealthy is a significanthandicap” to an entrepreneur because successisn’t absolutely
essential for wealthy people, and they are therefore not driven by an urgent need to
acquire and preserve cash. In the words of Jim Hammock, president of Silicon
Compilers (acquired by Mentor Graphics), “When we started up, the company was
all any of us had. We simply had to make it work. Often, fear of failure was our
strongest driving force.” It is possible to create the appropriate fear of failure in a
wealthy person, however, and thus overcome the ”handicap of wealth,” by having
that person invest a significant portion of his or her net worth in the new venture.
The CEO must have a very high energy level and be completely dedicated to the
company. Dedication means that the CEO should not be involved in more than one
or two outside organizations, since excessive outside involvement is irresponsible
and places his or her firm at significant risk. On the other hand, neither should the
CEO be overzealous-trying to do everything personally. Rather, the CEO must be
able to hire creatively, understand the responsibilities of every team member, and
delegate tasks appropriately. If the CEO is the founding entrepreneur, and is an
inventor or marketing visionary but not a manager, he or she may wish to delegate
a majority of the tasks through an intermediary manager-a chief operating officer
(COO).Under these circumstances, Silver (1985) advises hiring a manager who is
older and more formal, who has a great deal of energy and heart, and who is both
practical and thorough. Typically, a good manager for this position is a former
corporate achiever with a nonegocentricmind-set who became dissatisfied with his
or her environment.
Having both a COO and a CEO in a start-up involves a number of potential
dangers, however. This is essentiallya ”two-in-a-box”style of management, and the
12 ThePeople

COO and CEO may try to perform the same tasks, tripping over each other while
increasing costs and slowing decision making. Alternatively,the CEO may delegate
too much responsibility. Ideally, the CEO should rely upon the COO as one of the
chief members of the team to whom tasks can be delegated, but the CEO should
never delegate his or her primary responsibility, which is driving the company.
Another way to get into difficulty is to have a chief operating officer who
manages internal affairs while the CEO sells the company in various ways. Such an
arrangement stresses "selling" as a CEO's most important skill and thereby biases
the choice of a CEO, by limiting the field to candidates with a sales background.
Unfortunately, such individuals often find themselves incapable of hiring outside
the sales specialty and hence tend to populate the company with salespeople.
Although a CEO (and the rest of the team) should have some sales ability, the need
for such ability pales in comparison with the need for him or her to understand
finance, control, marketing, and products. Further, unlike a salesperson, who leads
and manages individuals, a CEO must create, lead, and manage teams of individuals.
In short, I believe that those involved in a start-up should think very hard before
selecting a salesperson or sales manager as a CEO.
Over time, I have concluded that the chief executive officer is typically the
weakest dimension of a start-up. The CEO holds a position of great influence, since
systems and controls for running the company smoothly are not yet in place.
Resource limitations compel the CEO to wear a number of hats, frequently in areas
where he or she has little expertise. One of the important hats is often that of
mediator,because intrateam disputes can have immediate(and possibly devastating)
bottom-line ramifications.The fledgling organization's inordinate dependency on
the CEO places a great deal of power and responsibility in this individual's hands-
perhapsmore than he or she has ever exercised.SomeCEOsget drunk on this power,
while others become frightened and paralyzed. Good CEOs are able to maintain a
certain measure of detachment and perspective and understand the need to drive
the organization.
The following list presents some key personal qualities exhibited by effective
CEOs. Readers are encouraged to rely on their own experience and intuition when
weighing them.

Intelligenceand energy: CEOs need intelligence so they can identify and prioritize
problems and set direction, and they need extraordinary stamina and com-
mitment because everyone in the company takes his or her lead from above.
When it comes to these two qualities, the higher a CEO's level of intelligenceand
energy, the better.

Integrity, quality, and working habits and environment: CEOs must be honest and
open in dealing with everyone, inside as well as outside the company. They
 TheCEO 13

must set a personal example that translates into both corporate and product
quality.

Openness: CEOs who encourage an ”open-door” policy, invite suggestions for

change and solutions from anyone, anywhere, and who are willing to openly
acknowledge their strengths and weaknesses tend to be good, honest leaders.
They have no ”hidden agendas” and demonstrate a realistic, appropriate pride
in their accomplishments. They usually get things done through the natural
processes of building interpersonal respect and recognizing competence.

Background: Good training and good role models, or mentors, are two of the
most common attributes of effective CEOs. The problem is that the great
companies don’t let their people escape. Thus, many of the available CEO
candidates may be the products of an inferior corporate background and
inferior professional role models. The best alternatives are often “virgin”
candidates with no preconceived company concept.

Team-building skills and ability to delegate: These attributes, which are actually
closely related to the team dimension, involve the CEOs personal ability to
create, motivate, and drive the team in a productive and organized way.

Ego and humility: Excessive ego or lack of ego can lead CEOs either to consis-
tently fail to delegate authority and responsibility,or to chronically overcommit
or undercommit to accomplish personal and company goals. CEOs must
therefore be able to restrain, but not eliminate, their personal and professional
pride. The accuracy of the CEOs’ assessment of the company’s (and their own)
strengths and weaknesses gives an indication of their true humility.

Just how critical is CEO selection?Dennis Gorman of Sevin Rosen found that
over 90 percent of the companies backed by his firm that went public were still
headed by the original CEO, whereas 25 percent of the companies that failed or were
floundering had retained the founding CEO.
In summary, James Swartz, past chairman of the National Venture Capital
Association, describes five attributes that a CEO needs to ”win a venture capitalist”:
leadership, vision, integrity, openness, and dedication.

CEO FLAWS

CEOs’ flaws are legendary, as countless newspaper and magazine articles have
chronicled with delight. Some CEOs have been victimized by technology’s moving
more slowly than they anticipated; others have met their fate at the hands of a fickle
buying market; still others have simply been losers. Unfortunately, the authors of
14 ThePeople

newspaper and magazine stories tend to simplify the issues by concentrating on

single events or single flaws and do not provide a holistic view. Book authors, in
contrast, have more time for research and more space to tell the story. As a result,
some books do provide a balanced picture of the right attributes, and many book-
length biographies and analyses present case studies illustrating effective CEO
performance.
The flaws summarized in the following subsections are, in effect, the reverse of
the virtues listed above. Since no one is perfect, the CEO is likely to exhibit at least
one of these flaws to some degree. How the start-up deals with a CEOs flaws is very
important, because a company that is weak in other dimensions may find these flaws
to be fatal. A flawless CEO is a rare phenomenon; however, Ken Olsen’ (Rifkinand
Harrar, 1988)comes as close to this ideal as any CEO with whom I have worked, and
his record of success is legendary.

Low Energy, Low Intelligence, andlor Low Integrity

The CEO may have either a low energy level (slow clock) or an inadequate time
commitment, low intelligence (what might, in computer lingo, be termed a slow
central processing unit, perhaps coupled with a small 640K memory), and/or
questionable ethics. This type of CEO tolerates nonegalitarian behavior, low quality
standards, poor work habits, and unrestrained company spending.
The criticalityof the CEO as the standards setter-the individual who establishes
the company’s clock-was discussed earlier. Almost all the dimensions encom-
passed in this flaw, ranging from intelligence and work habits to ethics, have come
up in ”judging” every CEO I know. A particularly annoying flaw for start-ups is the
CEO who treats the fledgling company as if it were a large, solidly established firm,
demanding all the perks. Individuals of this sort are readily identifiable, since they
insist on a large salary, absolutely must fly first class, require a carte blanche expense
account, and tend to be found quibbling over (or modifying) their original com-
pensation agreements with the company. Arrogance and greed drive such CEOs to
milk the very firms they were hired to nurture.
My own biases are clear: only become part of a venture led by a hardworking,
extremely intelligent, and highly ethical individual who knows how to establish a
dynamic, open company culture and can manage, lead, and sell.

. - -
1 Digital Equipment Corporation started in 1957 and ran well under Ken Olsen’s leadership until the
mid-l980s, when the advent of other forms of computing began to stall the company Product reve-
nues for 1990 declined from the previous year, and in the fourth quarter, the firm was unprofitable.
 TheCEO 15

Inability to Sustain the Cheerleader Role

The CEO may lack the stamina, energy, and ability to continually sell employees,
customers, and investors, throwing in the towel when the company fails to take off.
Given the brutal environment of a start-up, the CEO can never abdicate his or her job
as head cheerleader.

Inadequate Hiring Skills

The CEO may be unable either to make first-rate hires or to deal with the inevitable
hiring mistakes. Because this type of individual simply doesn’t know how to test for
and hire top-quality people, he or she often just hires former cronies, placing more
stock in allegiance than in competence. The company must continually seek and hire
only the best candidates. If the CEO is unable to accomplish this, then ”pygmy
hiring” sets in and the quality of the firm’s personnel enters a downward spiral.

Poor Managerial and Team-Building Skills

The CEO’s lack of managerial and team-building skills can manifest itself in
numerous ways. The company may operate in a state of continual chaos; the CEO
may reserve all control and decision making for himself or herself, thereby preventing
any of the subordinates from managing or developing; or the CEO may work all
issues one-on-one so that a team never has the opportunity to form and team
problem solving never occurs. This type of CEO may create either a ”political”
environment in which every decision hinges on the selling power of individual
personalities or a bureaucracy in which decisions take forever to be made.
The CEO who places a high value on ”being liked by everyone” will probably
create an environment in which staff-level decision making is impaired or futile. At
the other extreme is the tyrant who insists on taking and keeping control of every
area of the company personally, thereby impeding all progress. The CEO sometimes
does this overtly, by delivering imperial mandates at staff meetings; but he or she
can also achieve the same effect covertly, by allowing many issues to be left
unresolved. In the latter case, the CEO then avoids confrontation by “solving”these
issues outside of staff meetings, without buy-in from the parties who are most
affected.
Above all, the CEO has to understand the fundamentals of leadership and
management. He or she must be able to delegate, form a team, and get the team to
make extraordinary commitments. .
16 ThePeople

Inability to Build a Team or Keep a Team Together

It is sometimes possible to detect a lack of team quite readily. A venture capitalist I

know simply asks a direct question of a team member. If the CEO interrupts with an
answer, he suspects that the company is driven from the top down and lacks a viable
team.
When the founder and CEO is unable to keep the team together, he or she may
be ousted by a ”palace revolt” and replaced by a series of ill-conceived, board-
controlled actors and actions. For example, I know of a company that built real-time
laboratory computers based on the first 32-bit microprocessors. After two years,
when the firm was just beginning to reach its peak sales and was becoming
profitable, a ”palace revolt” prompted the board to replace the existing CEO. The
new CEO came from a very large computer company but was a sales-oriented
individual with no experience in the laboratory market area or in product devel-
opment. The organization subsequently declined to the point where it was forced to
merge with another floundering firm. Guess who the winner(s)were: (a) investors;
(b) the founding CEO; (c) the new CEO, who received a ”golden parachute”; (d) all
the founders and employees; (e) customers; (f) none of these.

Inability to Sell the Company to the Financial Community:

The ”Short-Socks Test”

A start-up may fail to secure funding for many reasons, not all of which are
necessarily relevant to the firm’s viability. The following is a case in point: After
visiting an entrepreneur, a New England venture capitalist commented to his
associate that the company wouldn’t be funded. “Why?“ asked the associate.
Replied the capitalist: “Because the president was wearing short socks.” Although
I’m sure that lots of California firms have obtained funding despite their founders’
wearing no socks at all, the basic principle still applies: when an entrepreneur is
initially seeking financing, first impressions really count-perhaps more than they
should.
Regardless of whether the precise reason for the CEO’s inability to sell the start-
up to the board and the investors is trivial (e.g., failing the ”short-socks test”) or
substantive (e.g.,not being a sufficiently persuasive advocate for the company), his
or her shortcoming will manifest itself through financing problems for the firm and
a lack of belief in and/or support for the CEO. This flaw really involves an inability
to manage the board and the investors. It is perhaps the most rapidly fatal flaw of
all those discussed, and its cost is quite simple: the CEO loses his or her job when an
impatient board finds it isn’t getting the response it believes it needs.
 TheCEO 17

CEO RULES

In some cases, a company is founded by an entrepreneur who has a technical or

marketing idea and who then serves as the acting CEO during the firm’s seed stage,
even though he or she may lack many of the required qualifications. This is a risky
way for a company to start out, because it may subsequently be forced to hire a new
CEO in order to reach more advanced stages of its development, and the transition
to the ”real” CEO can prove traumatic. Changing CEOs is similar to performing a
heart transplant: it takes a long time to find a compatible donor, the operation is
lengthy and complex, the body requires a long period of healing and adjustment
afterward, and there are no guarantees that the procedure will ultimately be
successful.Itwouldbemuchbetter tosearchfor anappropriateCEOfromtheoutset,
using the rules in this subsection as guidelines.

Does the CEO candidate possess the levels of intelligence, energy, ethics,
and quality that are required to establish the clock and culture for the
proposed company?

Although this rule can be stated explicitly, it is never really answered explicitly.
It is answered implicitly, however, by everyone-employees, investors, strategic
partners, or customers-who becomes associated with a particular start-up. Despite
its being wholly subjective, this rule tests the overall quality of a CEO candidate by
evaluating the individual as the prospective leader of the environment that he or she
proposes to create.
To satisfy this rule, the CEO candidate must provide solid evidence and
referencesthat testify to his or her past accomplishments.In particular,if a prospective
CEO has run another company and has led in the definition of its culture, then the
new firm is likely to be similar to his or her previous one. As the start-up ends the
seed stage, it will become increasingly clear to the employees, investors, strategic
partners, and customers-as well as to the CEOhimself or herself-whether the CEO
was well chosen.
A second, less subjective rule should also be applied to the concept stage
selection of a CEO:

Has the CEO demonstrated management, team-building, and leadership

ability involving product development, in a resource-constrainedenviron-
ment, and on a do-it-from-scratch (e.g., start-up)basis?

This rule really has three parts, since being able to manage, team-build, and lead
are all highly critical skills. Without managerial skills, the CEO will be unable to
18 ThePeople

establish any standards of commitment and follow-through.However, as discussed

earlier, the CEO could satisfy this aspect of the rule by delegating management tasks
to a COO, provided the company can afford the extra staff and there's a clear
understanding that the CEO is in charge. The second part of the rule tests whether
the CEO has experiencein technology and product development. The final part tests
his or her ability to operate with constrained resources.Ideally,the prospective CEO
will have gained that skill during a previous start-up, but a person who has begun
a small enterprise within a large company might be an alternative candidate, albeit
a risky one.

Can the CEO articulate and sell the company vision to attract the financing,
engineering, and other key talent needed for the (advanced or
predevelopment) seed stage?

The final rule for the concept stage evaluates the CEO's ability to act as a
salespersonin order to obtain seed stagefinancingand recruit outstanding employees
so that the seed plan can be carried out.

Does the CEO have extensive experience in management, and has he or she
demonstrated competence in product development, marketing, and sales by
adhering to the principal objectives of the seed plan?

This rule provides a simple test based on the CEO's most recent accomplishments
during the seed stage. Given the seed stage requirement of translating unique
technology into a product specification, it should be easy to determine whether the
CEO has in fact been successful in leading the company to this point.

Is the CEO a leader and team builder across departments, and can he or she
lead/manage the team and help attract key personnel at various phases of the
product development stage? This will be necessary in order for the company
to start building all the required functions.

This rule looks beyond accomplishments during the seed stage and examines
the likelihood that the CEO can continue to be an effective leader and manager
during the firm's future stages of growth. It is a rule that is often violated, because
many entrepreneurs do not have the time to receive management training (or to gain
its equivalent in terms of practicalexperience)before they begin running a company.
It is hard for an inexperienced CEO to manage a fledgling firm and get funding at
the same time. Michael Dell of Dell Computer and Bill Gates of Microsoft were
inexperienced CEOs who succeeded, but they did not have to obtain traditional
 The Team and Company Culture 19

funding, which is fortunate, since their youth and lack of experience might have
made it difficult.

Has the CEO been successful in attracting financing, recruiting key em-
ployees, and finding directors for the board?

The ultimate proof of the CEO's selling ability is whether key individuals have
signed up at the seed stage. There should be a "backlog" of people wanting to be
involved in the company.

Does the CEO have insight into the content, scheduling, and management
interdependencies of engineering and marketing in the early phases, and of
manufacturing and sales in the later phases?

In order for the CEO to build a team, he or she must understand the motivation
of the various functions and know how to get the team's members to work together
and resolve the conflicts that will inevitably arise. A good test of the CEO's skills in
this regard is whether both engineering and marketing have agreed to the product
specification by the end of the seed stage.

Can the CEO function actively as a company missionary in preselling,

negotiating strategicalliances,and lining up codevelopmentpartners during
the product development stage?

As noted earlier, a CEO must be able to sell the company to investors and the
financial community. Beyond that, however, he or she must also be able to sell to
customers and potential partners. In some cases, the "ideal" selling target is a
strategic partner who can invest in the new venture.

THE TEAM AND COMPANY CULTURE:

THE PARTS MUST FUNCTION AS A WHOLE

Lack of team is the number one company killer.

-John Shoch

Although teamwork is a critical aspect of an organization of any size, it is especially

important in a start-up.Teamworkis like a tree, with communicationas its trunkand
with mutual respect and recognition of common goals as its major root structures.
The leadership necessary to nurture teamwork starts with the CEO and his or her
20 ThePeople

direct reports, each of whom leads a team effort within a particular functionalgroup.
Although each direct report/group is measured independently, the groups must
realize that they form a team and that the results of the total team are what count.
There can be no such thing as saying ”Your end of the boat is sinking.”
Without integrated team effort, the company will be unable to understand and
resolve all the critical issues that cross organizational boundaries. Some of the issues
(financialcompensation,working environment, product quality) require the mutual
efforts of several groups, whereas others (product pricing, materials sourcing) can
be resolved by special pairwise relationships between groups.
Table 2-1 lists some crucial tasks that call for high levels of formal cooperation
and coordination.
To achieve the level of teamwork required to form and grow a successful
company, it is important that the top-level team (direct reports to the CEO) consist
of high-quality individuals with measurable experience and expertise. The head of
the start-up’s engineering department must have proven expertise in the company’s
technology/product domain; in addition, he or she must be able to perform a
function,such as design or analysis of some portion of the design. The top-levelteam
must also be “do”-oriented rather than ”management”-oriented. Each member
must be able to ”play” several positions on the team that reports to him or her rather
than just managing the team. This requirement implies specifickinds of competence
and serves to ensure that:

Members of the top-level team have an appropriate level of competence,ruling

out bureaucrats who come from large companies and possess the necessary
credentials on paper but often lack actual competence

The department head really knows what’s going on in the department, since he
or she functions as an active participant instead of just serving as the ”boss”

The organization is lean right from the start, since it does not have the separate
line (brawn) and staff (brain) components characteristic of many large, ”fat”
companies

A top-level team that passes these tests demonstrates competence, and compe-
tence is the basis for respect. Respect among the collected heads of the various
groups will ensure that they function as an integrated team rather than as a collection
of egocentric or warring individuals.
Even though the team operates in an integrated manner, each of its members still
has his or her own contributions to make. The measure of a team’s success is how the
contributions that its members make through their individual roles combine to
produce an overall result that is greater than the sum of the separate contributions,
 The Team and Company Culture 21

Table 2-1. Tasks Requiring Teamwork.

Task lnvolved Organizatioizs
Define the product for customers Engineering/marketing
Manufacture the product Engineering/manufacturing
Control the order-to-productflow Sales/manufacturing
Provide marketing information and Marketing / sales
establish order flow
Resolve customer problems Service/ manufacturing/
engineering

Meet corporate and departmental All departments/financial

operating and financial objectives organization
Maintain a commitment to All departments
corporate quality

due to the synergistic effect of teamwork. Table 2-2 summarizes the unique roles
played by various individuals as members of successful teams in some well-known
start-ups.

RESPECT FOR EMPLOYEES A N D THEIR PERSONAL TIME

The new company’s attitudes about how people will be treated begin to develop
during the seed stage. One of the most important and visible of these attitudes
involves the work ethic, as embodied in the firm’s working hours. A start-up must
strike an appropriate balance so that participants can have a life beyond the firm. The
successful start-up is often staffed with twenty-five- to thirty-five-year-oldswhose
families, including young children, can’t understand why they never see their
parents. It is unreasonable to establish a company culture in which, from the outset,
employees are routinely expected to work over eighty hours during six- or seven-
day weeks. One reason why a firm should avoid overscheduling its employees is
that it will have no slack-nothing to fall back on when the inevitable real crises
arise. However, the main reason for avoiding overscheduling is that burnout can
occur when employees work at such a pace for two to three years.
Hundred-hour weeks are inevitably required in even the best-managed start-
ups, but they should be the exceptions. In many new companies,staff members find
themselves working at least part-time on Saturdays, and it is not uncommon to hear
22 ThePeople

Table 2-2. Roles of Key Individuals in Several Well-Known Start-ups.

~ ~ ~~

Company Person Roles

Apple Jobs Founder, driving entrepreneur, and product

visionary
Wozniak Founding engineer and product designer
Markula Cofounder and source of financing and business
expertise
McKenna PR, unofficial member of executive staff, and
board of directors

Rock Funding and board of directors

Scott First president

Microsoft Gates Founding technical leader and visionary

Allen Technical Cofounder

Balmer Engineering operations
Shirley Business, marketing, and operations
Apollo Poduska Founder and company leader

Nelson Product visionary

Greata Product design

Spector First president until steady state
Vanderslice Second president, bureaucrat, sold floundering
company to Hewlett-Packard
Intel Grove Operations
Moore Overall visionary
Noyce Visionary and external spokesperson
Lotus Kapor Founding president and product visionary
Manzi Second president, during steady-state growth
Sun Khosla Founding entrepreneur and first president
Bechtolscheim Hardware product designer

JOY Software product designer and UNIX visionary

McNealy Manufacturing, with transition to president

Lacroute Operational management

 The Team and Company Culture 23

investors remarking about the number of cars in the firm’s parking lot on evenings
and weekends. In short, the start-up has a responsibility to establish reasonable
expectations with regard to work load and to clearly communicate those expecta-
tions to job candidates before they join the organization.

RESPECT FOR THE INVESTORS’ CASH

AND THE COMMITMENT TO PROFITABILITY

The firm’s attitudes about spending are another key part of its culture. Ideally, the
start-up should have a virtual reverence for cash, minimize spending (this includes
keeping salaries down), and maintain a clear focus on profitability.Investors respect
a new company that borders on being miserly. In contrast, they worry about a
company whose employees rake in high salaries and fill the parking lot with
expensive cars when the venture is not yet profitable.
I recently visited a chronically unprofitable company whose employees have
created a culture in which profit is disdained as if it were an unethical concept. The
organization, staffed with many talented artisans, came from a government-funded
research laboratory and now builds creative animation software, which it must sell
in order to survive. This firm must ultimately change if it hopes to remain viable,
since even the most gullibleinvestors reach the point where their patience wears thin
and their purse snaps shut.

TEAM FLAWS

Because a team can be undermined by almost anyone on it, the responsibility for a
team’s success lies with every one of its members.Whether or not those involved can
operate as a team depends on such factors as the extent to which they share a vision
of how to build a lasting company, the competenceof the individual team members,
the team members’ respect for one another, and the CEO’s leadership skills. Since
a discussion of all possible team flaws could fill an entire book, this subsection
describes only some of the most common ones.

A Mercenary Team

The problem with building a team using entrepreneurial mercenaries is that the
members’ motivation will be questionable. If the team’s aim is simply to make a
quick buck rather than to develop a unique technology and form a lasting company,
difficultieswill soon ensue. A similar flaw, forming a company with a questionable
motivation, is discussed in Chapter 3, ”The Business Plan.”
24 ThePeople

Conflicting Egos and Lack of Respect

In some cases, certain key participants, including the CEO, may be so egocentric that
the CEO cannot form them into a viable team. The first test of a group’s ability to
work effectively together as a team is when it has to prepare the company’sbusiness
plan and make trade-offs among various functions. If there is a problem with
conflicting egos and lack of respect, the team may simply fall apart at the concept
stage or the seed stage because of its members’ inability to get along while preparing
the plan. Alternatively,the team may break up during a later stage of the company’s
life, when the stakes are much clearer and the pressure for teamwork is even greater.
Lack of mutual respect is usually at the root of this flaw, although the problem
may give the outward appearance of ego conflict between the involved individuals.
It is common in high-tech organizations to find a lack of respect between marketing
and engineering personnel, which is almost certain to prevent effective teamwork.
Every possible effort must be made to overcome this flaw because although mutual
love is not a criterion for team membership, mutual respect certainly is.

TEAM RULES

The team is more than the sum of the founders or those who report to the CEO.
Although the CEO is ultimately responsible for the company culture, the entire team
must embody it. Team members must help define and promulgate the culture
throughout the firm by their actions. The likelihood of forming a successful team can
be analyzed by applying the rules presented in this subsection.

Do the two or three people current1y”onboard” at the concept stage have the
critical experience and expertise in technology/product/marketdevelopment?

The first rule tests whether the team has the individual and collectiveprofessional
capabilities to start up. Unless each member exhibits an outstanding level of
professionalism, the company does not have a solid foundation, and the lack of
competence and mutual respect is likely to prevent the formation of a team.

Is there evidence that the founders can function as a team? Tests: Have they
worked together productively for three to six months? Do they respect one
another?

The second rule checks for what might be termed “teamness” at the concept
stage. Without solid professional competenceon the part of each member, the team
will not function cooperatively to solve the inevitable conflicts, such as disagreements
 The Team and Company Culture 25

between engineering and marketing over the product requirements. This rule also
tests how compatible and comfortable the individuals are with one another in terms
of whether they can engage in joint problem solving and trust one other to manage
their respective areas. The simple tests include the team members' having worked
with one another long enough to be certain that they can build a company together.
Some investors insist on the team's having worked together either in a previous job
or for at least six months on the current start-up.

Does the team's orientation reflect an appropriate balance between "doing"

and "managing" that will enable it to begin establishing an action-oriented
culture? Tests: Can each of the top-level team's members "play" one or more
positions on his or her team as opposed to just managing a team of players?
Has the team managed comparable undertakings before?

This rule requires each member to function both as an individual contributor

and as a manager. Unlike managers in large, established companies, managers in
start-up firms invariably spend significant amounts of their time personally per-
forming their department's function, so they should be technically capable. On the
other hand, they should also possess managerial skills, since it is hoped that the
company will ultimately grow to the point where they will function primarily as
managers. It is, of course, difficult to find technological creativity and sound
managerial ability in the sameperson. Whenever technologically creative individuals
discover that they are weak in management, their first priority should be to hire their
own boss.

Do the reputations of the concept stage team serve to attract a first-rate

engineering team along with the critical marketing resources necessary to
achieve seed stage and product development stage objectives?

The team must have the individual and combined reputations (in terms of skill,
charisma, etc.) that will enable them to hire the critical people who will actually form
and carry out the company's main functions.

By the end of the seed stage, are the core leaders for the technology
development, product development, critical-process manufacturing, and
marketing functions on board? Are they operating as an integrated team of
six to eight people?

This rule, which provides yet another assessment of team formation, is tested
continuously during the seed stage, when the team members have an opportunity
26 ThePeople

to work together for several months-a vital step in team building. It is extremely
important that the founders be able to function as a team. If they show mutual
respect and the CEO is a good leader, chances are they will form a successful team.

By the end of the seed stage, have hiring criteria been established? Is a
systematic recruitment method in place?

Although each of the functions is responsible for recruiting in its respectivearea,

having companywide standards is also important to ensure that the first employees
are operating according to a single set of principles in establishing the company
culture. In start-ups, it is very easy to erect arbitrary walls and create different classes
of corporate “citizens” based on the way individuals are rewarded by various man-
agers. No matter how hard a firm may try, salary and stock ownership are likely to
become widely known. Although egalitarianism is not mandatory in order for a
start-up to be successful, rewarding on the basis of skill makes for a happier
environment.

By the end of the seed stage, if innovative manufacturing processes are

required (such as in semiconductor or disk manufacturing), is an experi-
enced manufacturing leader with a core team of functional specialists on
board?

If the company must undertake a manufacturing-intensive development pro-

cess, then the manufacturing leader must be part of the key hiring and team-building
effort right from the start.

By the end of the seed stage, have team members defined their desired
corporate culture? Is it compatible with what can reasonably be expected,
both from the company’s people and in terms of the overall professional
working environment in the firm’s geographic area?

All companies attempt to create a corporate culture that is uniquely their own.
The two key aspects of culture that must be defined at the outset are how the
company will treat its employees and how it will manage cash (the ever-present
symbol of its investors).
Interested readers can find many books and articles that discuss the culture-
formation process and/or analyze the culture of specific companies. Deal and
Kennedy (1982), for instance, have described various aspects of corporate culture,
including the case of Tandem Computers, which has the highest regard for its
 The Board of Directors 27

employees and is well known for its creative, healthy environment and its nearly
unique culture. Rogers and Larsen (1984) have described the culture of Silicon
Valley, and their work is required reading for anyone starting a venture there. And
finally, In Search of Excellence (Peters and Waterman, 1982) is the best-known book
on the subject.

THE BOARD OF DIRECTORS:

REVIEWERS, COUNSELORS, AND COMPANY
MISSIONARIES

The board of directors has the ultimate fiduciary responsibility in a company and
thus the ultimate responsibility for selectingthe CEO. However, once the board has
chosen a CEO, its members should function only as reviewers and counselors rather
than trying to run the CEO’s company for him or her. The only time the board
collectively,or its members individually, should play an active role in the firm’s day-
to-day operations is during those rare periods when the position of CEO is vacant.
Arthur Spinner of HambroInternationalVenture Fund summarizes the relationship
between the CEO and the board like this: ”If you are a venture capitalist [on a board]
and you want a company to run, go start one yourself.”
Spinner also cautions, “If you are an entrepreneur and you need direction rather
than support, you should not be running a company; you should be working with
one.” However, at various stages of the start-up’s development, the CEO may have
occasion to call on the board for review and counsel. Assistance may be required
initially in obtaining financing and later in taking a company public. Advice may be
needed in such areas as product and market development or selling to key customers.
The wisdom of experience may be useful in dealing with control and operational
problems. In each case, the board may provide its advice and counselby asking hard
questions and may help the firm achieve a more realistic perspective by offering an
alternative point of view.
Choosing board members is a critical process, because some may become
directors for life, and each must be considered a vital part of the company.
Unfortunately, the compositionof a board is frequently linked to financing, because
many venture capital firms make funding contingent on their being granted a board
position. In such cases, the member is often unable to make any contribution beyond
cash. Selectingboard members based on their ability to comeup with money or work
harmoniously with the CEO is usually a bad idea; rather, board members should be
selected based on the expertise that they can contribute. Even then, it will be rare for
a board member to have a broad range of applicable expertise unless he or she has
run a similar organization. I believe that a start-up should avoid choosing board
28 ThePeople

members who have not participated in the operation of a company or who possess
only a single area of expertise, such as the ability to raise money (unless it is
unquestionably clear that they can bring in cash easily).
Cautions have also been expressed about board members whose sole area of
expertise is the law. According to Gladstone (1988), many venture capitalists feel
that ”practicing lawyers make poor directors of small businesses” because ”busi-
nessmen. . .will help reach a consensus. . . [whereas]lawyers do not bring harmony
to the boardroom.”
A homogeneous board should be avoided, since this type of board is unlikely
to have the perspectives that a new company needs in such diverse areas as
operations, finance, technology, marketing, and consulting.A start-up whose board
consists of six near-clones is a recipe for disaster, because each member has the same
limited outlook. In fact, Spinner even argues that it is helpful for a board to have at
least one ”renegade of sorts who will consistently play devil’s advocate.”
In contrast, a heterogeneous board is the ideal (although heterogeneity should
not be carried to the point where board members cannot work together harmoni-
ously or communicate effectively). Such a board will find it easier to engage in a
variety of activities, ranging from simply serving as a support structure to shaping
external perceptions of the company (as Ben Rosen did for Compaq and Lotus). It
might also be useful to enlist members who have experience in working with
troubled firms and increasing their valuation.
The start-up should select board members who can spend the time necessary to
learn about the company’s business, its products, its competitors, and its customers.
They should understand the business well enough to detect danger signs and
recognize opportunities. Thus, people with time to do the job right may be much
more valuable than well-known individuals who already sit on a dozen or so other
boards.
When selecting board members, quantity should be considered in addition to
quality. Rosenstein et al. (1989) did a study of 162 start-ups in the northern
California, Boston, and central Texas areas, which revealed that board size tends to
increase as a company progresses from stage to stage in its growth process. (See
Table 2-3.)
Of the 162companies studied, the average board had 5.6 members, of whom 1.7
were internal members, 2.4 were venture capital principals, 1.2were venture capital
staff, and 1.8 had various other backgrounds. As companies grow, so do their
boards, and large, established firms have a mean board size of 13 persons.
The rather large representation of venture capital people on the boards surveyed
may be cause for concern, given the caution voiced earlier. However, it should be
remembered that the caution was against selecting board members exclusively as
sources of cash. If the company can find venture capitalists who have demonstrable
 The Board of Directors 29

Table 2-3. Board Size Versus Growth Stage.

Stage Average Board Size Standard Deviation
Seed 3.7 0.50

Start-up 5.0 1.24

Financing rounds 1,2, and 3 6.0 1.50

Financing round 4 6.0 1.40

expertise, they can make a valid contribution to the board. For example, in addition
to providing financing, these individuals can serve the firm in such capacities as the
following:
Developing the firm’s original strategy Monitoringoperations
Acting as a sounding board Monitoring financial performance
Recruiting and/or replacing the CEO Evaluating market plans

Recruiting (other than the CEO) Establishing customer contacts

Securing debt financing Developing new strategy
Securing equity (outside of venture channels) Serving as an interface with vendors

Serving as an interface with investor groups Assisting with crises

The value added by venture capitalists in performing these functions (as

perceived by the CEOs) was also tabulated in the Rosenstein et al. study. The study
concluded that venture capitalist board members made worthwhile (but not out-
standing) contributions, with the greatest contributions being made in the earlier
stages of company development. Also, no correlation was found between how well
the firm was doing and the CEOs assessment of its board, although the ordering of
the perceived value of each function did change slightly.Other functions performed
by the board were listed, too (evaluatingproduct/market opportunities, formulating
marketing plans, developing compensation plans, and assisting in the initial public
offering [IPO]),but these were deemed to be of negligible help. Steve Coit of Merrill,
Pickard, Anderson, and Eyre suggests that the venture capitalists on the board are
really the vice presidents in charge of financing and the IPO.
In order to maximize the board’s usefulness, the CEO must know how to
manage the board. For instance, the CEO should always raise issues rather than
adopting a defensive position. He or she should take care to meet the board’s
30 ThePeople

expectations,which means exceeding the requirements of the plan. Being prepared,

especially for the very first meeting, is essential. The agenda should include
information about progress and a summary of key issues that need to be dealt with.
In addressing key issues, the CEO should propose a plan for review as opposed to
asking for advice. The CEO who asks for advice will get it, and the CEO who does
so too often will find that the board or one of the directors is running the company.
Board meetings should be conducted in an atmosphere of openness. Both
Spinner and I believe that the company's vice presidents should attend board
meetings to make them aware of the boards views on various issues and to give the
board insight into the company's management team, one of its most important
assets. In contrast, CEOs who guard access to the organization are likely to be either
hiding something or insecure. In exchange for the CEO's policy of openness, the
board should deal with the CEO fairly and honestly, without wasting his or her time
on petty matters. Board members must realize that the CEO's time is a precious
resource, which they should conserve.

BOARD FLAWS

Individual competenceis at the root of having a great board of directors,just as it was

a key factor in having a great team. Not utilizing a competent board is merely a lost
opportunity, but certainly not a fatal flaw. The most serious flaw in this dimension
is simply having board members who are unable to contribute to the company,
either because they lack an understanding of the industry or because they possess
no knowledge and have only an ordinary level of intelligence. (The inexperienced
venture capitalist usually falls into the latter category.)
One of the CEO's most important jobs is to keep the board appropriately
informed and involved in the firm. Thus, the company must have a relatively
competent and cohesive board of a manageable size (about six or fewer members).
When the company goes out of control by missing its plan and board members are
surprised, the board oftentimes becomes involved in day-to-day operations.
The balance of this subsection describes typical board-related problems that
every start-up must guard against.

An Investor-Heavy Board with No Industry Experience

A board can have a very negative effect on productivity if it demands that the
company conduct its operations in a way that pleases the board instead of in a way
that will help the firm become a successful provider of goods or services. An
especially naive board composed of individuals who have had scant operational
 The Board of Directors 31

responsibility or who have a very limited understanding of the industry is likely to

have a net negative effect by creating “make-work.”One such board that I know of
contains a member who has no product, market, or technology experience and is
unable to make any valid contribution. Rather than having this individual tutored
“off-line,” roughly 30 percent of the board meetings are spent in his education.

A Board That Runs the Company

As noted above, when a board finds itself surprised by missed plans or faced with
operational uncertainty, it may get involved in the day-to-day management of the
firm, usurping the functions of the CEO and his or her team. A board that exhibits
this flaw is the riskiest type of board for the CEO to face, because it is just a step away
from firing the CEO.

No External Product/Market Review

Although the company’sproduct/ technology should routinely be subject to outside

review as the start-up develops its business, this may not be occurring because of
such factorsas (1)an uninformed or inexperiencedboard, (2)the lack of a technology
advisoryboard (TAB)or customer advisoryboard (CAB),or (3)operationalnegligence
on the part of the CEO and his or her team.
Every company needs an appropriate review mechanism to help direct its
efforts. In the case of an established firm, customers automatically provide such a
review through the marketplace. In contrast, a start-up is like a newly launched
missile, in that it must first be aimed in the right general direction and its trajectory
must then be continually corrected in midcourse if it is to reach the intended
destination. If board meetings are held only sporadically and communication with
the board is erratic and ad hoc, the board is typically out of control.

BOARD OF DIRECTORS RULES

A formalboard of directors is usually established with the first round of investment.

Although investors will naturally want to make sure the company is in control and
help it achieve its goals, granting board membership to inexperienced investors (or
to any other inexperienced individuals) won’t help the firm in the long run.
During the seed stage, the board should be structured to review the start-up’s
product and market plans in order to provide advice that will ensure the birth of a
healthy company.A customer or technical advisory board should alsobe established
during this stage.
32 The People

The following are the key questions that the start-up must address in setting up
its board of directors and its CAB or TAB.

Have board members with expertise in the key strategic areas outlined in the
business plan been identified to serve during the seed stage and later stages?

Although it is inappropriate to have a full-scale board of directors during the

concept stage, the company’sfounders should have some idea of whom they would
like and should have approached these individuals as the funding is finalized.
During the seed stage, the board will no doubt be composed of the two or three
founders and one or two investors. Since the goal of the seed stage is to reduce risk
and plan the start-up, it is worth having this critical sounding board to weigh ideas
about the start-up’s future direction.
During the concept stage, formal technology advisory and customer advisory
boards are probably inappropriate. However, if the company is entering an area
where the technological risks are especially high or where certain critical strategic
partnerships must be formed as part of the start-up process, it is prudent for it to be
working with a small group of key outsiders who will ultimately advise and assist
the firm during the seed stage and later stages.

Is a technology and/or customer advisory board in place by the end of the

seed stage?

This rule about having a functioning technology and/or customer advisory

board by the end of the seed stage is related to the preceding rule about having
identified potential board members with expertise in key strategic areas. It is
becoming increasingly common for start-ups to have a TAB and /or a CAB composed
of experts who understand the technology and advise the company on the formation
of the development team, reviewing the status of the technology and the competi-
tiveness of the proposed products. I recommend a single board composed of both
builders and users that meets regularly and whose members play an active role in
advising the firm, including serving as paid consultants.
The CEO must attend TAB meetings because they perform a critical review
function and provide feedback that the start-up may not get in any other fashion,
since its potential customers are often unwilling to tell the company’s marketing
staff the truth about their products. Also, the market input may get garbled as it
passes through various individuals who are grinding their own axes and who may
be unable to communicate effectively with engineering. A TAB should be free to
conduct its critical review of the company’s technical and applications directions
 Conclusion 33

without restrictionsthat may hamper its effectiveness.No topic should be off-limits,

including how the firm designs products, to whom it sells, or how it conducts its
business.

By the end of the seed stage, does the board include members who have
appropriate operational experience related to product and market develop-
ment in addition to the investor representatives?

At the seed stage, the board is likely to be overstaffedwith investors whose only
function is to keep an eye on their money. An ideal board would contain no more
than two investors, the CEO, and one or two outsiders. The two investors should
have previous operational experience in related businesses. The outsiders should
have experience in the product, service, or market area and should have invested
enough through sweat or equity to ensure that they are involved and concerned.
In 1990, most venture capital companies are staffed with people who have had
successful operational experience.This reflects a change in the composition of these
firms that occurred in response to the often-expressed criticism that they were
staffed with fresh MBAs who had no previous experience in operations or in the
industry. Although being lucky in a few previous deals is a necessary prerequisite,
it is not in itself a sufficient qualification.

CONCLUSION
The CEO of a new start-up was lamenting to his board about the difficulty of hiring.
A wise venture capitalist advised: “It’s not only hard; it’s your only job, because if
you are successful, everything else is easy.” The top-level people-the CEO, the
team responsible for carrying out the major functions, and the board of directors-
constitute the start-up’s three most important dimensions.
The CEO establishes the standards for the company and serves as its team
leader. The vice presidents for engineering,manufacturing, marketing, and sales are
the ”CEOs” for their respective functions. This top level of management must
operate as an integrated team and ”drive” the organization to achieve its business
plan and establish a healthy company culture. The CEO reports to the board of
directors, where the ultimate fiduciary responsibility for the venture rests. In the
ideal firm, the board merely helps and advises the CEO and company rather than
participating actively in the start-up’s management.
Chapter 3

THE BUSINESS PLAN:

A ROAD MAP
AND A SCORECARD
FOR THE FUTURE

The dream is what I look for more than anything else.

-Matsuda-san, Kubota Limited

Investors usually take the advice given in Chapter 2 and study the people associated
with a proposed company very carefully before making a commitment, since they
realize that a great team with a great product can recover from substantial adversity,
including the setbackscaused by a faulty business plan. This is not to say that a great
team and a great product don’t need a great business plan, however, because the
plan serves both as a road map for guiding the company’s current operations and
as a scorecard for subsequently determining how well those operations met their
objectives.
The business plan serves many critical purposes. It is:

A set of guidelines for operating the company

The standard of record against which the firm expects its results to be measured

A sales brochure directed at potential investors (although the downside and the
risks the company faces must also be covered)

A place where the founders can describe their vision for the firm

34
 The Business Plan 35

Because of the unlimited range of technologies and market approaches that a

start-up can employ in creating a market and a company, there are no hard-and-fast
rules governing the size, creation, and contents of a business plan. Experience
shows, however, that short plans (ten to thirty-five pages) are better than long plans.
Short plans are easier for potential investors to read and comprehend, since
significant points are made quickly and succinctly and can be readily understood
within the context of the overall plan. Furthermore, because it is quite difficult to
create a good short plan, the process of doing so forces the entrepreneur and his or
her team to think in an organized way. Entrepreneurs who are unable to make a case
for their proposed new venture within a handful of pages need to do more
homework. Finally, unless a plan is short, it cannot be easily referenced and updated
as a working document.
Thebusiness plans written by Bill Poduska-a founder of Prime, Apollo, Stellar,
and Stardent-have all been short and successful. In 1983, he offered the following
format for a successful business plan containing no more than ten pages:

Summa y (one page).

Market briej Who will buy and why-characterized as a new or existing product
type,foraneworexistingmarket. (Poduskafavorsanewproductforanexisting
market.)

Product briej The what, why, and how of building the product.
People: The who of building the product, the rule being to use only grade-A,
experienced individuals.

Financial projections: Both a statement of a practical strategy that can yield high,
yet realizable returns and a tool that can be used as the operational yardstick.

In many ways, the ability of a CEOand his or her top-level group to writeagood
business plan is the first test of their ability to function as a team and to run their
proposed company successfully. If a firm’s founding CEO can’t understand, build,
and operate the financial model for the company’s business, he or she should not be
the CEO. If the team has trouble writing a simple business plan, which is the first step
in running a business, then it’s quite likely they won’t be able to make any plans, and
they should give up the idea of starting a company until they get their act together
(which may mean forming a different team).
Despite the importance of a good business plan, a few companieshave managed
to become successful without the benefit of such a plan. For example, Gateway
36 The Business Plan

(discussed in Chapter 11)succeeded with no written plan. In some circumstances,

individual entrepreneurs can also get away without a business plan. I recently
advised an engineer who had a working product prototype to simply make a data
sheet, price the product at four times his cost, and then sell a few to see how users
like the product. If the results seem promising, he should then get a partner who can
handle the business aspects of founding a company.
Early in the planning process (i.e., at the concept stage), the company needs a
financial model that makes sense, showing how it can become profitable and stay
profitable.This model of the profit and loss statement, balance sheet, and cash flow
is just as important as the plan for designing and selling the product. Founders
should ”hang it up” right at this point if they see no way to create a viable model
based on reasonable assumptions about costs, prices, and market sizes, because the
initial business model is probably the most optimistic one the company will ever
have.
Surprisingly few variables drive the financial model. The key ones are:

Fixed assets and overhead costs, such as rent, telephones, and equipment

Variable costs based on head count, together with associated overhead (e.g.,
equipment, insurance, travel), for the fixed components of the organization-
research and development, manufacturing, marketing, sales, and administra-
tion

Variable costs for manufacturing the product, including work in progress and
inventory

Average sellingprice, salesproductivity,order-gestationtime, accounts payable,

and cost per salesperson

Requirements for cash to fuel the enterprise

THE COMPANY VISION

The company’s vision of its future is the most import part of the business plan.
Without a dream, the firm is unlikely to excite either itself or potential investors. It
should be possible to state the vision with equal facility in a single sentence, a
paragraph, or a slide show; the vision should be articulated in the plan; and finally,
it should be embodied in the product or demo. At each level, more should be
revealed so that investors, customers, and the press maintain their interest in the
product and want to see even more. For example, Ardent has a demo of its graphics
 Business Plan Formats 37

supercomputer that displays a simulated American flag waving in the breeze. The
resolution and clarity of the demo’s graphics prove that the supercomputer is fully
capable of delivering the 100 million floating point operations per second that the
simulation requires.
Digital’s “VAX strategy,” which guided the company throughout the 1980s,is
an example of a simple, yet powerful vision offered by an established firm:

Provide a set of homogeneous, distributed-computing-system products based

on the VAX-11 so that a user can interface, store information, and compute,
without reprogramming or extra work from the following computer sizes and
styles:

via [a cluster of] large, central (mainframe) computers or networks;

at local, shared departmental/group/team (mini) computers [and evolving

to PC clusters];

with interfaces to other manufacturers and industry standard information

processing systems; and

all interconnected via the local area Network Interconnect [Ethernet] in a

single area, with the ability of interconnecting the Local Area Networks
(LANs) to form Campus Area and Wide Area Networks.

The essence of the strategy was described in just a single page, and the entire
document (including the rationale and details) was only seven pages long. The
detailed plan was updated annually to reflect the tactics needed to respond to
changes in the marketplace and technology. The plan’s simplicity enabled over a
hundred thousand employees and customers to understand and support the
company’s effort.

BUSINESS PLAN FORMATS

All recommendedformatsfor businessplans include Bill Poduska’sfive components,
listed earlier in the chapter: summary, market, product, people, and financial
projections.
Beyond these basic components, various authors recommend specific options
and structures. Gladstone (1988), for example, presents a strong argument for
starting with a capsule presentation (i.e.,a summary),followed by a section entitled
“The Business and Its Future.” This section is probably the most extensive, because
38 The Business Plan

it covers such topics as the nature of the business, history and future, uniqueness,
product/service, customers, industry and market, competition,marketing, produc-
tion, labor force and employees,subcontractors,equipment, property and facilities,
patents and trademarks, research and development, litigation, government regu-
lation, conflicts of interest, backlog, insurance, taxes, corporate structure, and
detailed rksumks. Gladstone also recommends separate sections describing the
financing, risk factors, return on investment, and exit (how the investors obtain
liquidity).
White (1977) presents a number of heuristics about generating a quality plan
that will pass seven hurdles in the funding, from initial evaluation to financial
evaluation and final negotiation. He recommends that the plan include sections on
the history of the start-up, its manufacturing methods, quality assurance and
reliability, money-leveraging strategies, proposed distribution of ownership, and
founders’ stock incentives. He also suggests an extensive set of appendixes that
examine how the proposed company will be managed, including the use of
management by objectives.
Nesheim (1988) proposes two more sections (in addition to Poduska’s five)-
one dealing specifically with strategy and milestones and the other dealing with
operations, including engineering, manufacturing, finance, and administration.
BizPlanBuilder (JIAN, 1988) is a ten-point format that can be run on a PC or a
Macintosh. It provides a plan outline that anyone can build on directly simply by
editing the plan file, filling in the answers to critical questions, and completing a
spreadsheet. BizPlanBuilder’s ten points supplement Poduska’s format by adding
situation audit, objectives, and manufacturing sections; separating the topic of
marketing into two subtopics: analysis and strategy; and ending with a summary.
Although BizPlanBuilderis more suitable for lower-tech companies, its usefulness
can be expanded by adding a product development section. With this tool and
pruning, an author can develop a twenty-five- to fifty-page plan (not including
financials and appendixes). BizPlanBuilder can also be used to check any business
plan and make sure it covers all vital areas.
Content is the key to a good business plan, regardless of whether that content
is prepared manually or with the assistance of a spreadsheet program. Venture
capitalists rightfully complain about the quality of writing in plans. Thus, a plan
should be written so clearly that any of your friends or relatives who don’t work in
the high-tech field could easily understand it. Authors should not be lulled into
complacency by the nifty presentation possibilities of desktop publishing, spread-
sheets, and graphics. Venture capitalists and funders will not grade the plan
according to its thickness or sparseness, according to its flashiness, or according to
whether all possible questions (even irrelevant ones) have been answered. They will
grade a plan according to its integrity and the ability of the company’s founders to
back up absolutely every statement made in the plan.
 Successful Plans 39

SUCCESSFUL PLANS
Having gone to considerablelengths to describemodel formats for drafting business
plans, I must admit that all three of the successful plans discussed below differ from
the models in certain respects. However, they all show major elements of the model
formats, and they demonstrate the variety of approaches that can prove successful.

THE APOLLO BUSINESS PLAN (JANUARY8,1980)

Figure 3-1 shows Bill Poduska’s business plan for Apollo (referred to as ”Nuco” in
the figure).In addition to the summary, market brief, product brief, and people brief,
over half of the handwritten document (six pages) consisted of a five-year financial
plan. The financial section included such information as the projected profits and
losses, a proposed balance sheet, and the scheduled head count (the key cost-control
item in a start-up). The financial plan was used as a blueprint throughout the self-
funded seed stage, while the three key technical founders worked on the product
concept.Four months after the business plan was written, the first round of funding
closed, having raised $1.6 million, which represented 60 percent of the value of the
company. The Apollo business plan was brief because it assumed readers would be
completely familiar with the computer marketplaceand understand what is required
for development. The plan was thus designed to convince both founders and
funders of the company’s viability. It contained no near-term milestones (which is
at variance with the recommendations made earlier in the chapter), only the first
ship date. No other plan was made.

Nuco is formed to create a Profitable, Major Computer Company which

Manufactures and Sells, High-Technology, Low Price Computer Systems.

Profitable means 20% Pretax and 10% After-Tax.

Majormeans $50 million in 5 years, poised to grow to $1 billion in an orderly

way.

Manufacture and Sell includes Design, Fabrication, Direct Sales and

Distribution, Installation, and Service-Nuco will sell both hardware and
software.

High-Technology means the explosive technology of the 1980’s including

LSI Processors, Large Memory Systems, Extensive Mass Storage, Highly
Interactive Human Interface, and High-Bandwidth Network for Distributed
Data Processing.
(continued)
40 The Business Plan

(continued)
Low Price means under $20,000 for a Personal Computer System to
$70,000 for a Central Computer System.

11. Overview of The Business Plan

A. Marketplace

The Marketplace for Nuco is the community of users who now use
Time Shared Systems. These users want and demand high levels of
performance,functionality, and interaction in order to increase human
productivity. But the era of Time-sharing is ending. With the rapid
decline in the cost of computer hardware, it is no longer necessary to
share the cost of a large computer among many users, who then suffer
the inevitable delays and poor response of a shared system. The
future will be dominated by powerful Personal Computer Systems
designed to maximize individual productivity. These systems will be
integrated into a unified, but distributed computing system by a high-
bandwidth network. This Network will also include Central Computer
Systems which provide great computing power as well as support for
large central files, and sharing of expensive peripherals.

B. Product

The ProductLineconsists of two basic product systems: 1. a Personal

Computer System and 2. a Central Computer System. A typical
installation would have several PCS systems in a network, roughly
one per computer professional. A larger system might have one or
more CCS systems attached for greater throughput.

Recent advances in computer technology now make such systems

economically feasible. The most important of these advances are:

1. LSI Processors: Both Custom Gate-Array LSI and Standard 16/32

bit Processors and Peripherals.

2. 64K RAM: Making extensive Local Storage Practical.

3. Fixed Media Disk:WinchesterTechnologymaking 30-60 MB of Mass

Storage per PCS Economically Feasible.
 Successful Plans 41

4. UNlXet al.: Software Technology to Provide Highly Interactive User

Oriented Services.

5. Network Technology: High Speed Local Networking to Distribute

the Resources while Maintaining a Community of Users.

The Technology Advances have not gradually emerged but are

dramatically coming available in the period mid-1980 to mid-1981.
Thus the timing of this Venture is most appropriate and Nuco could
achieve a dominant position in the marketplace.

C. Business Plan

The Business Plan is structured to rapidly enter the marketplace and

to grow to a volume of $50 million in year 5. The Model Plan has the
following salient characteristics:

Yr Rev. Earn's Paid in caD. Comments

1 000 -900 1694 Design Products,
Build Organization
2 2060 -1339 3574 Build Sales Rapidly,
Continue Product Development
3 8230 -240 7129 Build Volume, Breakeven Q4,
Public Offering

4 21666 2547 71a9 Build Profits, Sustain Volume,

Reduce Costs
5 53164 6509 7239 Sustain Growth and Profit Rates,
Introduce 2nd Generation
Products

The Marketing Organization is structured for very rapid growth in

Year-2 and will continue to operate a high expense level. Promotion
and Merchandizing of the product are crucial to reaching such a
vigorous marketplace and will be heavily funded. Expectations are
that 65% to 75% of sales will be generated by a direct commissioned
sales force, and about 25% to 35% [of] sales will be generated by
distributors, dealers and/or representative[s]. European Sales opera-
tions will be initiated early and will account for 20% to 30% of sales in
Year-2. Software OEM [original equipment manufacturer] buyers; Le.

(continued)
42 The Business Plan

(continued)
companies which buy Nuco products, add software, and resell; may
become an important sales mechanism. Field Engineering and Ser-
vice are part of the Marketing function to insure rapid response to
customer needs.

D. Financial Plan

The Financial Plan for Nuco is to finance the rapid growth of the
business by Equity and limited Debt. The Model Plan calls for
investments of $3.0 to $3.5 million in the first two years, and a Public
Offering . . . in the third year. The Venture Capital is to be raised in
several steps with one or two lead investors in the beginning.
Additional rounds of financing are planned every 6-9 months which
will include up to six additional investing firms.

Founders and early employees will be offered stock purchases at very

attractive prices. Such stock will vest to the employee over a four or
five year periodproratedquarterly. Some founders and board members
may also participate in early rounds of financing on an equal basis with
the investing companies. The Model Plan uses debt sparingly in the
first several years in anticipation of forbidding[ly] high interest rates.
However, debt may be used more aggressively with favorable condi-
tions up to about a 1 :1 Debt/Equity ratio.

Figure 3-1. The Apollo (Nuco) Business Plan. (Reprinted with permission from Bill Poduska.)

SUN MICROSYSTEMS SEED PLAN (FEBRUARY12,1982)

An outline of the six-page Sun Microsystems seed plan is shown in Figure 3-2. This
plan is interesting because it is, in principle, exactly in line with the idea of a seed
stage business plan. Using this plan and seed stage funding, the company went
directly to break-even within a few months. One element of the plan is unusual,
however-namely, the fact that the first product marketed was a university ”labo-
ratory product” (from Stanford).

AUTODESK

The AutodeskFile (Walker, 1987)containsWalker’sfirst working paper proposing the

company in January 1982 as well as other key documents covering until 1988.
Autodesk was funded ($59,030) by its founders, a group of talented programmers
who built applications programs and marketed them through retail dealers whom
 Successful Plans 43

Mission statement: “Develop, manufacture, market, and support graphics

workstations for the OEM CAD/CAM marketplace. Evolve a family of
compatible graphics workstations. Maintain lead with the best cost/perfor-
mance product on the market.”

Objectives for four months: “Take current laboratory product to market,

begin to develop [a] workstation, assemble [a] team, build a plan, and obtain
financing.”

Tentative two-year plan.

Product: The Sun workstation, including key competitive advantages. (This

section also covered standards and current availability.)

Market: OEM workstations.

Summary of marketing approach.

Competitors.

Patents and other rights.

Current team.

Appendix A: Costs.

Appendix B: Financial requirements until May 1982.

Appendix C: All the tasks, with dates and resources.

Appendix D: Staffing for sixteen months.

Figure 3-2. Outline of the Sun Microsystems Plan. (Reprintedwith permission from Vinod Khosla.)

they knew. The company vision was that a PC revolution would occur and that this
group would simply capitalize on it. One of the firm’s first products was AutoCAD,
a program for architectural and engineering design. One of the founders’ major
goals was for the venture to be profitable from the start, and Autodesk did
essentially achieve profitability during its first year of operation. When Autodesk
went public in June 1985,each $1initially invested was worth $165, and in mid-1990,
the firm’s value was over $1billion. Given Autodesks success, and the orderly but
unorthodox way in which the company was started and funded, readers are urged
to study Walker’s book and to be equally creative.
44 The Business Plan

BUSINESS PLAN FLAWS

The company’s business plan is important because it is used in such a variety of

ways-as everything from an informal contract to the ultimate scorecard.The plan’s
main purpose is to serve as a blueprint for operating the firm (Chapter 4 discusses
how the plan is used as a control mechanism).
There are many reasons why entrepreneurs feel compelled to write unattain-
able plans. In some cases, the firm is attempting to pander to the greed of potential
investors and ends up producing a plan that is not only immoral but also potentially
illegal. In other cases, the company is simply fooling itself and its investors through
ignorance and incompetence. Hence, one of the most serious flaws is to start a
venture by accepting major product development financing without having an
adequate development plan-a plan that could have been written had the company
taken the time for a seed stage.
The following subsections present some common plan-related flaws and de-
scribe a number of classic cases of flawed firms whose difficulties stemmed from
poor business plans.

UNREALISTIC PLANS

Proceeding according to an unrealistic plan is one of the greatest risks a start-up

company can take. Two frequent causes of unrealistic plans are investor greed and
technology miscalculation.
Start-ups often prepare absurdly aggressive and optimistic plans, which have a
very low likelihood of success, just to maximize the company’s perceived dollar
value. In succumbingto venture capitalist greed (”Showus a plan that creates a $100
million company in five years”) and funding requirements (”We don’t put less than
$X million into a deal”), the company establishesexpectationsthat cannot be met. In
1990, the canonical business plan projects revenues of $50 million in five years. Such
”plans” are often the product of misused spreadsheet technology and venture
capital funding patterns rather than being the result of careful,realistic planning by
the company’s founders.
Greed may continue to dominate even after a company fails to meet its initial
business plan and requires additional funds. In such cases, the plan is often revised
to project even higher revenues, in order to prevent the firm’s valuation from being
lowered. This spiral continues, with accelerating expectations, until the funders at
last tire. When the funders do tire, the valuation is finally corrected for the next
funding round, and the organization and its management are restructured to
operate at a substantially reduced scale.
 Business Plan Flaws 45

Table 3-1. Example of Company Valuation at Variance with Actual Performance.

Round PricelSkare a No. of Shares $s/Round Company Total
Valuationh Raisedb
12/83 0.3300 0.42 0.140 0.14 0.14
7/84 1.oooo 7.70 7.700 - -
common 0.0100 2.60 0.026 10.70 7.90
9/86 2.0000 3.20 6.400 29.80 14.30

3/88 0.3000 10.80 3.200 8.00 17.50

8/89 0.0750 20.00 1.300 4.20 18.80

7/90 0.0125 240.00 3.000 3.70 21.80

dollars.
millions of dollars
‘In millions.

I speak from firsthand experience, since I once invested in a company whose

operations aptly illustrate the spiral-of-accelerating-expectationsscenario. This
firm, whose funding history is shown in Table 3-1, was even backed by a foreign
government.
Table 3-1 illustrates a typical case of a company’s valuation getting out of line
with its actual performance. The price of the stock rose initially, while the firm
developed its technology and first product, led by the founding president, a
technologist. The second president, a former CEO, was then brought in to run the
company ”professionally.”He was able to sell stock at $2 a share and proceeded to
spend the company’s funds by buying another firm and investing in several
products that were orthogonal to the primary venture’s technology, marketing and
sales expertise, and general business. The third president operated a demoralized
and disorganized firm for two years while the original product slowly gained
market acceptance.
In 1989,one of the company‘sventure capitaliststook over as president, reduced
expenses to a minimum, and delegated responsibility for planning and operations.
Finally, the firm started selling in sufficient volume to attain profitability. The plan
on which the July 1990 funds were raised was to finallybecome profitable and then
sell the much-devalued company for several tens of millions of dollars, thereby
46 The Business Plan

allowing the last round of patient (or foolish) investors to attempt to make a
severalfold return in order to rapidly recoup their original investment.
Although the company was founded during a period of bountiful capital,
common sense would rule out a venture that required multisite international
operations from the outset. What looked like a sure money maker-with compelling
technology and engineering, backed by government funding, and having a built-in
home custom market-was derailed by the existence of too many agendas (multisite
international operations, custom and standard products, doing research contracts)
and other management-related factors. The early investors’stake was diluted by the
fact that they did not continue investing in the company in the final financing
rounds. The early founding employees (common stockholders) have negligible
equity except through common stock that was issued in the final round. Despite
these shortcomings, however, the firm did have sound technology and a viable
product.

DOING RESEARCH AND CALLING IT PRODUCT DEVELOPMENT

Figure 3-3 is derived from the financial data of a company that was formed in the
early 1980sto build a human input device. The figure compares the actual operating
revenue (the plain in the foreground), which rose to nearly $6 million per year
during the first four years, to the ever-increasing mountain range of plans in the
background, which represents the firm’s dreams for its sales. The first two-year plan
projected sales of over $10 million, even though market forecasts in the 1980s (still
unrealized in the early 1990s)for the product showed sales in the billions of dollars.
The second and third plans, which enabled $15 million to be raised, projected $40
million in revenue. Finally, in order to make the last two financings of over $11
million, projections of more than $50 million in sales were required. Altogether, $38
million was raised over a period of seven years. The question the company still faces
is, if and when the technology matures to the point where it can serve a large market,
will the firm be able to respond, or will some other competitor, such as IBM, come
in and take it all?
More recently, a new company announced its intention of introducing a
computer that accepts handwritten input. The questions associated with its
product/market viability are exactly the same as those for the firm shown in Figure
3-3: (1)can the company be kept under control while its product is sold to an infant
market willing to pay high prices for a new technology product, and (2) will the
required technology mature rapidly enough for the product to decline in price
sufficiently to make it attractive to a broad, general market?
 Business Plan Flaws 47

50

40
50

30
40

20

30
10

20
0
U!
w9
10

Figure
48 The Business Plan

LOSING TOUCH WITH REALITY

A plan contains both a spending stream (reality) and a revenue stream (a desire).
When the two streams diverge significantly, a new plan is needed. Here are a few
popular rationalizations for why the two streams may have diverged:

”All our expenses are on plan. We do have a ’top-line’ problem, however.”

(When a company fails to meet its revenue plan while proceeding with its
expense plan, the difference shows up immediately on the bottom line.)

“We‘re selling the right quantities, but the discounts are much higher than
expected.”

“We‘re selling and people are buying, but we still can’t produce the product.”

“We’ve met our hiring and spending plan, but the product still isn’t quite ready
to ship.”

The common thread running through all these rationalizations is the (midbelief
that the plan is more real than the tangible results. This form of dishonesty occurs
when the company refuses to face the facts that it sees in hiring, schedules, costs,
sales, etc.; believes that the plan is reality; and views the actual facts as anomalies
(expressed by parentheses or minus signs) that will eventually go away. The
frequent result is that the CEO and board refuse to create a new plan, and the
company goes blissfully on, runs out of money, and returns to the investors with an
even more aggressive plan, as shown in Figure 3-3.

LACK OF A SUSTAINING TECHNOLOGY OR PRODUCT

The most common flaw of a high-tech business plan is to center the plan on a ”one-
shot” technology or product, without providing for subsequent products. Such a
plan may be able to validly project a market victory for the first product, but it has
no enduring vision for the company and no strategy for how the company will win
in the long term. A firm has no lasting advantage if it is based on a transient product
or on a distribution scheme designed simply to fill a niche left by a dominant
supplier.
When integrated circuits first appeared on the market in the 196Os, a huge
number of minicomputer ventures sprang up, ready to do battle with the existing
suppliers, pinning their hopes on their first (and only) product. A majority suffered
from the lack-of-sustaining-technologyflaw and failed. A similar situation has
occurred more recently, with the first computers that utilize RISC (reduced instruc-
tion set computer) technology. In this case, companies have formed to exploit a
 Business Plan Flaws 49

particular technology in an existing, filled marketplace where all the players are
ultimately likely to catch up, and have been doing so.
The lack-of-sustaining-technologyflaw is also discussed in Chapter 5 ("Tech-
nology and Engineering"), Chapter 6 ("The Technology Balance Sheet"), and
Chapter 8 ("The Product"), since it is also a technology flaw and can cause a flawed
product idea, thereby resulting in the creation of a flawed company.

QUESTIONABLEMOTIVATION

Business plans flawed by questionable motivation (and no product vision) are

written by:

Chronic entrepreneurs who start many companies but do not build organiza-
tions that last

Entrepreneurs who start a company to "get even" with a competitor (usually a

former employer)

Entrepreneurs who want produce a perfect "Product X in a field of fifty com-

panies that already build a "Product X"

Most business plans exhibiting the questionable-motivationflaw result from a

need to satisfy the ego and personal desires of the founder(s).Chronicentrepreneurs
enjoy starting ventures but lack the determination to stay with a small firm and
nurture it into a larger, healthier, and more successful one. These people get high on
the rush of excitement associated with any start-up, including that of a hot dog
stand. Although they can be useful in the start-up process, they are unlikely to build
the infrastructure of a lasting company. Another questionable, but nonetheless
frequently observed, motivation is to start a firm in order to compete against a
former employer. In such cases, revenge rather than vision is the driving force, and
failure is the usual result. The third questionable motivation, perhaps the most com-
mon of all, is a form of technical arrogance-the not-invented-heresyndrome. In this
scenario, engineers enter a crowded field to build an incrementally better product.
Naive arrogance of this sort accounts for most start-up failures, including the failure
of many of the one hundred firms that were founded to build minicomputers.

SKIPPING THE SEED STAGE

Skipping the seed stage could be called a "lack-of-refinement" flaw. By rushing into
the product development stage without a seed stage business plan, the company is
relying completely on the untested assumptions made in the concept stage business
50 The Business Plan

plan. Because its plan lacks the refinements normally incorporated during the seed
stage, the firm faces higher market and development risks. The results are usually
the same: it costs more, takes longer, and requires more resources to market the
product (which may then fail the market test). Skipping the seed stage is especially
apt to be fatal if the founders haven’t done a similar product before and the proposed
product involves a significant amount of development that relies on several tech-
nological breakthroughs.

MULTIPLE AGENDAS

During the operation of a company, new ideas frequently emerge, and the firm may
allocate resources to these new endeavors. If the new endeavors are outside the main
thrust of the company’s principal business, even a well-established organization
may run into trouble. For example, in the 1970s, Control Data Corporation (CDC)
acquired a wide range of businesses to become a conglomerate.It also tried to reform
education by investing hundreds of millions of dollars in the development of the
Plato Computer Aided Instruction system at the University of Illinois, operated
inner-city factories, and promoted trade with Russia to fight the cold war. It even
had a hydroponic garden on one of its buildings. These multiple agendas eventually
had an adverse effect upon CDC’s core business-computers,
For a start-up, the pursuit of multiple agendas is particularly troublesome, since
it spreads the fledgling firm’s already-thin resources even thinner. Start-ups cannot
afford to try to become conglomerates.Furthermore, a start-up that follows ”other
trails” after creating the seed stage business plan is usually on morally and legally
shaky ground, because that plan is a contract that the company has made with its
investors.

TWO OR MORE START-UPS IN ONE

Most successful ventures began with one idea and product: Intel with memory
chips, Microsoft with a Basic compiler for Altair, Apple with a home computer
board, NCR with cash registers. Divisionalization and a multiplicity of products
came after profitability and established success.
An organization founded to conduct two parallel and independent projects,
each of which might in itself be the basis for a company, faces a high probability of
failure. In effect, two start-ups are being managed under the umbrella of one firm.
Such a plan is predicated on greed and a naive misunderstanding of the difficulty
of starting a company. For one thing, two projects usually cost just about twice as
much as one. For another thing, redundancy does not result in a lowering of the
 Business Plan Flaws 51

company’s risk. In fact, dual projects involve considerably more than twice the risk
(the risk may be as much as the square of the number of projects per firm).
The multiple-start-ups flaw can manifest itself in either of two ways:

The company may engage in top-down, or backward, integration as a system

supplier to reduce product cost before a business exists. For example, the firm
may attempt to develop all the components that it could buy and therefore be
unable to concentrate on what it intends to sell.

The company may engage in bottom-up, or forward, integration from a com-

ponent technology.For example,the firm may have a new and unique component
whose sale could be the basis for a successful company, but the firm tries instead
to capture the ”whole market” by using the component to make systems.

An interesting example of a company exhibiting the top-down, or backward-

integrating, flaw was Cydrome.

Cydrome. Cydrome was formed in 1983 and produced an impressive

minisupercomputer that exploited parallelism by executing several instructions at
once. The computer consisted of a numeric portion and a UNIX front end (for
communicatingwith users, networks, and managed files).The front end was similar
to UNIX-based multiprocessors available from Arete, Encore, and Sequent. As a
founder of Encore, I attempted to convince Cydrome to use an Encore computer for
the firm’s front end, because it could save time and about $15 million in development
costs by doing so. However, Cydrome’s management felt that the Encore machine
was too costly and that Cydrome could make contributions in front-end design.
Development of the numeric part went well, but the front end was late and
proved to be more costly than anticipated, causingthe total system to be late and less
competitive.Thus, Cydromeexhibitedthe classictop-down,or backward-integrating,
flaw by devoting resources to developing something the company could have
bought rather than concentrating on its principal field of expertise, which was the
numeric portion of the machine. Unfortunately, the flaw was fatal, since Cydrome
folded after five years.
A good example of the bottom-up, or forward-integrating, flaw can be seen in
the case of Vitesse.

Vitesse. Vitesse began as a manufacturer of gallium arsenide (GaAs) circuits, for

which a small market existed. Vitesse then funded the development of a computer
to exploit the circuits it had developed. When problems arose with GaAs circuits in
terms of cost, size, and availability,the computer division switchedto complementary
52 The Business Plan

metal oxide semiconductor (CMOS) technology. Fortunately, the firm recognized

its mistake in time, closed the computer division, and switched back to its roots,
manufacturing GaAs semiconductor parts.
A third example of the multiple-start-ups flaw is Modular Advanced Design, a
company that exhibited some elements of both bottom-up, or forward, integration
and top-down, or backward, integration.

Modular Advanced Design (MAD) Intelligent Systems. MAD is a Silicon Valley

start-up funded to build a workstation that would use rule-based systems to access
knowledge bases and databases. It began by building (rather than purchasing) an
ordinary PC as its workstation base. It then implemented a new (rather than an
existing) version of LISP, which it integrated with its own design for a database.
Finally, it built the shell, a rule interpreter, which it considered to be its proprietary
and unique technology.
Having thus exhibited the top-down, or backward-integrating, flaw by devel-
oping components that it could have bought, MAD then proceeded to exhibit the
bottom-up, or forward-integrating, flaw by trying to use its unique component (the
rule interpreter) to make systems and thereby capture the “whole market.” To take
the product into a variety of markets, MAD had to build operational prototypes of
several applications, a very expensive and resource-draining process. In 1990, the
company is selling financial-services software products written in C.

Although a company that concentrates on what it does well may start smaller,
the likelihood of its succcss will be four times greater than for the company that tries
to do everything. When Digital Equipment Corporation was funded in 1957 with
$70,000from American Research and Development, for 70 percent ownership of the
firm, its plan was to start by developing transistorized digital modules and eventu-
ally develop computers. In the first year, it was profitable based on sales of the
modules. In 1960, it introduced a computer that used the modules.
As readers will see in the following examples of MIPS and VPL, it is possible to
avoid both the multiple-start-up and the one-shot-technology/product flaws by
going deeply into technology, and it is also possible to develop and maintain the
technology base by establishing partnerships with other companies. But as the
Gyration example will illustrate, such a plan is not easy to sell to investors because
it implies a lack of focus.

MIPS Computer Systems, Inc. MIPS was successful because it developed RISC-
technology microprocessor chips on which to base its computer board and system
products. MIPS knew that making a great computer meant starting with the silicon
and having the best chips, unlike the first RISC companies-Pyramid (now a MIPS
customer) and Ridge (defunct). In 1985, MIPS initially designed and sold chips,
 Business Plan Flaws 53

boards, and systems.By 1988,it was no longer sellingchips but instead licensed chip
designs to semiconductor companies for a license and royalty fee. By 1990,it became
a supplier of systems and applications software that it and third-party software
suppliers created. In this way, MIPS users had a single standard, providing a large
market for software suppliers.

VPL. VPL started up to build a "virtual-reality" system that allows a user wearing
a special helmet with displays for each eye to "walk through a three-dimensional
space. Navigation is controlled by head movement and by a special glove. The
company needed a program for displaying a 3-D space, so it wrote Swivel 3D and
licensed it to Paracomp. In the short term, the market for Swivel 3D is substantially
larger than for any virtual-reality product. VPL also designed a simplified glove for
Nintendo games. Although the helmet and glove are sold as components, the vision
for the company is still to create a virtual-reality system.

Gyration. Gyration started up to make a 3-D pointing device-in essence, a 3-D

mouse (perhaps it might be termed a "bat"). The founder, a successful entrepreneur,
hired a graduate fresh out of Stanford who invented a revolutionary gyroscope.
Their patented invention reduces both the cost and size of gyros by large factors
while increasing their accuracy and stability. The company has both the need and
the opportunity to become a significant gyro manufacturer because it has created a
breakthrough component. As an enabling technology, Gyration's invention could
replace many existing gyros, and it could form the basis for fundamental new
products such as television remote controllers, 3-D pointers, and controllers for
industrial manipulators, mobile robots, and vehicle navigation units.
In the summer of 1990, the venture capital community was reluctant to fund
Gyration,however, because every venture capitalistknows that software is "in" and
that all low-cost, volume-produced products have to be designed and manufac-
tured in Japan. The venture capital community offered Gyration the following
recommendations,despite the fact that, paradoxically,it is also unwilling to fund the
proposed course of action:

First, the venture capital community urged an initial,unlikely plan under which
the company would build and sell computer pointing devices-a potential, but
currently nonexistent future market that no one can prove will actually materi-
alize. In fact, at this stage, companies should focus on only one application.
Second, it ruled out other, equally feasible product applications areas because
companies producing components are even less in vogue than companies
entering new markets. Conventional wisdom says that components sell at a
lower price and have a diffuse market.
54 The Business Plan

Under the second operational plan, Gyration will prove the efficacy of its
technology by building the pointing device product while at the same time selling
gyros for every potential new application. American military vendors, who want
and need a gyro (e.g.,for use in low-cost, small, smart missiles) are too bureaucratic
to allow themselves to invest in or use the new technology. Meanwhile, Japanese
component and system suppliers are happy to fund the company and to get the
rights to build small, accurate, inexpensive gyroscopes.
Gyration intends to make both the technology and the company successful. In
the process, the country and infrastructure that supported Gyration’s invention may
reap none of the rewards-xcept through products purchased from foreign firms.
A sad, but typical scenario. Stay tuned for further developments.

BUSINESS PLAN RULES

Although there are stories of business plans being written while entrepreneurs are
part of another organization, this is not the norm. Furthermore, writing a plan while
one is part of another organization is immoral and potentially illegal, and should
never be done!
Another warning: Although investors and other reviewers may sign nondis-
closure statements, a start-up’s business plan will nonetheless end up as a public
document and will be seen both by its traditional competitors and by others who are
starting up with similar ideas. Therefore, the founders should avoid including
excessively detailed and critical information about development schedules, stock-
holders, budgets, market projections, marketing approach, proprietary products, or
proprietary processes. Such details can be presented verbally on a ”need-to-know”
basis in the course of the due-diligence process.
In the case of successful companies,a small core of founders leave their jobs and
write a completebusiness plan on a full-timebasis. Finishing the plan typically takes
from three to twelve months, depending on technology, market uncertainty, com-
plexity of the product, and complexity of the manufacturing processes.To determine
whether they have written an effective concept stage business plan, the founders
should ask themselves the following question:

Is the plan’s summary of technology, product, market, and formal business

plan development short (six to ten pages, excluding appendixes), and does
it contain the following elements?

Statement of the proposed company‘s vision, mission, and business

 Business Plan Rules 55

Product concept (what the product is)

Technological uniqueness that will sustain the firm beyond the initial
product
Rationale (why people will buy)

Gross estimates of the target market (who will buy)

Simple ”market map” (how the product will be sold)

Plan for reaching the seed stage, with objectives and milestones

Outline of a financial plan for the company

Resources, in terms of dollars and people

All nine parts of this question must be answered, and the answers must be
detailed enough for the concept stage. Of the nine parts, the plan for reaching the
seed stage is especially important, since this is fundamentally what investors are
buying during the concept stage round of financing.Thus, the concept stage plan is
both the first draft of a traditional business plan (which convinces investors of the
company’s potential) and a ”plan for a plan”-i.e., a plan for the company’s “real”
business plan.
Near the end of the seed stage, the start-up prepares for entry into the product
development stage. One of the tasks that must be accomplished at this point is to
upgrade the concept stage business plan to a seed stage business plan. To determine
whether they have succeeded in performing this upgrading, the founders should
ask themselves the following questions:

Has the concept stage plan been updated, expanded, and confirmed as a
result of the seed stage? Is the plan now twenty to thirty pages long (not
counting the financial appendixes)?Does it contain the following elements?

Statement of the proposed company’s vision, mission, and business

Product concept (what the product is)

Technological uniqueness that will sustain the firm beyond the initial
product

Rationale (why people will buy)

56 The Business Plan

Gross estimates of the target market (who will buy)

Simple “market map” (how the product will be sold)

Financial plan and details of company ownership

Description of the firm’s people and operating philosophy

Key milestones in product development and company growth

The first six items on this list are the same as those for the concept stage business
plan, and the overall plan has a fundamentally similar format. The major difference
is that earlier assumptions have now been verified, and the company should have
enough information to more accuratelyplan the development of the product and the
market.
In addition to the items on the preceding list, the founders need to ask
themselvesa few other questions about the seed stage business plan. Several of these
questions partly overlap the items on the list but are sufficiently important to
warrant examining them in greater depth.

Does the company have a formal financial plan that includes the strategy
and timing of present and additional funding rounds, types of backers
being sought, etc.?

To support the financial information in the business plan, the company should
have a position statement about its intended method of financing and a ”sketch
plan” schedule of the financing requirements for the first five years.

Does the plan clearly demonstrate that the company is sustainable and
verify the assumptions initially made at the concept stage? (E.g., is the
technology implementable, is the engineering plan valid, and has the firm
determined why customers will buy?)

Technology and market verification were carried out during the concept stage.
The seed stage business plan simply has to present convincing arguments for why
the company is sustainable.

Does the plan refer to a detailed plan for the next stage of the start-up (the
product development stage), including a list of objectives, a schedule with
milestones, and allocations of the required financial and human resources?
 Conclusion 57

Although the seed stage business plan need not contain details of the product
development stage, except for a few key dates (the completion times of the four
phases of the product development stage),the company should have a development
plan. As will be discussed in Chapters 5 and 6, the creation of a start-up company
has a technology dimension in addition to the business plan dimension being
discussed here. If a development plan is not availableduring the seed stage,both the
technology and business plan dimensions of the start-up process don’t measure up
to the ideal.

Are the product development times, product cost, product performance, and
external risks (component or process) clearly identified, and are they ac-
counted for in the plan’s funding?

A business plan should not identify risks merely so the company can tell
investors that a potential problem materialized as predicted. Rather, the plan must
contain adequate backups and contingency provisions to enable the firm to deal
with the problems. If the company has made no plans for creatively managing the
problems that will surely arise, the product introduction schedule and the product
cost will be adversely affected. The inevitable result of missing the schedule and
blowing the product cost is dilution of the company and reduction of market share.
At the seed stage, the founders need not state every possible risk with the rigor
that is normally required in the prospectus of a company about to go public.
However, they definitely must indicate all foreseeable problems that may affect the
plan’s outcome, if only to defend themselves against potential lawsuits.

CONCLUSION
A company must always have a single vision and a common business plan, for
without this guiding focus, it will be directionless. (Imagine, for example, that
instead of consistentlysailing west, Columbus had asked his officers and crew each
morning which way they felt like sailing that day!) Although such a plan can be
changed, one and only one plan must exist at any given time, and everyone in the
organization must be trying as hard as possible to carry out that plan.
The business plan serves many purposes: first, it is the document that the
company uses to secure funding; second, it is the plan for operating the company;
and finally, it is the yardstick against which the company is measured.
The company’s vision is a statement of its image and trajectory within an
industry, reflecting the essence of what it is attempting to be. Such a vision must be
58 The Business Plan

incredibly simple. In fact, the larger the firm grows, the simpler the vision must
become. Apple, Digital,IBM, Lotus, Microsoft,and most recently, Sun Microsystems
all used single product lines around which to rally resources and focus effort.
In short, the company that has a well-thought-out vision and a truly effective
business plan understands the purpose of its existence and knows where it is going,
how it intends to get there, and how it will demonstrate that it has accomplished
what it set out to do.
Chapter 4

CASH, FINANCEABILITY,
AND CONTROL

A company’s financial health is determined by three factors:

Cash: the funds that the firm has on hand or can obtain rapidly (i.e.,in less than
three months)

Finunceubility: the company’s ability to raise cash in the short term (i.e.,in three
months or longer) and in the long term (i.e., over the life of the firm)

ControZ: the company’s ability (going substantially beyond simple financial

control) to operate according to a plan that specifies income, spending, and
overall results

The following sections examine these three determinants of a start-up’s financial

health and present the flaws and rules applicable to each one.

59
60 Cash, Financeability, and Control

CASH: FUEL TO GET TO THE NEXT MILESTONE

Cash is more important than your mother.
-AI Shugart

Inadequate cash for growth is the number two killer.]

-John Shoch

Cash is a crucial resource, because only cash can buy the fledgling company time to
search for answers to such hard questions as: When will a viable business plan be
ready? When will the technology work? When will the product work? What is the
market for the product? How long will customers take to decide to buy the first
product, and when will they and their colleagues buy more? When will customers
pay cash for the product?
In many ways, cash and time are opposite sides of the same coin, since time
sometimes “buys cash.” With time, a critical problem can be solved so that a product
can become operational or go from unacceptable to great. Also, with time, a
customer may pay a bill, a large order can come in, or more financing can be
obtained. Nearly all failed companies claim that if they had simply had more cash
and had not run out of money, the venture would have worked. On the other hand,
cash can also prolong the inevitable demise of ill-conceived, cash-rich firms.
Ideally, the company will at all times have more cash on hand than what is called
for in the business plan-i.e., the company will be ”above plan.” Having adequate
cash permits the firm to control the timing of its next request for financing,such that
the request is made when it is in a strong negotiating position. For example, the cash
available during a start-up’s product development stage must last through the
alpha-testing phase, so that investors will be convinced that the product is sound
and will therefore regard the company’s next round of investment as worthy of a
higher valuation.
In contrast, lack of cash could put the company in the unenviable position of
needing money immediately in order to meet its payroll. When the firm is thus
pressed to the wall, its negotiating position is nil, and investors and bankers can
make the price of money almost anything they want, including below the price of the
previous round of financing.Sy Kaufman of Robertson Stephens, when counseling
a company about the need to accelerate its funding plans, stated: “The pain caused
by running out of money is just unbelievable and unbearable. Don’t ever let this
happen to you.”
A company in the market development stage can find that lack of cash is the
primary limitation on its growth and success. Banks are usually unwilling to
-
1 . Lack of team is the number one killer.
 Cash 61

provide a line of credit to unproven ventures and are unlikely to make loans against
collateral that consists merely of accounts receivable and customer purchase orders.
Lending money to established small businesses with a good cash flow is usually
much more profitable for the banks and does not involve the risks inherent in trying
to understand a complex industry. There are exceptions, however, and some banks
in high-tech areas (such as the Silicon Valley Bank) are aggressive pioneers in
making conventional and equity-backed loans to start-ups.
Having adequate cash and negotiating additional financing from a position of
strength also permit the start-up to grow without giving up substantial ownership
to the investors. Nearly all the companies with which I have been involved have
pursued this goal. Unfortunately, in order to achieve this goal, the CEO and the CFO
(chief financial officer)are often forced to spend almost full time looking for money,
even in firms that fund their growth primarily through their existing investors.
In short, start-ups face the same financial paradox that individuals do when
dealing with banks: "The only time you can borrow or raise money is when you
don't need it."

CASH FLAWS

Three of the following four flaws involve a lack of cash. If the start-up doesn't have
enough cash, it may be unable to get off the ground or advance to the next stage; and
if an already-established companyburns off its cash by being out of control, its board
of directors may take over active management of the firm. The final flaw involves
having too much cash, which, surprisingly enough, can also be detrimental to a start-
up's health.

Inability to Pay the Founders During the Concept Stage

Perhaps the most common cash flaw a start-up can exhibit is simply having no way
to support its founders while they make a creditable business plan. Thus, the
fledgling organization is faced with a dilemma: a plan must be written, but the
founderscannot leave their present jobs without support, nor can they write a plan
while they are part of another company. The only way out of the dilemma is to have
one of the founders write the plan while not working for an existing firm.

Having Inadequate Cash to Move to the Next Stage

In either the seed stage or the product development stage, the start-up may have
inadequate cash to move to the next stage and demonstrate the firm's competence
62 Cash, Financeability, and Control

or efficacy.This flaw can also manifest itself at a later stage in the company’s growth
if it fails to deliver on its product or market development promises.

The Investor-Run Company

If the start-up’s cash declines to the point where investors repeatedly have to put in
more money on an emergency basis, they-rather than the CEO, CFO, and team-
can end up running the firm. When this happens, investors may keep the purse
strings very tight, doling out funds one phase at a time in an operational fashion and
even making such decisions as when to buy parts in order to make the first
prototype. This is an inherently poor way to run a company.

Having Too Much Cash

As noted above, being cash-poor can stifle a firm’s development. It might therefore
seem as if there would be no such thing as having too much cash. But the cash-rich
organization runs the risk of becoming careless. The company may start off on-plan
but then slip into operating in a sloppy fashion that will ultimately require a major
adjustment.
Although the publicridiculesthe extravaganceof somelarge, wealthy companies,
small start-ups may exhibit equally foolish spending habits. The on-plan venture
that has much more cash than it needs can easily get into trouble, because it is quite
likely to begin spending without the appropriate planning.JohnGrillo-former CEO
of SPSS and Tesseract and venture capitalist at Robertson, Stephens & Company in
San Franciscdescribes this condition as “financing-induced brain damage.” A
cash-rich company is prone to acquiring many bad habits, including inadequate
control of spending, unwillingnessto continually prune growing expenses (including
people), and a general inability to run lean and mean. In short, all organizations,
regardless of their size, must watch their outlays.
Some good spending habits for start-ups that are almost never practicedby large
companies include such tactics as planning trips wisely and in advance to save time
and money, not booking businessclass, choosingreasonably priced accommodations
(e.g., Days Inns), and buying nonmatching, used, or auctioned-off furniture. Most
important, salaries must be based on value to the company and performance rather
than on traditional salary-hierarchyformulas. Suhas Patil, founder of Cirrus Logic,
believes that the founders establish a start-up‘s salary standards and argues that the
pay scale for hiring should be governed by need and performance, not by hierarchy.
There is some evidence to support the idea that less money is better. Objectivity,
a Silicon Valley firm building an object-oriented database, studied a number of
 Cash 63

software companies and found that many of the most successful did not have a
significant amount of venture financing in their early years. The Objectivity team
hypothesized that the absence of a large bankroll forced these organizations to
behave in several ways that tended to promote their success:

Smaller projects had to be undertaken to get products to market sooner,

reducing development risk.

Development teams remained small and focused, an approach widely believed

to be most effective for software.

Product definitionand developmentwere often customer-funded,which ensured

that the resulting products would meet real market needs.

Getting to market sooner with useful products enabled the companies to

establish a lead in the race for market share.

CASH RULES

When it comes to starting a venture, ”cash is king.” Without cash, there is no

company. During the concept stage, the founders can trade off their personal time
for cash, but once full-time employeesare hired, cash is required to fuel the firm. The
following rules test whether, at each stage, the organization has met its objectives
and has enough cash to carry itself through to the next stage.

During the concept stage, do the founders have sufficient (usually personal)
time and cash to be able to write the business plan for the seed stage?

The only mechanism for funding the concept stage, during which the first plan
is developed, is the founders themselves. They must be capable of sustaining
themselves while they write the seed stage plan and look for seed or start-up
financing.Having adequate time and cash for this indeterminate period, which may
last up to a year, is essential. At this first stage, out-of-pocket, personal cash is
synonymous with financeability.

Have the founders obtained the cash to execute the seed stage plan?

The preceding two rules test whether the company has the cash and time to start
up. Without funds or some way to support the founders during the concept and seed
stages, the firm will be unable to get off the ground.
64 Cash, Financeability, and Control

By the end of the seed stage, has the seed stage funding been sufficient to
enable the start-upto meet its objectives and milestones for that stage?Does
the company still have enough cash on hand to sustain itself for a period of
up to three months while it pursues product development stage financing?

At the completionof the seed stage, the start-up may require more time to search
for additional cash than was originally anticipated. For this reason, the company
should always be in the position of having up to a three-month supply of funds in
reserve. Furthermore, it should be understood that receiving a commitment for
funding is not the same as having cash in hand, since several months can elapse
between agreement and the actual availability of funds.

FINANCEABILITY:
VIABILITY THROUGH THE ABILITY TO RAISE CAPITAL

Nowhere is the expression ”timing is everything” more applicable than in financing

a new venture. The principal measure of financeability is the ability to raise capital
in a timely fashion at a fair market value. Furthermore, raising capital in a timely
fashion is not a problem that the start-up will face only once. Rather, it is an ongoing
process, because the more successful the firm, the more additional capital it will
need to sustain its growth. Even with continued and increasing profitability, a
company’s capital requirements will rise sharply, especially if it is operating in a
capital-intensive field such as manufacturing (e.g., semiconductors or disks), in
which large amounts of cash are needed to finance plant and inventory expansions.
Financeability is perhaps the most difficult dimension to measure, since it
depends both on circumstances that are entirely beyond the company’s control
(exogenous factors) and on circumstances that the company can directly control
(indigenous factors). The exogenous factors that affect financeability include the
following:

The condition of the overall economy, together with the apparent market for the
firm’s proposed product or service

The number of competing companies engaged in businesses similar or identical

to that of the start-up

The financial community’scurrent health and willingness to fund the business

sector

The desire of a given source of funds to be in a particular industry sector

 Financeability 65

The indigenous factors on which financeability depends include the following:

The company’s ability to operate according to its business plan

The competitivenessof the product (i.e.,the product position) and the likelihood
of the firm’s achieving its planned results

The organization’sability to create a competitivemarket in which investors will

find the company attractive

The company’s perceived intangible value, including any synergy with other
companies in an investment portfolio

The return that the firm offers its investors on their investment

During any round of financing,the company’s valuation is determined by all of

the factors in the preceding lists, together with the vicissitudes of the negotiating
process. The best way for a new venture to get a high valuation is to offer investors
a great company and then have lots of firms that want a piece of the deal, such that
the offering is oversubscribed. Attracting a large number of buyers is also a good
way to ensure that they are not illegally colluding to establish the price of the
offering.
Two calculations are especially important in determining the company’s
valuation:

The cost that was required for the start-up to attain its current position,
compared with the cost required to finance a similar company to a similar point

The expected return

The former calculation is critical,because if a company is founded on a great idea

that is both simple to implement and offers a high expected return, other similar
companies will probably be formed. (Remember Silver’s [1985] axiom for the
venture business: “Anything worth doing is worth duplicating.”)The possibility of
other similar firms being funded will tend to drive down the valuation of each of the
individual firms.
The expected return is also an important consideration in valuation. It should
increase in proportion to risk and will vary depending on where the company is in
its growth process. Several books on entrepreneuring (such as Gladstone, 1988)
discuss the process of determining appropriate valuations based on the expected
return. Although in the 1980s, venture capital firms average about a 22 percent
annual return, the contributions of the various companies in their portfolios may
66 Cash, Financeability, and Control

vary widely. For example, Saratoga Venture Finance (Nesheim,1988)estimated that

in a typical venture capital portfolio, only 10 percent of the investments were
substantial winners, with 6 percent returning 50 percent annually and 4 percent
returning over 100 percent annually. Winning investments such as these offset the
investments in firms that go bankrupt (60 percent!).
Given these statistics, it is easy to see why venture capital firms need to pick as
many winners as possible and avoid paying too much for either the winners or the
losers. Silver (1985) offers some advice to assist these firms in valuing a company
appropriately. He starts by looking at the firm during five risk periods-product
development, manufacturing, marketing, management, and growth, which roughly
correspond to our stages-while applying three laws of venture capital:

1. Accept no more than two risks per investment.

2. Valuation = P X S X E , where P is the size of the problem being solved, S is

the elegance of the solution, and E is the entrepreneurial team; the S factor
is further defined to be:

S = B X T, where B is the Business Plan, and T is the technology.

3. For companies where the above formula yields comparable results, invest
in the big-P companies because the public market will accord them unrea-
sonably high valuation, irrespective of S and E.

Silver looks at the following eight DEJ (demonstrable economic justification)

factors in determining the value of S and P:

1. Existence of qualified buyers.

2. Existence of qualified sellers.

3. Homogeneity of buyers.

4. Large number of buyers.

5. Lack of institutional barriers to selling.

6. Word of mouth is principal form of advertising.

7. Optimum price/cost relationship.

8. Whether invisibility of the new company can be maintained.

In 1990, many entrepreneurs are saying that venture capital has turned itself
inward and operates more like a bank, with venture capitalists being unwilling to
 Financeability 67

finance an enterprise unless it has a proven operational prototype, orders in hand,

and a guaranteed market. In fact, in the case of Gyration (page 53), the venture
community is unwilling to invest in what may be a breakthrough technology even
though the company already has a bank line of credit to build prototypes.

SOURCES OF CAPITAL

Two major sources of capital exist:self-fundingand externalfunding. Although self-

funding is by far the preferred method, as Gateway (see Chapter 11)and Microsoft
have shown, if that approach is not possible, the company will have to seek
a financial partner. To do so, it must establish criteria for a desirable partner and
decide on a financing strategy; otherwise, it will waste a great deal of time searching
for the ”golden goose.” The selling and final negotiation processes determine the
terms for the “deal” with the start-up’s financial partner.

Self-Funding

I strongly recommend some form of self-funding, if at all possible. Self-funding

permits the founders to spend more time on their business plan and on product
development without having to devote major amounts of time to wooing financial
backers. Self-funding also fosters a strong sense of discipline with respect to
spending control,shortens the time it takes to reach the market, minimizes extraneous
marketing and sales overhead, and keeps the firm from being overly concerned with
product elegance. On the one hand, a self-funded company does sacrifice some of
the advantages that venture capital or other investors can offer. On the other hand,
it avoids burdening the board with three or four venture capitalists who lack any
understanding of the product or market-a surefire recipe for disaster.
Self-funding has been quite successful for a number of companies, especially
software firms. One example, described more fully in Chapter 11, is Dragon
Systems. Dragon was founded in 1982 by Jim and Janet Baker and has become a
leader in speech-recognition products. The company, which bootstrapped its fi-
nancing without outside investment, is employee-owned, wholly supported by
customer revenues, and net-profitable.

External Funding Sources

Although the preceding subsection emphasized self-funding, and this chapter in

general has focused on venture funding, there ure other sources of cash, and all
should be pursued with equal vigor. White (1977)presents a relatively large list of
sources, including the aforementioned self-funding and venture capital options.
68 Cash, Financeability, and Control

The important sources, arranged roughly in order of growth stage, are:

1. The founders’ savings, including borrowing on assets

2. Family and friends

3. Formal investment groups, including venture capital concerns and com-

panies that specialize in the private placement of stock

4. Foundations

5. Grants and small-businessloans from various government agencies, such

as the National Science Foundation’s SBIR (Small Business Innovative
Research) contracts for transforming technology into products

6. Having the company’s employees buy equipment and lend it to the firm

7. Obtaining the firm’s capital equipment through bank loans and leasing
companies

8. Forming research and development partnerships with investment compa-

nies to do incremental development

9. University endowments

10. Large companies that enter into a venture-investment phase and pension
funds

11. Strategicpartners that are potential customers and want early access to the
start-up‘s product

12. Strategic partners that are manufacturers whose products would be en-
hanced by the start-up’s product

13. Strategic foreign investors that want access to technology, ranging from
simple distribution to complete rights, through manufacturing

14. Foreign governments, companies, and banks that might be interested in

building a local joint subsidiary to produce or market the start-up’s
product

15. Banking institutions that invest working capital based on firm orders

16. Equipment suppliers and vendors that may help a new company get
started
 Financeability 69

17. Customers, including other start-ups, that may pay in advance for product
or for a development contract (i.e., use someone else’s venture capital)

18. Going public or being acquired by a larger, more cash-rich company

There are, however, some reservations that founders should keep in mind when
considering funding from some of these sources.For example, funding from ”family
and friends” may be desirable during the concept and seed stages. However, these
people will alsobe called on to provide emotional support during those periods, and
it may be too much to ask them for their capital as well.
In the case of venture funding, it is common for venture capitalists to invest in
companies that run out of money and to do so at prices substantially lower than
previous rounds. These financing rounds are called ”cram downs” or ”washout
rounds,” and they have the effect of devaluing the previously issued stock. Even if
the start-up ultimately succeeds, early investors are unlikely to get their money back.
This possibility is another reason why it is critical for the founders to think twice
about taking money from their family and friends.
The third funding source on the list-” formal investment groups, including
venture capital concerns and companies that specialize in the private placement of
stock’-also has its potential dangers. When it comes to securing funds, there is no
such thing as “easy money” or ”dumb money.” The oft-written-about private
placement specialistswho obtain funding from doctors and dentists are seldom able
to deliver either in time or on reasonable terms. Furthermore, many of these
investors can be naive and dangerous. The founders should remember that the start-
up makes a contract with every source of funds and therefore has a contractual
obligation to succeed-an obligation that nonprofessional investors may take quite
literally. Although when all is going well, as in the case of an initial successful
financing, everyone is a friend, when things are not going so well, nonprofessional
investors are likely to turn on the start-up and its founders. Naive investors tend to
build up expectations based on hope rather than on the start-up’s business plan.
When the financial results do not live up to such inflated expectations,the investors
conclude that the company has failed.

Finding a Financial Partner

If self-financing is not possible, founders might be well advised to consider review-

ing the list presented in the preceding subsectionwith an eye to identifying financial-
partnership possibilities, keeping in mind the cautions just voiced about family,
friends, and naive investors. John Grillos has provided the following helpful guide
70 Cash, Financeability, and Control

to the selection process by listing ten attributes of a desirable financial partner.

According to Grillos, the ideal prospective partner:

1. Fits the financing model and strategy for the company

2. Has valuation expectations consistent with [those of] the current owners

3. Has resources to play its role in the strategy-i.e., has ”deep pockets”

4. Has a philosophy that fits with [that of] the company and its current owners

5. Can aid in future financing

6. Has liquidity preferences consistent with [those of] the current owners

7. Can afford to take the loss if the business fails ([which]may rule out family
and friends)

8. [Is] nice to work with and is likely to be with you when the chips are down
9. Understands the business

10. Can contribute [morel than [just]money: consulting, sales, contracts,board

membership, joint venture and strategic partnerships

PRESENTING YOUR CASE

TO A PROSPECTIVE FINANCIAL PARTNER

Almost every entrepreneur who finds venture capital funding advises that the first
few contactsare a learningexperienceand that they will probably result in turndowns
until the entrepreneur’s story comes together. Mike Hackworth, former CEO of
Signetics and current CEO of Cirrus Logic, notes:

You have to tell the story of a company in [the] language of the business you’re in. But
it has to be done in such a way that the financial person can picture it. That means
defining-up front-risks, milestones, and critical dependencies. The market content
is more important than anything; the capitalists assume the technology is there.
Further, investors are interested in getting to know you.

One way for the founders to plan, and subsequently present, a financing
strategy is to pattern it after the strategies of successful ventures that have something
in common with their own firm. In many cases, the people responsible for successful
financing strategies are more than happy to talk about, and relive, their previous
successes. Also, they are likely to want to invest in the proposed start-up if the
 Financeability 71

founders can demonstrate that what they’re doing now is similar to what the
established company did earlier.
As for how long this process will take, the founders should be advised that
financingis more complex and time-consumingthan they could ever have dreamed.
In 1990, the most optimistic scenario for a ”perfect” company seeking venture
financing is that it will take a minimum of three months from the time a preliminary
business plan is available until the cash is in the bank. The amount of time required
depends on such factors as the number of investors involved in the deal, the
financing round (first, second, third, etc.), and the existence and seriousness of any
differences of opinion between the entrepreneur and the investors concerning the
appropriate valuation. Until there is agreement on the value of the company, there
can be no deal. The incredibletime demands of raising money clearly engender poor
behavior and always prompt the firm to go after more money than is actually
needed. This phenomenon is known as Kleiner’s law: “When the hors d’oeuvres are
passed, take two.”
Funding becomes very complex in later rounds, when the terms for the liqui-
dation of the company (it either fails, goes public, or is purchased), based on the
current round and all previous rounds, are written into the financing.

HOW MUCH FUNDING AND HOW MANY ROUNDS?

The amount of funding and the number of rounds are very difficult to determine,
since in each case, the total varies substantially between labor-intensiveenterprises
such as software companies and capital-intensive efforts such as disk or memory
companies.Midway between these extremes is the computer systems firm. By 1990,
a well-run, successful computer systems firm typically required about $50 million
before it achieved profitability. Having $50 million to spend does not guarantee
success, however, because many companies spent that amount (or more) but still
failed. ETA (1983-1989)had an average of four hundred employees and probably
spent on the order of $200 million before it was closed, having shipped a dozen
supercomputers. Trilogy spent nearly $300 million and never could get its technology
to work, let alone ship a computer.
For software companies,self-fundingis best if the founders can manage it. In the
case of small software projects involving a singleteam of five working for two years,
only a few million dollars may be required from start to profitability. In the case of
large software projects involving half a dozen teams of five, a minimum of $10
million may be required to reach profitability. In 1983, Lotus spent $6 million to
launch its 1-2-3spreadsheet and thus preempt competitors.
As for the number of rounds required, a company rarely achieves profitability
with just one or two financing rounds, and the norm is more like three or four
rounds. The funding model advocated herein assumes at least four rounds: seed;
72 Cash, Financeability, and Control

product development with alpha testing; beta testing until market calibration; and
market development, including profitability until the steady-state stage is reached.

FINANCEABILITY FLAWS

When a company has trouble obtaining funding, it is often hard to determine

whether the problem involves a lack of financeability or whether potential investors
believe the firm is inadequate in one of the other dimensions, such as people,
product, or plan. This section deals with the most common exogenous and indig-
enous flaws involving the financeability dimension per se rather than financeability
problems that stem from a start-up’s inadequacies in other areas.
One major exogenous flaw is a dearth of capital in the overall market caused by
a lack of confidence in the economy, region, product, or market. For example, one
large New York- based venture capital company has gone from having a third of its
investments in New England in 1985 to having less than a tenth of its investments
in that region in 1990.During the same period, in contrast, its investments in Silicon
Valley have risen from a third to over half. Another serious exogenous flaw is the
existence of a crowded product area in which the start-up is one of many “me-too”
players.
One major indigenous flaw can arise when the company first starts up if the
entrepreneur and investors cannot come to terms about a fair market value for the
firm. Another major indigenous flaw can manifest itself in later stages when several
rounds of financing have been required because the company failed to meet its
plans; in this case, investor fatigue sets in, and the investors finallysay ”no.” The firm
is then closed, and its assets are sold.

A Dearth of Capital

Unlike most of the other flaws discussed in this book, having the supply of capital
dry up is almost completely beyond the company’s control. Rather, it is caused
either by an overall economic shift toward recession or by a shift in investment
strategies away from certain technology sectors. The result is that no funding is
available, and the start-up’s founders will just have to wait until a more favorable
time.

An Overly Crowded Product Area

Like the dearth-of-capitalflaw, the entry of too many firms into the product area is
primarily the result of circumstancesbeyond a company’s control. This problem is
common, however, because lots of people often come up with the same idea at the
same time. It should be obvious that entering the market with a ”me-too” product
 Financeability 73

under these circumstances is dangerous. Unfortunately, even entering such a

market with a compelling technological advantage may be unsuccessful, since
investors are often wary of a market they consider to be overcrowded.
As an example, I recently looked at a radical (yet simple) disk design that halved
the number of parts required. Although building a radical new disk structure is
risky, the cost advantage was quite compelling. Many firms, including current disk
companies, had examined the patented design and felt fairly certain that it would
work. Despite all these favorable factors, no venture capital company or industrial
partner would fund the start-up. They wouldn’t even fund a seed stage effort to
examine the design in more detail, evidently because they felt that the disk field was
already full and that it would therefore be impossible for the new product to take
market share away from the entrenched leaders.

Failure to Come to Terms with Potential Investors

Although in a few rare cases, the entrepreneur and investors feel equally pleased
with the valuation placed on a company when a round is closed, the norm is for
disagreements to arise over the firm’s valuation. The line between fairness and
exploitation is thin enough that such disagreements can stop the negotiating
process, and a potentially profitable venture may fail to obtain funding. In most of
these cases, the problem stems from the entrepreneur’s having an overly inflated
view of the start-up’s value, and it’s just as well that the firm doesn’t form. In other
cases, the entrepreneur goes on to self-fund the company and is better off without
external funding.

Investor Fatigue Resulting from the Start-up’s

Failure to Exhibit Integrity or Self-Control

The final flaw-a start-up’s lack of integrity or self-control-brings us full circle to

a scenario dominated by indigenous factors. When it comes to financing, the
combined advantages of luck, positive exogenous factors, and a good product can
be totally counteracted by a company’s lack of integrity or by its inability to control
or sell itself. Once these shortcomings become apparent to investors, they can be
expected to lose patience with the firm and withdraw their financial support.

FINANCEABILITYRULES

At any stage of a new venture’s growth, the issue of financeability essentially boils
down to the basic question of whether the firm got the financial support it needed
in order to continue. The rules presented in this subsection are aimed at testing the
start-up’s financing readiness and its quality in the minds of potential investors. For
74 Cash, Financeability, and Control

example, it is critical to have the right experts ”bless” the company. A knowledge-
able expert’s personal financial backing and commitment to spend time count far
more than words coming from a paid consultant.

Are the present plan and people sufficiently compelling to facilitate raising
capital for the seed stage (usually $100,000 to $1million, depending on the
company‘s scope) at the desired price level and also produce a waiting list
of additional investors who want to be part of the start-up round?

Getting financingfor the seed stage is, by definition,the only real test of whether
the concept stage has been successful. Independent of whether the company has
obtained seed stage funding, the following two rules diagnose the likelihood of its
successfully achieving concept stage financeability.

Has the start-up gained the support of at least three reputable, known
outside individuals-persons whose backing would tend to lend credence
to the technology, product, market, and company concept?

The venture must have the support of at least three respected individuals who
are willing to attest to the company’s efficacy and the feasibility of its proceeding to
the seed stage. It is helpful at this point if the outside sources are also willing to invest
their personal capital and time.

By the end of the concept stage, are the critical founders prepared to commit
to a full-time effort during the seed stage?

This second rule for the seed stage tests whether those founders who might be
considered critical have made a commitment to carrying out the seed plan on a full-
time basis. After all, the founders are what the company is selling as a start-up. Their
commitment is usually conditional on obtaining funding-i.e., if the funds for the
seed stage arrive, the founders will leave whatever they’re doing and begin to
develop the plans for the company. However, if the founders are not personally
committed, it is unlikely that investors will be either. In effect, seed stage investors
are buying both a potential idea and a team.

Have the formal business plan and seed stage proved salable, such that an
excess of investors have signed up to provide financing for the next stage
 Financeability 75

(i.e., product development) at the sought-after price, resulting in an

overcommitment of funds?

The principal proof of financeability is as simple as it was for the seed round: the
company got the money.

Was the financing sought at the end of the seed stage in line with the
objectives, milestones, and resources required to complete the product
development stage?

Ideally, enough funds should be obtained to complete all phases of the product
development stage, includingbeta testing.For large development projects,however,
two or more funding rounds may be required merely to finish the product. The first
nonseed round just covers the product specification,basic design, construction, and
preliminary alpha testing. A second round would cover beta testing and the first few
months of the market development stage.

When the company starts up (i.e., at the end of the seed stage), is its valuation
in line with reality as compared to similar endeavors?

The most convincing plan is to present a comparison with other, similar

ventures. Depending on the capital market and the firm’s perceived market posi-
tion, the valuation will be disproportionately higher or lower by factorsof more than
3 (or about an order of magnitude between the highest and lowest valuations) for a
comparable stage. The CEO or financial person should understand how other
companies have done and use this information to set realistic valuation goals across
the board.

Is the funding picture (in terms of availability of funds, state of the economy,
and market and product segment) adequate to sustain the company’s need
for capital? Test: Is the product and/or market area still sufficiently unique
and ”in fashion,” or has the once-”hot” area suddenly become ”cold”
because of an overabundance of suppliers or a long market-gestation time?

During the seed period, the funding picture can suddenly become unfavorable
for the start-up through no fault of its own. Funds can dry up for many reasons,
including investors’receiving requests for cash from companies they started earlier
76 Cash, Financeability, and Control

that are now doing poorly. For example, a semiconductor start-up may place heavy
demands on cash just when the economy and spending take a downturn.

CONTROL: DOING WHAT THE COMPANY SAYS IT WILL DO

Income 20 pounds, expenses 19 pounds, 19 shillings, and 11 pence-result

happiness. Income 20 pounds, expenses 20 pounds and 1 shilling-result
misery.
-Charles Dickens

A system is under control when it is operating predictably by producing outputs in

response to a variety of inputs. The most important element of being "in control" is
first and foremost having a plan that describes the relationship between resource
inputs (time, cash, people, etc.) and outputs (specifications, products, paying
customers, profit, etc.).Without this critical plan (described in Chapter 3),there can
be no control because there is no standard of measure. The second most important
element of control is simply operating the company in such a way as to ensure that
it meets its plan and changing the plan appropriately when exogenous factors
dictate the need for change. In addition to financial control, it is equally important
for the firm to achieve qualitative control, by employing such techniques as
management by objectives (MBO).

THE PROFITABILITY HABIT

The preceding sections of this chapter have emphasized the importance of cash and
have shown that cash depends on successful rounds of financing.In turn, successful
rounds of financing depend on credibility. To achieve credibility, a company must
be in control of what it is doing: it must be able to make a plan and operate according
to that plan. Furthermore, when conditions change, as they will if products or
markets shift, the company must able to adapt quickly. Any firm that wishes to
sustain its profitability must operate according to the following maxim:

Profit is habit-forming. So are losses. Therefore, be profitable from the outset.

PLANNING: THE KEY TO CONTROL

In order for an organization to be in control, it must measure its operations against

a plan. A high-quality company will have a high-quality plan, and that plan will be
a key document for keeping the enterprise in control. A company whose plan is
 Control 77

infeasible or not believed is a company that is inherently out of control. A firm with
no plan at all is really out of control and should consider itself to be in a stage of
unfettered research.

FINANCIAL MANAGEMENT OF THE INPUT AND THE OUTPUT

The financial portion of the plan is a major part of the firm’s control dimension. A
company’s financial operations have two basic components: the input side and the
output side. The input side is the cash generated by the financing rounds. The key
test of input-side control is whether the chief financial officer understands the plan
and is in control of spending the company’s cash according to that plan. To achieve
control, the company must establish control mechanisms for hiring (salaries and
stock), consulting personnel, other temporary personnel, benefits, office expenses,
capital equipment, purchasing, travel, and entertainment. The output side is the
firm’s production of goods and services. The key test of output-side control is
whether the company produces its contracted output-e.g., completing projects,
delivering against purchase orders, and building the agreed-upon products at the
right price and on schedule.
To achieve control of the input and output sides of the company’s finances,
every part of the organization (and nearly every individual) must plan, operate
according to the plan, and adapt to changing conditions. The best, and perhaps the
only, way to achieve this sort of control is to establish formal systems whereby every
individual makes weekly objectives that support the overall plan and reports on
them. Management by objectives,management by exceptions, or some other control
scheme is necessary.
The organization’speople can learn to plan and operate by the numbers if they
are required to measure themselves against a concrete numeric standard. For
example, I measure an engineering group’s planning ability by a single number, the
schedule fantasy factor (SFF),which is calculated by dividing the actual time it takes
to achieve a given milestone by the planned time for achieving that milestone.

MANAGEMENT COMMUNICATIONS AND CONTROL

The company must hold staff meetings and minutes should be taken to chronicle the
topics discussed and the resulting ”action items.” The easiest way to tell how a firm
operates is to examinethe minutes of its staff meetings and review thereports of each
critical function. Are these meetings held on a regular basis? Does interfunction
communication occur? Is the performance of each function tracked? Are critical
issues identified and resolved? Does the company somehow deal with every crisis
rapidly and efficiently,or is it in a constant state of crisisbecause problems rarely get
resolved?
78 Cash, Financeability, and Control

Decision making must be rapid, yet not precipitous. I believe that every major
decision should be made over at least a one-day period, with time allocated for
reconsideration.When changesin the situation necessitate changesin the company’s
plan, the new plan must be carefully thought out and appropriate in light of the new
situation; otherwise, the plan is foolhardy.
The key to planning is easily stated (as I did at Digital in 1973):“He who plans,
does.” The only way to achieve total commitment to carrying out a plan (control)is
for everyone responsible for the plan’s execution to participate in its formation. The
staff knows when it has a decision or a plan that it can implement, and will commit
to such a plan. In contrast, the staff usually recognizes an unrealistic and ill-
conceived plan, and will not give it their full commitment; the plan will then have
to be remade.

THE LINK BETWEEN PROFITABILITY AND CONTROL

Concern for profitability is the basis of control. As noted earlier, profit is habit-
forming. Concern for it must therefore pervade the organization from the day the
doors are opened and the company starts spending money. Everyone has to
understand that every dollar spent must ultimately be repaid by product revenue.
Although profitability, like quality, has to come from the top, it must also be
ingrained in every single individual in the organization. It is to every employee’s
benefit to keep profit in mind, because a chronically unprofitable company is
generally an unhappy place in which to work. In particular, the founding team
knows that the firm is being fundamentally dishonest and deceiving itself if it must
keep creating a succession of new plans to convince investors that profitabilityis just
around the corner. Furthermore, this dishonesty is likely to repeat itself, since an
unprofitable company must return to its funding sources over and over again for
more cash. Each trip to the investors will be increasinglyunpleasant and very time-
consuming, because the investors will ask nagging questions about those earlier
plans for profitability.Also, unless the CEO has a stranglehold on the firm, through
ownership or technology blackmail, sustained unprofitability will ultimately cost
his or her job.
An often-successful way to improve the chances for profitability is to delegate
the responsibilityfor it below the CEO. Thus, several people share the responsibility,
and the CEO merely helps them achieve the desired goal. This argues for a quasi-
divisional structure, a technique used during Digital’s period of greatest growth
(1966- 1984). During that time, the company was organized around a collection of
about twenty “product lines,” which were responsible for various market segments.
The segments included professionals (e.g., laboratory, engineering, industrial
control, commercial); customers (e.g., government, small business); buying chan-
 Control 79

nels (e.g.,original equipment manufacturer [OEM], components, retail); and prod-

ucts (e.g.,DECsystem-10).This structure allowed a series of closed loops to control
the allocation of resources and the delegation of profit responsibility. With the
advent of the VAX architecture, all customers were buying similar products. The
opportunity to differentiate the product and to address the respective markets
diminished, and the organization abandoned the product-line structure in favor of
a more traditional functional organization. Profitability then depended upon over-
all control of every organization. In 1990, DEC once again began to reorganize
around a product-line structure in order to be able to delegate control and respon-
sibility for profitability. ,

CONTROL FLAWS

As noted earlier,controlinvolveshaving a good plan and carrying it out. Conversely,

control flaws involve having a poor plan and/or being unable to carry it out. The
planning flaws range from having a foolish plan to having a plan that the company
is not confident it can execute. Although the control flaws usually stem from
shortcomings in the way the organization manages itself, the firm may be doing
everything it thinks is right and still fail to meet its sales projections. Other control
flaws range from having total anarchy to being overly bureaucratic. The CEO and
team are at the heart of a start-up’s (in)ability to be in control.

Trying to Operate with a Plan That Has Lost Touch with Reality

A company may continue to spend according to an operational plan that has become
unrealistic, oblivious of the fact that it is failing to meet important milestones.
Although unrealistic plans were discussed in Chapter 3, “The Business Plan,” this
flaw is worth mentioning in the present context as well, because it is also related to
the control dimension, a relationship best illustrated with the following true story.
In early May 1990,I attended a Dutch-treat dinner celebratingthe end of the seed
stage for a company about to enter the office automation market. The team had
performed admirably during the concept stage and had come up with a good plan
for a product and a company. The CEO had obtained ad ice from a small group of
I

competent friends in the industry (called the KBOD, for “kitchen board of direc-
tors”) and had conducted numerous interviews with potential strategic partners
and customers to help develop the product specification during the eight-month
concept stage.
When the firm proceeded to the seed stage, it continued promoting the product,
working on selecting early beta sites, writing draft manuals, etc. However, it had
created a complete fantasy. On the one hand, the company appeared to have a lot
80 Cash, Financeability, and Control

going for it. It had a team with a board (KBOD), a product spec, a demo (which it
came to believe was the product), a process for improving the product, customers,
and a support team. The only thing the company lacked was an actual product.
Worse yet, it continued to be unable to hire a person to be responsible for building
the product. The firm was spending its precious seed money and time (using up
credibility with its investors by jeopardizing its seed plan) to do work that was
irrelevant at that stage. It was simply addressing the wrong problems, and the
various parts of the organization were out of synchronization with one another.
Furthermore,by broadcasting its plan widely, the company was giving competitors,
both potential start-ups and existing firms, an opportunity to build the product first.
Fortunately,the company recognized the flaw and by fall that year was on track with
two product developers who produced a great prototype to use in closing the
product development funding.

Having No Measures or the Wrong Measures

A company is, and gets, what it measures. A firm that operates with no measures or
with the wrong goals or measures is likely to produce either nothing or the wrong
thing. Each part of the organization, especially engineering, must have appropriate
measures, such as schedule and product quality. The Ardent product development
story (page 122)illustrates a case in which focusingon a singleproduct-performance
metric nearly proved fatal for the company. Every department that concentrates on
just one metric to the exclusion of all others runs the risk of being similarly flawed.
For example, a customer service organization may assess its performance only in
terms of an overall customer service index and not bother measuring the myriad
factors that are reflected in that rating, such as response time, mean time to failure,
and number of outstanding errors.

Inadequate Financial Control

When a company operates without adequate financial control, it may embark on a

spending spree not covered by its operational plan. For example, one CEO went out
and bought an expensivecomputer-aided design (CAD)system, using 15percent of
the start-up funds, as his first executiveact. It was a system that his engineering team
neither wanted nor needed. Later on, the firm purchased an expensive tester that it
could have rented. The organization was never brought under financial control
because the investors were always willing to provide additional funding in return
for promises.
 Control 81

Overly Bureaucratic Planning and Decision Making

The opposite of an out-of-control organization is an organization controlled so

tightly that it cannot spend enough time generating output. Instead, it wastes
valuable time adhering to procedures that require needless forms and approvals,
appearances before committees and working groups, etc. Although bureaucracy
usually goes hand in hand with large organizations, start-ups can often evolve
rapidly into ungainly bureaucracies. Because engineers frequently complain about
bureaucracy, one would expect a venture led by engineers to be streamlined and
efficient. This is not necessarily the case, however. In practice, engineers are
inherently the world’s greatest bureaucrats because they are so good at designing
organizations and processes.

Lack of Support for the Plan

The only way to ensure a company’s commitment to the operating plan is to have
its entire staff participate in making the plan. Any other planning approach will
result in an out-of-control situation. Thus, companywide support for the plan is a
necessary condition (though not a sufficient one) for achieving control.

Top-Level Flaws in Planning and Decision Making

Planning and decision-makingflaws that originate at the top can take several forms.
An autocratic CEO or department head may insist on being involved in every
decision. An anarchic CEO or department head may fail to make timely decisions,
believing instead in the ”Bo Peep” school of management (”Leavethem alone and
they’ll come home. . . ’’I. A mercurial CEO or department head may make decisions
capriciouslybecause they are fun to make, oblivious to the fact that all the decisions
must then be remade. Or a top-down CEO or department head may make a plan that
lacks “buy-in” from those who will be responsible for its implementation.
I know of one CEO who was a fine leader and salesman for the company but
possessed all of the above-mentioned traits to some degree. This individual made
certain decisions in an autocratic and mercurial manner, was unaware of some
important decisions, was inconsistent about management practices, and made top-
down plans and pronouncements that no one could believe in or carry out. I know
another truly anarchic CEO who was simply dumb, perhaps lazy, and an inept
manager and leader, but he didn’t possess any of the above-mentioned flaws. This
may have made him somewhat less dangerous, because his team rallied and was
able to manage itself.
82 Cash, Financeability, and Control

Experiencing a Major, Unexplained Slip in the Plan

The company’scredibilityis establishedby how well it meets its commitments-Le.,

how successfully it achieves its plan. A major slip can occur, though, when the
assumptions about a product or market turn out to be invalid, and the firm then
attempts to recover or respond to the new information.Slipping the schedule during
the seed stage is not a fatal flaw if it happens for good reason, since during the seed
stage, the organization is still permitted to engage in open-ended work (including
advanced development and market research).
However, such a slip raises a couple of important questions: Is everyone in the
company operating according to some formal schedule and management by ob-
jectives? Is everyone informed about the firm via effective staff meetings during
which review, direction setting, and problem identification/resolution take place?
If these questions cannot be answered affirmatively, the slip could be the first sign
that the organization is operating in a potentially open-ended fashion.

CONTROL RULES

The following rules provide the guidelines for evaluating whether a new venture is
in control at both the concept and seed stages. In these early stages, it is difficult to
determine whether or not a company is in control. In fact, at the concept stage, the
only real indicator is whether the CEO and individual team members have a history
of being able to perform their respective functions in a controlled fashion.At the end
of the seed stage, however, the firm’s ability to control itself can be measured in
terms of how it performed during the seed stage.

At the concept stage, does an examination of the founders’ reputation and

past achievements disclose solid evidence that they are capable of accom-
plishing the following?

Hiring top-quality people

Demonstrating technical and marketing expertise

Producing successful products and planning effectively

Achieving schedule milestones and meeting budgets

Although control is hard to measure at the concept stage because there does not
yet exist an actual company with committed resources, a diligent effort should be
made to ascertain whether the founders have run comparable firms successfully in
 Control 83

the past and whether the group is likely to be able to meet its own timetables, both
for the seed stage and for subsequent stages. Thus, it behooves a start-up to
accurately record its progress in achieving the goals to which it has committed itself.
This gets the company into the habit of clearly understanding and continually
monitoring its own abilities and accomplishments so that it can be ”in control”
during later stages.

Did the team meet its timetable for making the seed plan?

By the end of the concept stage, a preliminary estimate can be made of whether
the company is likely to be able to meet its commitments and remain in control. If
the firm can’t plan effectively enough to create its seed plan on time, there is little
reason to expect successful results from any other activities that require planning.

Has the company been operating according to an overall plan, and has that
plan been changed only minimally during the seed stage?

The first part of control is having a plan. The second part is sticking to that plan.

Were seed stage objectives met without major milestone slips? If milestone
slips did occur during the seed stage, were the backers’ expectations man-
aged and recast to their satisfaction such that they are willing to continue
investing during the product development stage?

The new venture’s track record during the seed stage is likely to be the best
predictor of its future performance.

Does everyone in the company operate according to some formal schedule

and MBO? Is everyone informed about the company via effective staff
meetings during which review, direction setting, and problem identifica-
tion/resolution take place?

The firm’s credibility is first established during the seed stage, based on how
well it meets its commitments. However, a major slip can occur if the assumptions
about a product or market prove invalid and the company attempts to recover or
respond to the new information. As mentioned above, slipping the schedule during
the seed stage is not a fatal flaw if it occurs for good reason, but an inability to answer
these two questions affirmatively could be the first indication that the organization
is starting up in a potentially open-ended, out-of-control fashion.
84 Cash, Financeability, and Control

Does the company have a controller, or a control mechanism, with budget-

ing, hiring, and spending processes in place to manage cash for both the
current and next stage and control its spending as it enters the product
development stage?

When the company opens as a full-scaleentity, it is likely to be flush with a large

amount of cash ($5 million to $10 million). At this point, there will be an enormous
pent-up demand to hire direct and support staff and buy equipment, parts, and
computer resources. Unless the firm has controls in place or is operating according
to a detailed budget, it can easily begin spending vasts sums right away for critical
items and thus find itself out of control from the moment it starts up.

CONCLUSION

Cash, financeability (the ability to get more cash), and control (theability to produce
results, including profit, with the cash and resources a company has) are all inter-
related.
Without cash, the venture cannot proceed beyond the ”kitchen table” planning
stage. At least one founder must be self-supporting while the business plan for the
seed stage is written. Later on, when the founders have spent several months in the
seed stage planning the company and writing the business plan, the start-up must
have a large enough cash cushion to wait out a period of at least three months while
financing is sought. However, even if an enterprise has the necessary cash cushion,
if it lacks control, its cash will ultimately decline to zero and the firm will be
unfinanceable.
Without financeability, a company cannot continue to implement any plans or
vision that it may have. Financeability is determined by both exogenous and
indigenous factors. Although the firm can only respond to what it believes are the
exogenous factors, it has total controI over the indigenous factors-its business, as
embodied in the technology it selects, the product it builds, its plan, its people, and
other dimensions. The best guarantee of financeability is for the start-up to be in
control and to have adequate cash through planning and profitability. Getting out
of control and starting to deplete cash will put the venture on a downward spiral.
Control is the start-up’s combined ability to make an effective overall business
plan and then be able to operate in such a way as to achieve the plan. Control is
measured quantitatively by how closely the company’s actual operations match its
plan in both the expenses and revenues lines on the profit and loss statement.
Control is also measured qualitatively by how the company manages itself with
respect to the objectives it has established for products, employee satisfaction,
service, etc. Being in control is at the heart of preserving cash and financeability.But
being in control is moot if the company is out of cash and not financeable.
Chapter 5

TECHNOLOGY AND
ENGINEERING

The technology and engineering dimensions of a high-tech venture are so important

that it takes two chapters to describe them fully. This first chapter covers the role of
technology in product development, technology progress in logic and memories, and
various aspects of technology creation and transfer. The following chapter covers
technology and engineering flaws and presents the ”technology balance sheet,” a
framework for analyzing a company’s technology and engineering.
The technology dimension is reflected by a firm’s abihty to assimilate and utilize
scientific and engineeringknowledge as embodied in components,processes, and the
”know-how” of its people. The engineering dimension is reflected by its ability to
produce specifications(oractual informationin the case of software)for a manufactur-
ing organization. Since technology is the basis for engineering, technology will be
discussed first, starting with its role in product development.

TECHNOLOGY

The technology needed to develop products can come from a range of sources.When
technology emerges solely from science and engineering, the technology is pushing
products into the market. In contrast, when technology is required in order to satisfy
needs, the market is pulling to create technology. One difficulty with new technology
is its acquisition. I firmly believe that the best way (and in some cases, the only way) to
transfer technology is to transfer the people associated with the creation of that
technology. Because technology is rarely measured, many companies start up without

85
86 Technology and Engineering

knowing how much technology they need, how long it will take to acquire it, or how
much it will cost. The foundersof a new venture must understand the firm's technology
well enough to measure it. The following subsections should help them achieve that
level of understanding.

TECHNOLOGY PUSH AND MARKET PULL

Product development is usually characterized as involving either technology push or

market pull. According to the technology-push model, products originate with a flow
of ideas that starts in research, progress through advanced development and product
development,and ultimately reach the customer.According to the market-pull model,
products originate with customers, who specify their requirements to a marketing
organization,which, in turn, tells a developmentorganizationwhat to design and build,
with research and advanced technology playing only a minor role. Strictly speaking,
neither of these two models is correct, even for limited classes of products. Companies
that operate exclusively according to either of these models are doomed to fail,because
product and market responsibility must be disseminated throughout the entire orga-
nization in order for top-quality products to be created.
During the past decade, every company has attempted to characterize itself as
"market-driven." Unfortunately, that often translates to "marketing-department -
driven." In such a firm, the marketing department talks to some users and comes to the
engineering organization with a comprehensivelist of requirements for the proposed
product. The list is inevitably embellished with marketing's own ideas about how the
product should be designed,sincemany high-techmarketingpersonnel are alsoformer
engineers. Products that are specified in this fashion have the following predictable
attributes:

They are priced lower than all current competitors.

Their performance is greater than that of any existing product.

Their features and functions represent the total of those of all existing products.

They possess unique, discriminatingfeatures designed to enable them to "knock

off" other products, often by differentiating themselves just enough to allow
government buyers to avoid competitive procurement.

The desired delivery time is yesterday.'

1. Marketing wants it yesterday, engineering will have it tomorrow, and science is still working on it.
 Technology 87

Engineering is expected to design and specify a product that satisfies these

requirements and conforms to the original design provided by the marketing depart-
ment. This approach to designing and building a product ignores technology and
engineeringinnovation.Worse yet, it relegates the engineeringorganizationto the role
of a nonthinking automaton that simply builds products in response to what other
companies produced several years ago. Without question, a firm driven solely by its
marketing department is fatally flawed.The followingtwo stories, the first apocryphal
and the second true, illustratethe potentialperils of basing a product on marketingideas
that lack sufficient engineering input.
Let’s start with Ken Olsen’s tale of a revolutionary new wallpaper remover.
Accordingto Ken, a marketingperson approached an engineer one day with an idea for
a very powerful wallpaper remover that, when applied, would immediately take the
paper off, leaving the wall spotless and ready to paint. The impressed engineer
exclaimed: ”Great, let’s build a plant to manufacture and sell it. We’ll get rich! What’s
in it?”Replied the marketing person: “Darned if I know. I thought up the idea. You tell
me what’s in it and design the plant to manufacture it!”
Then there’s the tale of the ”recordingwhteboards.” In early 1986, just before the
arrival of ”recordingwhiteboards” from Japan,I was asked to become involved with a
concept stage venture that proposed to manufacture such a product. The start-up
wanted advice about its design. Two marketing people, with virtually no technical
background, had identified the need for the product and had just spent six months
investigatingthe market forit.They presented a plausiblecasefor why it would be such
a great product: who would buy at various prices, how many they would buy, and how
it would be sold. We then got into how the product would be built and what it would
cost. It turned out that the marketing people who originated the idea had little
understanding of technology and cost. After a few phone calls, we ruled out their
approach (using conductive fiberboard) because of its resolution and questionable
reliability.I described a scroll-scanningapproach, made possible with new photodiode
arrays that were part of the fax revolution.They felt that this would be too expensiveand
too complicated. Fortunately, Japanese manufacturers introduced the product I de-
scribed before the company was able to obtain funding or hire gullible engineers who
might have been tempted to try its flawed approach.
These storiesare not intended to absolveengineeringfrom allblame, though. Firms
driven solely by their engineeringdepartmentscan fail because of an inability to satisfy
the customers’ needs and are therefore also flawed.
The successfulstart-up must strike a balance between marketing and engineering.
Striking a balance, however, does not mean eliminating legitimate and healthy com-
petition between the two organizations.Each organizationmust fulfillitsresponsibility
with wholehearted commitment and a high level of drive. The conflict that inevitably
88 Technology and Engineering

arises when engineering’s product definition clashes with marketing’s requirements

often produces precisely the spark needed to generate premier products. The most
appropriaterolesfor the engineeringand marketingorganizations,and for the CEO, can
be summarized as follows:

Engineering: The engineering organization must acquire the technology and the
engineering talent. It must then design the product to meet the cost and schedule
goals at the highest possible quality level.

Marketing: The marketing organization must identify the product require-

ments, including what (the price, performance, features, and functions), why
(thebuying rationale, expressed in terms of benefits),and who (precisecustomer/
application profiles).
In the last stages of development, marketing must also deliver the necessary
support material for the resulting product, so that the sales organization can
generate profitable orders.

CEO: The CEO must arbitrate conflictsand deadlocks between engineering and
marketing to arrive at a common product and marketing plan and ensure that
each group carries out its respective responsibilities.

TECHNOLOGY PROGRESS

Technology progress, which can occur in both an evolutionary and a revolutionary

fashion, results from twobasic factors: ( 3 ) the increased density of semiconductorsand
magnetics and (2) the quest to build and exploit computers with new applications.
Additional factors that drive progress include all the forms of research, development,
and manufacturing.

Revolution and Evolution

Figure 5-1 shows two models of progress (Gomoryand Schmitt, 1988).One model is a
”ladder” of scientific revolution based on important milestones in computer technol-
ogy, while the other is a ”wheel” of evolutionbased on continuousrefinement of a basic
design or process. The ”rungs” of scientific revolution are somewhat arbitrary. Fur-
thermore, the dates given are for the introduction of a particular technology into
computers, not for the initial availability of the technology itself. For example, vacuum
tubes were used in radios long before 1944. These observations aside, the most
interesting aspect of the ladder is that it shows no computer-technologyrevolutions
 Technology 89

Year Of First Use Milestone

In Computers
Fabrication Architecture, design
Molecular?
&lor bio-?
technology
. tools, construction,
and manufacturing

Optical?

Integrated Use
circuit

Transistor
/
New

c
needs
stimulate
1946 Stored program
1944 Vacuum tube
\ Sell and build in
high volume
Evaluate and
Electromechanical understand

Turbine wheel model of continuous evolution

1930
1890 Mechanical

Lad er model of scientific revolution

Figure 5-1. Revolutionary (Ladder) and Evolutionary (Wheel) Models of Technology

Progress

since the introduction of integrated circuits in 1967.Although optical technology (now

used extensively in communications) may eventually find its way into computers,
products based on this technology are unlikely to appear during the 1990s,since there
is a substantial delay between laboratory development and product introduction. I
quantify the typical laboratory-to-product lag like this: roughly a decade (or one
technology generation) usually elapses between the time a significant laboratory
inventionoccursand the time that inventionis used to any significantextent in products.
Despite the fact that marketing organizations often use the term revolution to de-
scribe new products or developments in computing, evolution is a more realistic word,
because progress is generally based on well-establishedtechnology and a set of design
principles. In particular, circuit and memory technologies (i.e., the technologies in-
volving the physical components that actually process and store information) are the
key determinants of a computer’s performance and cost, and during the past twenty
years, progress in these technologies has been considerably more evolutionary than
revolutionary. Unfortunately for the United States, which excels in revolutionary
inventions,Japan excels in evolution.
The cycle of evolution in computer technology is driven by the interactionof many
processes.New basic materials and circuits,along with advancesin fabricationtechnol-
90 Technology and Engineering

ogy, make possible new architecturesand new ways of producing the next computer.
The process of selling,building in higher volumes, using, evaluating, and understand-
ing computers raises aspirations for the next cycle of evolution. Some of these factors
involve computer manufacturers, some involve users, and some involve the formal
study of computers in computer-sciencecourses.With the advent of increased capabili-
ties comes the discovery of new uses and needs, which unleashes more funds to fuel the
next cycle.

Semiconductorand Magnetic Density Evolution

Many developments have permitted the computer to evolve rapidly, the most impor-
tant being density increasesin semiconductorsand magnetics.Althoughimprovements
in these technologieshave been evolutionary (i.e.,conforming to the ”wheel” model in
Figure 5-1),their impact on computer architectureand applicationshas paved the way
for revolutionarychanges (i.e.,conforming to the ”ladder”model) in those areas.At the
present rates of progress in semiconductors and magnetics, the cost of hardware for
computers of the type and sizecommonlyused today will be near zero by the end of the
century. Semiconductorpeople often make the analogy that ”If cars evolved at the rate
of semiconductors, we would all be driving Rolls Royces today that go a million miles
an hour and cost $0.25.” The difference lies entirely in the technology: Maxwell’s
equations governing electromagnetic radiation, which moves at the speed of light,
versus Newton’s laws governing the motion of objects with mass, which move at far
slower speeds.
The integrated circuit was invented in 1958, the year when discrete transistors first
started being used in computers. Every year from 1958until about 1972,the number of
transistorsper diedoubled.Startingin1972,the numberbegandoublingonly everyyear
and a half, or increasing at roughly a rate of 60 percent per year, resulting in a factor of
100 improvement each decade. Gordon Moore of Intel posited two laws based on this
phenomenon:

Moore’s law (1964): The density of chips doubles every year.

Moore’s law (1975): The density of chips doubles every 1 ‘I2years.

In recent years, the use of memory circuits that require only one transistor per bit
stored (plus some capacitance)have made bits per chip rather than transistors per chip
a more interestingmeasure, but density has continued to double every year and a half,
which means that it quadruples every three years. This three-year pattern is illustrated
by these statistics on the number of bits per chip and the year in which each chip was
 Technology 91

introduced: lK(1972),4K(1975),16K(1978),64K(1981),256K(1984),lM(1987),and4M
(1990).The following equation applies (note that t equals the current year):

Number of bits/chip = 1K X 2''~1972)/1.5

This trend seems likely to continueuntil the year 2000, when extrapolationsuggests
that a single memory chip will store 256 million bits. The 256-million-bitfigure may be
slightly optimistic, however, since Meindl (1987)predicts that growth will slow down
from60percent tobetween20percent and 35percent beginningin 1992- 1998.However,
Meindl sees 20 percent to 35 percent growth persisting for another twenty years, in
which case, a single die will store between 1 trillion and 100 trillion bits.
Both this past history and the future of the entire industry can be seen in the
following graphs. The first graph (Figure 5-2), based on data from Intel, shows the
number of transistorsper die forvarious-sizememoriesand microprocessorsduring the
period 1970- 1990and projects the growth in density through the year 2000. The graph
indicates a logarithmic increase in density over time. This has allowed computers to
operate faster while costing less, because of the following two rules:

The smaller everything gets, approaching the size of an electron, the faster the
system behaves.

Miniaturized circuits manufactured in a batch process tend to cost nothing to

produce after the factory is in place.

The cost impact of the increased densitiesshown in Figure 5-2 is reflected in Figure
5-3, which shows changes in the relative cost of scientificcomputing from 1950 to 2000.
The cost has declined over five orders of magnitude during that period, representing a
price drop of 20 percent per year.
Not all of the cost benefits of increased memory chip density have translated into
a reductionin system cost, however.Someof the cost benefitshave translated into larger
memories, since the advances permit a given computer to have more memory for a
constant price. In the forty-five-year period shown in Figure 5-4, primary memories
have grown by over six orders of magnitude, representing an increase in size at the rate
of 35 percent per year.
In summary, the semiconductordensity evolution has been extremely dramatic. It
has spawned whole new classesof computers,new computer systems,new companies,
and new opportunities, many of which are discussed later in the chapter.
However, semiconductor memories are only one part of computer memory
systems,which can be thought of as a hierarchy (Figure5-5).Information pertaining to
92 Technology and Engineering

109

Microprocessor
lo"
A Memory

107

.-
0)
U
L.
%
g
Y
106
.vYl 2501

3
10s

1
l(r

8080
8006
'4004
lo! I I
1970 1980 1990 2000

Figure 5-2. Transistors per Die Versus Time for Various-Size Memories and for Several
Intel Microprocessor Chips. (Courtesy of Intel Corporation.)

a present computation is stored in fast registers that are part of the central processing
unit (CPU), while recently referenced informationis held in cache memories. Informa-
tion referenced less often is stored in primary (semiconductor array) memories.
Infrequently referenced information is stored using electromechanical technologies
that record information on magnetic disks, magnetic tape, and electro-optical media.
Although each lower level in this technological hierarchy is characterized by slower
accesstimes, the cost per bit stored is correspondinglylower.Technologyprogress at all
levels of the hierarchy has driven down the price of computing systems memory, as
indicated in Figure 5-6.
 Technology 93

Derived from a graph by Victor Peterson

NASA Ames

10

0.1

0.01

0.001

0.0001
1950 1960 1970 1980 1990 2000
Figure 5-3. Relative Cost of Computation Versus Time for Leading-EdgeScientific
Computers.(Courtesy of Askmar. Reprinted with permission.)

Just as increasing transistor density has improved the storage capacity of semicon-
ductor memory chips, increasing areal density2 has directly affected the total
mformation-storage capacity of disk systems. Figure 5-7 shows lines of constant areal
density for disks and tapes. Notice that IBMs 1957disk file, the 350 W A C , recorded
about 100bits along the circumferenceof each track and each track was separated by 0.1
inch, giving an areal density of 1,000bits per square inch. In early 1990, IBM announced
that one of its laboratories had stored 1billion bits in 1 square inch. This technology

2. The amount of information that can be stored per unit area.

94 Technology and Engineering

100 GB

10 GB
c
1 GB

8 100MB
.#
tn
t
10MB
3;:
?!
E
.#
2 1MB

100 KB

10 KB

11111111111111111111llllllllllllllllllllllll
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Figure 5-4. Primary Memory Size Versus Time for Mainframes and Microcomputers.
(Courtesy of Askmar. Reprinted with permission.)

progression of six orders of magnitude in tlurty-three years amounts to a density

increase at a rate of over 50 percent per year.
Increases in areal density have led to magnetic storage systems that are not only
cheaper to purchase but also cheaper to own, primarily because the density increases
have markedly reduced floor-space requirements (which are a substantial item of
expense in many environments).Figure 5-8 shows changes over time in the amount of
information that can be stored in various-size disk memories. The first disk files
occupied over 2 square meters of floor space but held only 5 megabytes of information.
 Technology 95

Larger Capacities
Slower Access Times

Figure 5-5. The Various Technologies That Form the Computer Memory Hierarchy.
(Coutesy of Askmar. Reprinted with permission.)

By the mid-l960s, large disks occupied less than a square meter of floor area. With the
introduction of 8-inch-diameterdisks (not shown in the figure)in the mid-l980s, six 500
megabyte disks could be rack-mounted in a cabinet occupying 1square meter of floor
area, a twelve hundred- fold improvement over the early disks.
Modern 5 'I, -inch and 3 '1, -inch drives can be mounted within a workstation, and
without such high-density disks, the modern workstation environment would be
impossible. In 1990, a 2 -inch 20-megabyte disk drive occupies an area of less than 10
square inches at a height of less than half an inch, permitting the disk to be mounted
96 Technology and Engineering

t Derived from a p h b Frank Ura

&ewttt-Packard

Mainframe DRAM

\
:y
14" Disk
IBMDrives \
WORM
0
\
Midrange DRAM

Microcomputer DRAM

DRAM Chips

1970 1975 1980 1985 1990 1995 2000

Figure 5-6. Price of Primary (Dynamic RAM), Secondary (Disk), and Tertiary (Write-
Once) Memory Versus Time. (Courtesyof Frank Ura, Hewlett-Packard.)

entirely on a printed circuit board and thereby making laptop and notebook-size PCs
possible. Soon,electro-opticaldisk technologieswdl provide a gigabyte of disk memory
at the cost of a compactaudiodisk,makingit economicallyfeasiblefor PC or workstation
users to have roughly four hundred thousand pages of pure text or ten thousand pages
of pure image data instantly available.In short, along with semiconductorsand display
devices, disks have been a key enabling technology for a number of computer classes,
including PCs, workstations,and laptops.
 Technology 97

1M

lOOK

10K

1K

100

10
1 10 100 1K 10K lOOK
Track Density per Inch (TPI)
Figure 5-7. Density for Various Secondary (Disk) and Tertiary (Tape) Memories.
(Courtesy of Askmar. Reprinted with permission.)

THE ROLE OF TECHNOLOGYPROGRESS IN FORMING START-UPS

In the previous subsection,technology progress was said to be the result of two factors:
the increased density of semiconductors and magnetics, and the quest to build and
exploit computers with new applications.That is perhaps a simplistic view, however,
because the synergisticrelationshipbetween start-ups,established companies,research
labs, and academia plays an equally important role in technology progress. Defining
98 Technology and Engineering

E ASKMAR Printed Circuit Board Mounted

Rack Mounted
10 GB
Floor Mounted
11

b
.I

%P, 1GB
J
k
iz
5 1°0MB

10 MB

IBM 350 5-1/4"

1 MB
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Figure 5-8. Secondary (Disk) Memory Capacity Versus Time. (Courtesyof Askmar.
Reprinted with permission.)

that relationshp is a bit like addressingthe 'Which came first, the chicken or the egg?"
question,because established companies, research labs, and academia create advances
in technology,which provide the impetus for the formationof start-ups. Start-ups then
create additionaladvances in technology,which are further researched and developed
by established companies, research labs, and academia.Tlus process is consistent with
the "wheel" model of product evolution shown earlier in the chapter (Figure 5-1).
Start-upsform to exploit the challenge of a new product idea that is based on one
or more of the following (examples are shown in parentheses):

Basic or applied research performed at university, government, or industrial

laboratories (Valid Logic: simulation, timing verification, and design; Silicon
 Technology 99

Compilers: automated chip design; NChip: multichip packaging; and Silicon

Graphics:3-D workstations)

Applied research directed toward the development of a new product (voice

typewriter)

A new manufacturing process (metal oxide semiconductor [ M a ] )

A new component (integratedcircuits;microprocessors,including RISCs [reduced

instruction set computers])

A new architecture (parallelprocessing)

A new standard (Ethernet, ISDN, UNIX)

A new de facto standard that fills an early need (AdobePostscript)

A new paradigm for computing (Visicalc,Hypecard)

A new generic application made possible by a new computer (word processing)

A new professional application (movie making, molecular modeling)

A new user-specific product or requirement (GMs MAP protocol)

A new military or government requirement (ADA, Posix, VHDL)

User-developed software that serves as a demonstration, first prototype, or first

release (Nastran, DBASE 11)

To explore each of these factors in depth would require hundreds of pages, but a
quick review of several of them would be useful. Since the first two items on the list
concern research, let’s begin there.
The revolutionary and evolutionarychanges in technology discussed earlier in the
chapter have been the result of research conducted in a number of environments:
government-funded research in universities, government laboratory research, corpo-
rate research, and individual research. The following subsections briefly examine the
technologicalchanges that have resulted from each of these types of research.

Government-Funded University Research

The great new form of computing have come from government-funded university
research. DARPA (DefenseAdvanced Research Projects Agency)has been the primary
funder of large projects, while the NSF (National Science Foundation) has funded
100 Technology and Engineering

smaller projects, university infrastructure, individual researchers, and the training of

engineersand scientists through research projects. Table 5-1 shows some of the critical
inventions that have been produced by government-funded research.

Government Laboratory Research

Sevenhundred national laboratoriesin the United Statesand a smaller number in other

nations are important sources of research activity and technology training. Occasion-
ally, they also provide prototypes or ideas that stimulateproducts. For example, a large
number of algorithms and programs for scientific computing have come from the
Lawrence Laboratories at Berkeley,Livermore, and Los Alamos. NASA's laboratories
have pioneered many programs in computational fluid dynamics, and the standard
program for finite element analysis (Nastran)is a product of NASA. In the computer-
hardware area, NASA stimulated the development of the first integrated circuits
because it needed compact, powerful computers. Several of the NASA products were
spun off and further developed by private industry. For example, Nastran was
transferred to McNeal Schwindler Corporation for continued development and com-
mercialization. Similarly, Ashton-Tate exploited JPLDIS, produced by research at the
NASA-funded Jet Propulsion Laboratory, by reimplementing the Fortran version in
assembly language and marketed it as DBASE 11.
In many cases, laboratorieshave served as knowledgeableand demanding users,
and companies such as Control Data Corporation (CDC) and Cray Research have
formed to build computers that would meet the laboratories' needs. These laboratories
have been the first users of nearly all innovative computers and hence have supported
the industry by acting as risk-taking customers. If government laboratories had not
played this role, high-performance computers and many new technologies simply
would not exist.
Althoughsuccessinthe 1990swillrequirebeingcompetitiveinconsumerelectronics,
small and large computers (includingsupercomputers),and complex semiconductors,
America now faces serious and widespread deficiencies in its manufacturing ability. A
concerted industrial-government effort, including a thoughtful industrial policy that
looks beyond the strong de facto policies of significant military funding, might galva-
nize action and help reinvigorate the continually eroding U.S. industrial base.

Corporate Research

In the past, invention has been characterized as proceeding through a well-defined set
of stages from basic to applied research, to advanced development, and on to product
development, and this is still the case in certain industries. Given the nature of basic
 Technology 101

Table 5-1. Inventions Produced by Government-Funded Research.

Date Funder Where What
1546 Army University of ENIAC, the first large-scale,
Pennsylvania electronic calculator and stored-
program computer

1952 ONR” MIT Whirlwind computer with core

memory, first compiler, CAM
(computer-aided manufacturing),
interactive computing,
demo. Air defense, air traffic
control

1563 DARPA University of Project MAC and Multics, time-

California,Berkeley; sharing, stimulated AT&T’s UNIX
Michigan; MIT
1965- DARPA CMU; MIT; Stanford Artificial intelligence leading to
robots, and expert and speech
systems

1966 DARPA University of Utah Graphics and the training of the key
graphics scientists and engineers
byDave Evans and Ivan
Sutherland, formation of Evans and
Sutherland

1967- DARPA SRI Human-interface experiments

leading to Xerox PARC’s Alto and
on to Apple’s Lisa and Macintosh
1972- DARPA BBN; University of ARPAnet as first packet switch;
California, stimulation of Ethernet and packet
Los Angeles; radio, the forerunner of cellular
Network phones
Analysis Corp.; etc.

1580- DARPA University of UNIX 4.x to exploit and evolve

California, Berkeley UNIX

1981- DARPA University of Computer-aided design for VLSI

California, Berkeley; (very-large-scale integration),
Cal Tech; CMU, including silicon compilers
MIT; Stanford; etc.
1985- DARPA University of Strategic Computing Initiative.
California, Berkeley; Connection Machine, RISC
CMU; MIT; Stanford architecture, Systolic Processing

1985 NSF, DOE Cal Tech Hypercubes as the first large

multicomputers
aTheOffice of Naval Research, a precursor to DARPA, funded early basic research.
102 Technology and Engineering

research, products are often an unplanned side effect of industrialaccidents.In the case
of computing,though, traditional industrialresearch plays a less significantrole than it
does in other industries, such as chemicals. Furthermore, since Nobel Prizes are not
awarded for computer-science inventions and discoveries, there is no established
method for recognizing research accomplishmentsin this field. However, there have
been some noteworthy achievements at industrial labs, including Bell Lab's transistor
and UNIX, IBMs work on RISC, and the first development of a distributed workstation
environment at Xerox PARC.
Major electronicscompanies spend an impressive amount of money on research
and development. In 1989, Electronic Business reported on the top research and devel-
opment spendersin the United Statesand Japan (showninTable 5-2),which spent $13.6
billion and $15.2 billion, respectively.
Occasionally, corporate product development establishes new directions in com-
puting.Someof themoreimportantfirst developmentsincludedisks,printing,relational
databases, and the System/360 architecture by IBM; minicomputers, time-sharing,
networking, and the VAX architectureand its homogeneous computing environment
by DigitaI Equipment Corporation (DEC);and the Intel 80x86 architecture combined
with Microsoft's operating system as a basis for the evolution of the personal computer
or, alternatively,the IBM-compatiblePC.
Not all "corporate research takes place in laboratories such as those just
mentioned, however. Many technological advances attributable to corporations
actually originatewith the user base at those companies.The applicationof computers
has been a prime source of new product ideas. Products often progress from a
specific program at a firm such as Lucasfilm Ltd. to the formation of a company
(Pixar)to exploit the product on a wide-scalebasis. The largest softwareorganization
(ComputerAssociates)came from the process of productizing programs encountered
by an organization serving IBM customers.
Since computers either supplement or supplant other information-processing
systems, includinghumans, the potentialfor computing is very large (i.e.,as large as the
information business itsem. Computers will eventually be involved in the creation,
storage, or transmission of nearly every bit in the universe. Applications designed to
exploit that potential are therefore a major source of ideas for new companies.

Individual Research

Although virtually all the research and development that has resulted in significant
products can be traced back to an individual or a project leader, some of the inventions
are particularly noteworthy because they led to a new kind of computer or a new way
of computing. Most of these advances involved new programming languages or new
 Technology 103

Table 5-2. Top U.S. and JapaneseResearch and Development Spenders.

American Expenditures Japanese Expenditures
Company Amount ($ billions) Company Amount ($ billions)
IBM 5.9 NEC 3.7

Digital 1.3 Hitachi 2.8

Hewlett-Packard(HP) 1.0 Matsushita 2.4

Xerox 0.8 Fujitsu 1.9

Unisys 0.7 Toshiba 1.7

Motorola 0.7 Sony 1.1

Hughes 0.6 Sharp 0.5
Texas Instruments (TI) 0.5 Canon 0.5
NCR 0.4 Ricoh 0.4

Control Data
Corporation (CDC) 0.3 Omron 0.2
Honeywell 0.3

Intel 0.3

hTational 0.3

Apple 0.3
Wang 0.2

uses for computers.Other advances, such as new architecturesor large computers,were

team efforts. Table 5-3 lists some outstanding personal contributions to computing.

New Components

In addition to the research organizations mentioned above, the introduction of new

components is another important source of inspiration for the formation of start-ups.
When a new component(e.g.,Motorola 68000)becomes available,many companieswill
simultaneouslyform, all of which claim to be unique and ideally positioned to exploit
the novelty. For example, when the availability of the Intel 80860 component was
104 Technology and Engineering

Table 5-3. Outstanding Personal Contributions to Computing.

-~ - _ _ ~ -
Who Wheve What
_~__________ -
Backus IBM Fortran

Bricklin and Franksten - Visicalc, the first spreadsheet

Cocke IBM RISC

Cray CDC/Cray RISC, vector architecture,

innovative circuitry, and
high-density packaging

Iverson IBM APL-A Programming

Language
Kemeny and Kurtz Dartmouth BASIC

McCarthy MIT Lisp

Thompson and Ritchie Bell Labs UNIX and the C language

announced in April 1989, I was invited to become the CEO of a company whose stated
purpose was:

developing and marketing a new category of computer system to address a substantial,

emerging market opportunity: to bring supercomputing solutions into the mass market
of the desktop era . . . with the first line of extremely high performance servers.

Unfortunately, more than a dozen other groups (including a few start-ups) had also
formed to build roughly the same product and attempt to enter the same market.

New Architectures

Most recently, the idea of parallel processing has given rise to almost a hundred
hardware start-ups aimed at supplying high-performance computing by linking
hardware componentsof various lunds.These new ventures have produced significant
advances in the ability to operate a large number of processors, processing elements, or
computerstogetheron a singleprogram. But sinceparallelism is not a market per se, and
the difficulty of solving parallelism is so great (includingthe issue of retraining users),
any company that markets parallelism without solving real problems whose solution
will produce a significantpayoff is certain to fail.Furthermore,given the long lead time
required to establish a market, large firms can adopt the concept on an evolutionary
 Technology 105

basis, after the hard work of developing the technology and markets is done. Despite
these drawbacks, all forms of parallel computingwill exist by 2001, along with enough
users to exploit the various structures.

N e w Standards

In the case of almost every new standard, new companies form to exploit the time
advantage that comes frombeingfirst to market witha product that meets the standard.
In effect, these start-ups are betting against the long or infinite product-gestation time
of large engineering organizations in established firms.
Standards take many forms:

Usergroupgropes (alargegroupof people withvaryingqualificationsgropingfor

a product design) surrounding the standards committees of ANSI (American
National Standards Institute), IS0 (International Standardization Organization),
and the IEEE (Institute of Electrical and Electronics Engineers) that produced
languagessuch as ADA, Algol, and Cobol;interfaces,such as dialectsof W;and
various communicationprotocols, including OSI.

e "Industry-compatible," a euphemism for "IBM-compatible."

De facto, based on a particular part or convention of a dominant supplier. In this

case, users and competitive suppliers agree to employ a particular interface,
such as the Small Computer Storage Interface (SCSI). Start-ups are likely to
create de facto standards.

Consortiaof users and suppliersthat posit a standard, which is then processed and
formally accepted by national and international standards bodies. Ethernet (IEEE
802.3)was developed in this fashion by Digital, Intel, and Xerox.

Establishment of a product acquisitionand development group that specifiesnew

product standards. The Open Software Foundation (OSF)was formed to evolve
UNIX independently of AT&T and Sun Microsystems.

THE ROLE OF START-UPS I N CREATING TECHNOLOGY PROGRESS

The previous subsectiondiscussed the role that technologicaladvances have played in

the creationof start-ups.Thisprocess hasbeen a two-way street,though, sincestart-ups
have also played a major role in the creation of technologicaladvances. All new classes
ofcomputers,from supercomputersto personal computers (but excludingmainframes),
originated withstart-upcompaniesthat used new semiconductortechnology.Start-ups
106 Technology and Engineering

have been pivotal to computer development, as proved by a number of substantial

innovations,including Microsoft's evolution of MS/DOS as a basis for IBMs Personal
Computer evolution;the development of relationaldatabasesby Oracle and other start-
ups; and the introduction of floppy and small disks using IBM-developed Winchester
recordingtechnology.Start-upshave created genericword processing,communications,
mail, and spreadsheet programs as well as profession-basedapplicationprograms such
as mechanical and electricalcomputer-aided design.

TRANSFERRING TECHNOLOGY FROM ITS SOURCES

Successfully securing technology from the sources mentioned above requires an

understandingof the laws that govern the flow of technology.These laws deal with the
substance of the technology, the transfer process itself, and the inevitable competition
to develop products for selected applications.If a new venture is to be viable, it must
address all three of these issues.
The followingare some common technology-transfermethods, listed in increasing
order of effectiveness:

Papers and conferencesat which ideas and algorithms are presented

Industrial programs for exchanging ideas, people, etc.

Industrial research consortia

Direct funding of projects whose mission is to produce a specific result

Consulting with a high degree of responsibility

Computer programs that can become industrial-strength

Transferring trained people, together with a new technology

Transferring trained people, together with a prototype or an operational product

The best way to transfer technology is to transfer trained people. This method of
technology transfer is especiallyeffectiveif the people can bring a prototype of the new
idea with them. Although this is generally not possible if they are coming from a
government or commercial laboratory, it may be possible if they are coming from an
academic environment.The concretenessof the idea is essential. In the words of MIT's
David D. Clark "One artifact is worth a thousand papers." The prototype not only
demonstrates feasibility,it also demonstratesa potential new product in an application
context.If the technologyembodyingthe idea cannotbe demonstratedat the outset,any
 Engineering 107

company founded to exploit the idea is likely to find itself doing a lot of fundamental
research.
Novel conceptsor artifactsinspirenew product ideasin the minds of every engineer
or savvy marketeer, and do so in direct proportion to the media attention devoted to
these new developments.Furthermore,at least two redundant east/west firms will be
created in responseto these conceptsor artifacts.The total number of companiesformed
will be roughly proportional to the amount of investment capital available from all
sources.In short, entrepreneurswho think their proposed start-up will be alone in the
field are probably mistaken.

ENGINEERING

The two types of engineering done in nearly all companies involve hardware and
software. Hardware engineeringis the process of designing and building an ultimate
physical object,such as a computercomponent,a computer,or a manufacturingprocess
or plant to produce products. The fmt stage of hardware engineering is to build a
working simulation or model of the product or plant using a computer before any
physical construction is undertaken. The final stage is to build and test the physical
object itself.
Softwareengineeringis the process of building a program or product that operates
wholly within the confines of a computer or computer network. Since all engineering
requiresan understandingof software,the engineeringprocess describedin this section
will focus mainly on the formal steps of softwareengineering.

HARDWARE ENGINEERING

Hardware engineeringis the process of utihzing technology to create a new product or,
more precisely, a set of documents and specifications from which a manufacturing
organization can build the product. A good hardware engineer has vision tempered
with judgment, the capacity to deal with endless detail, and the fortitude to stay the
course despite setbacks.
During a company’searly stages, the foundersand hardware engineersmust select
the right technology to employ in the product. This is a criticaljudgment call, since the
firm’s success depends on having the “right tech.” Technology cannot be too high
(approachinginfinity,as in research)or too low (approachingzero, such that anyone can
replicate it). Furthermore, it has to be timed right. No other factor in a company’s
development operates in such a delicate balance. On the one hand, investors want
demonstrably unique and proprietary technology, and on the other hand, they are
unwilling to invest in any risky research to develop such technology. Thus, really
108 Technology and Engineering

successfulstart-ups are llkely to come directly from the laboratory,where the technol-
ogy is demonstrable, as in the case of Silicon Graphics and Sun Microsystems,whose
founders developed prototypes while at Stanford. The best technology/ product, as
noted earlier, consists of a prototype together with the people who created it.
As a company develops additional products, the hardware engineers must make
this technology decision again and again, knowing that their opposite numbers at
competingfirms are malung similar decisions.Typically, a new technology or new part
is ”almost available.” If the engineers choose to use it and it is delayed in reaching
production, their company will have no new product. If they choose to be too
conservative,the competition may succeed in offeringcustomersboth the new technol-
ogy and the ”bigger bang for the buck that it represents, leaving the conservative
engineers with an uncompetitive product.
In general, the hardware engineers at small companies will be more daring than
those at large companies, since they need smaller volumes of the new parts and they
must take market share away from established firms by offering more value via high
technology. Furthermore, the fact that product cycles are longer at larger companies
tends to increase the lead time between the ”technology decision point” and product
availability. Because of the longer lead time, engineers must forecast technology
availability further into the future, increasing the risk and generally resulting in more
conservativetechnology decisions.
In addition to making technology decisions, which many regard as the “fun part”
of engineering, hardware engineers must evaluate every aspect of their design for
operationat maximum and minimum clock speeds,temperature, componentvariation,
etc. Although many simulation tools now exist to aid in this task, it still involves
considerable drudgery.
Finally, in hardware engineering,Ohm’s law and Maxwell’s equations pale in im-
portance and influence next to Murphy’s law. Even with simulation tools, a number of
“gotchas” lurk in every design effort. Some are technical; some are personal; some are
political.The good engineer will stay the coursethroughoutthese setbacks,which is one
reason why high-tech ventures like to hire engineersfrom top universities, where com-
petition is fierce and courses are difficult-the graduates have been “fire-hardened.”

SOFTWARE ENGINEERING

In the 1990s, a computer system organization, including one that builds PCs, must
understand software.The companyis usually responsible for unique programs that are
part of the machine (e.g.,firmware),plus the softwareused in the design,manufacturing,
and testing of the systems. Furthermore, a great deal of what was formerly hardware
design has now become software-oriented.For example, much digital system design is
now done in a completely symbolic fashion whereby a ”program” is compiled into a
 Engineering 109

chip. Thus, every high-tech company must be intimately involved with the develop-
ment, use, and maintenance of software whether the firm likes it or not.
Some people regard the development of software products as an unstructurable,
unschedulable, unmanageable, and highly creative process that is the province of the
last great American inventors. Others see the process as resembling a factory, with
thousands of programmers worlung in one large, open room, all using the same,
impersonalbureaucraticprocess,and turning out thousandsof lines of code, as in Japan.
The latter model generates almost twice as much code per person as the former model,
and at a qualitylevel that is two to three times better.In order to produce at the necessary
volume and quality level, start-ups must use methods that are closer to those of the
Japanese than to those of the lone inventor.
Since everything about the design and fabrication of a product can be considered
in factorylike, or at least job-shop, terms, the idea of developing software on a mass-
production basis is indeed tempting. This approach, however, is beyond the means of
most new ventures, which must follow the ”invention”model rather than the ”factory”
model in developingtheir software.Fortunatelyfor thosestart-upsthat must rely on the
”invention”model, though, users are interested strictlyin the end result (i.e.,the unique
characteristics,functions, performance, and quality of the product itself), regardless of
whether that result was achieved by a creative, inventionlike design process or a
standardized, factorylikedesign process.

PHASES OF A HARDWARE OR SOFTWARE ENGINEERING PROJECT

Every product passes through a series of predictablephases (or stages)from concept to

retirement,and most companieseventuallydevelop a phase-review process to track the
development of their products. Table 5-4 shows these phases of a product’s develop-
ment, together with the corresponding stages of the start-up’s growth.
Aproduct is conceivedduring the company’s conceptstage(I),proceedsalonguntil
the seed stage (II),is developed in the product development stage (IIIa, IIIb, IIIc, and
IIId), and ultimately reaches a phase during the market development and steady-state
stages (IV and V) in which it is produced, sold, and maintained. In its final phase, the
product may be enhanced and improved for some time before it ultimately passes into
a state of retirement at the end of its life.The followingparagraphs exploreeach of these
stages in greater detail.
During the concept stage (I), the idea for a product must be explored and dem-
onstrated to some degree. This demonstration can take the form of a feasibility
prototype or model, or it can consist of either a demo by a key engineer indicating that
such a design is possible or a careful analysis of critical technology.
During the seed stage (IT), the requirements are spelled out in a product-
requirements specification, a system production definition, and a preliminary user’s
 Table 5-4. The Phases of an Engineering Project, Company Stages, and Software-Engineering Process.

Bell-Mason Market Development

Diagnostic Stage Concept Seed Product Development and Steady State

Hiring,
Specifications, Designing
and Schedule and Building Alpha Testing Beta Testing

Preliminary Detailed Implementation System testing Acceptance

specifications specifications
and schedule and schedule
~

Company stage I I1 IlIa 11% IIIC IlId IV and V

Formulate Plan product Assemble Build Alpha-test Beta-test Produce and

technology. and project team. simulation of product via product with sell product by
based on product, internal use real users in maintaining
Gather requirements. Define followed by with internal their product and
market prcduct. first running users. working correcting
requirements. Demonstrate system. environment any errors.
technical Prepare
Outline product. feasibhty. formal plan Maintain product's
life with
Plan technical enhancements.
feasibility

Software
EnginePring Phase I I1 IIla IIIb IIIC IIId IV and V

Concept Planning and Hiring Implementation Integration, User Fixing,

explanation definition of designing, and testing acceptance adapting, and
requirements replanning Unit design, testing enhancing
coding, testing
Preliminary and
detailed design

outputs: Feasibility Product- Architectural First qperahonal Mcdfications Producible and

product, prototype or and detailed functional product to adapt to supportable
manual, and model specification design product from user needs product with
specifications specifications tested and minor release
System integrated User-accepted
production User's manual components product
definition

Preliminary user's
manual
Table 5-4. (continued)

Software
Engineering Phase I I1 IIIa 11% IIIC IIId IV and V

Plans Preliminary Final project Acceptance- Product- Product- Enhancement

project plan plan test plan release plan support plan plan for next
release
Preliminary Final Installation
verification verification and training
plan plan plan

Reviews Product- Preliminary Source-code Product- Product- Project

requirements and critical review release review support review postmortem
review design reviews
Acceptance- Installation
Product- test review and training review
feasibility review

Resources Prototypes and Network Host compute1

first computers

Standards Program standards Configuration

and style guide management

People Architect and Project leader Development

product designer team
112 Technology and Engineering

manual. These are described in the specifications section (page 1181, and it may be
worthwhiletocheck therequirementsagainsttheactualproduct definitionaspartof the
process. A seed stage is especiallyvital to a start-up involvinginnovativesoftware (i.e.,
softwarethat is being developedfor the first time),because this is the stageduring which
the planning for the product is done. Without planning, the software schedule will be
unpredictable.
During the first portion of the product developmentstage (1114,the main part of the
design is carried out, starting with preparation of an architecturaldesign specification,
which is subjected to a preliminary design review.This is followed by preparation of a
detailed design specification,which is subjected to a critical design review. Successful
completion of this latter review is the main exit criterion for t h s design stage. A user’s
manual is written and made available. The verification plan is developed, and the
project plan is updated.
During the implementation phase (IIIb), the design for each of the software
components is prepared, and the coding and testing are completed. Formal design
reviews are the best way to evaluate a complex hardware or software project. They are
also the cheapest way to debug a program or system. Programs can be assessed by
conducting code walk-throughs in which the designer ”walks” a team of four to six
people through the design. Alternatively, an inspection team can be appointed to go
over the code, examining it and comparing it against various criteria. After the design
reviews are complete,the programs are tested individually.Hardware implementation
consists of two phases: first, an implementationof the design;then, a simulationbefore
physical hardware is constructed and assembled. The exit criterion for substage IIIb is
the existence of working subcomponents.
During the alpha-testing phase (IIIc), the subcomponents are integrated into a
single system so that system testing can begin. After the system passes the specified
internal acceptance tests, it can be released for first use by customers.
During the beta-testingphase (IIId),the system is given to customers for actual use.
This phase involves working closely with an appropriate number of customers (e.g.,
three for computer hardware or large softwaresystemsand twenty for mass-produced
components or software) to gain an in-depth understanding of how well the system
meets the expectationsof real users. In this stage, serious errors that result in unreliable
operation must be fixed immediately.Critical features that were overlooked for some
reason may have to be added. Thus, the product-support organization is first tested
during a time when it is building its formal plan for providing support.
Theoretically,the product is announced upon completion of beta testing, when the
company has satisfied itself that customers are happy and the product is viable in the
marketplace.Actually,nearly all start-upsannounce the product during, or even at the
beginningof, beta testing.Anythingless conservativethan announcingat the end of beta
testing is a flaw.
Finally, the product is released, and the firm enters the market development stage.
 Conclusion 113

CONCLUSION

Semiconductor and magnetic densities, as measured by the number of bits stored per
unit area, are llkely to increase at their current exponentialrates on into the twenty-first
century. These increases will provide opportunities for new hardware systems, which,
in turn, will permit the development of new software products (as will be discussed in
Chapter 8).
Technology involves the ability to design and build high-tech products. Many
developmentsprovidean opportunityfor technologicalprogress,includingcomponents
such as semiconductors,standards,customerapplicationneeds, and genuineinventions.
Technological advancements come from having trained resources and concentrating
those resources on discovery. A significant portion of the world's research and devel-
opment capabilitiesare available to entrepreneursand intrapreneursin various forms,
ranging from papers, demonstrations, and consortia to trained people. Technology
transfer is best accomplished by transferring people, as occurs when people leave a
laboratory where they have developed an idea and form a separate group or a new
company to commercializethe idea by creating a product.
Technological know-how is a necessary prerequisite for a new venture, but it is by
no means sufficient. The start-up must also have a mastery of engineering that will
enable it to successfullyconvert its technology into products in a prehctable and timely
fashion.
Chapter 6 presents the technology balance sheet, which can be used to break a
company's technology and engineeringabilities down into twelve separateaspects, or
dimensions, that are analyzed to determine the status of these two critical areas.
Chapter 6

THE TECHNOLOGY
BALANCE SHEET

Just as it is essential to understand an organization’s financial health, it is equally

necessary to understand and measure its technological health. The first section of the
chapter describes the technology balance sheet, a useful approach to measuring a
company’s technology. The second section presents a number of classic technology-
related flaws, ranging from requiring infinite technology (i.e., attemptingto develop a
product that is predicated on a fundamentaldiscovery or technologicalbreakthrough)
to having no sustainingtechnology.The final sectionlists the rules for evaluating a new
venture’s technologicalposition at the end of the concept and seed stages.

THE TECHNOLOGY BALANCE SHEET

The technology balance sheet evaluates each of twelve key dimensions of a start-up’s
technology. Readers may notice that the dimensions used on the technology balance
sheet to assess a firm’stechnology are very similar to the dimensions used throughout
the book to assess a firm’s overall status.
Figure 6-1 lists the twelve dimensions to be considered and measured:

Technology base

Standards

Design, quality, and other processes

114
 The Technology Balance Sheet 115

Plan. with schedule

Design,quality, and Engineeringspecs:

and otha pmesses
\ I *Thedesign as seen by the user (data sheet, e=.)
/ * How the product is architecled for design
External (indusoy)
and internal standards Manufacturingspecs
(i.e., how to produce the
product)

Indigenous
(i.e.. skills) ana Chief technicalofficer
exogenous (vicc president of engineering)
technologybase

Operauonal Team and

management
he.. the abllity
io meet plans) 1
Technology
fUtUn
I
Technicalresources
' enginecringculture

Archilect(s)and archileclural
definition process
(people, consultants,
computersand software.
cools. and l a h w
equipment)

Figure 6-1.Technology Balance Sheet Plotted as a Relational Graph.

Plan, with schedule and resources

Engineering specifications

Manufacturing specifications

Chief technical officer

Team and engineering culture

Architecture

Technical resources

Technology future

Operational management

These dimensions will be discussed in the following subsections.

TECHNOLOGY

The technology dimension includes internal and external sources of components,plus

"know-how," as representedby critical personnel, patents, processes, etc. The company
must examineevery facetof the technologythat it needs to build a product and then rank
each technology source as objectively as possible.
116 The Technology Balance Sheet

STANDARDS

Standards should be regarded as a critical aspect of product design. Establishing

uniform ways of doing things (such as having an exact dialect of a language for
expressing a program and having programming style conventions) permits rapid
progress because standardized components can be interconnected and built on one
another.Although standards are inherentlyconstraining,Dave Nelson (oneof Apollo’s
founders) believes that constraints are what really breed creativity. In designing a
product, it is inherently more difficult to start with a blank slate than to start with some
restrictions, because in the absence of established criteria, nearly anything is possible.
Effort will therefore be squandered exploring an almost infinite number of options
rather than channeled and focused in the most productive directions.
A start-up (or a company of any size, for that matter) must understand and
implement both external and internal standards. Major aspects of product design are
determinedby external(industry)standards coveringsuch areasas inputs,outputs,cost
(in memory size),and speed (fiftythousand lines per second).For example, a compiler
may be specified as having to accept ANSI (American National Standards Institute)
standard C language input, produce code for the Motorola 88000 chipset that is better
than the existingcompilers,occupy no more than 100lulobytesof memory,and compile
at over fifty thousand lines of code per second.
Internal standards are equally important and range from how logic design or
programming is done to line width on printed circuit boards. Internal standards must
be specified, published, and enforced in a formal manner. For instance, when Digital
Equipment Corporation (DEC)first started, the Engineering Committeetook responsi-
bility for creatinga set of design standards that covered everythingfrom how a physical
environmentwould be specifiedand tested (power,temperature,humidity, etc.)to how
a copyright statement would be placed in memories and programs. Internal standards
also include a list of the components that are permissible in new designs.
Upon seeing such standards and component lists, the first reaction of most
engineersis that they are bureaucraticand constrain creativity.However,standards are
simply a statement of decisions that have been made regarding good practice, which
means the designerdoesn’t have to think about these more mundane aspects of a design
(such as the temperature at which the product should be designed to operate) and is
therefore free to concentrate on the truly creative aspects.

DESIGN PROCESS

The design process, which specifies what tools engineers use to create and check each
part of their product design, must be documented and managed. The design process is
intimatelytied to the resources a company has to aid designers.Much has been written
 The Technology Balance Sheet 117

about software engineering, and there are any number of valid models for how
programming should be done. The important thing is simply to pick a model that is
appropriate for the team and operate according to it.
The SoftwareEngineeringInstitute (Humphrey,1989;derived from Demming and
Juran) has established a five-level ranking to characterize how effectively a team is
functioning in terms of its process capability:

1. Initial: There is an ad hoc process. Formal procedures may exist, but there
is no management mechanism for traclung results against the procedures.
The team rarely makes and meets plans.

2. Repeatable: A process exists that deals effectively with routine programs but
produces unpredictable results with new programs or new tools.

3. Defined: A qualitative description of the process exists.

4. Managed: The process contains a minimum set of measurements to define

quality and cost; a process baseline exists; etc.

5. Optimizing: There exist sufficient quantitative measures for each part of the
process to allow the process to be completely understood and fine-tuned.

Humphrey describesa method for evaluatinga company’s process capabilitiesand

also recommends various processes, standards, and methods for attaining software
process control.The organizationwith which he is associated,the SoftwareEngineering
Institute, can audit a firm to determine its level of process control, and some members
of the institute did so as part of a 1990 trip to Japan sponsored by the Ministry of
International Trade and Industry. While there, they found that many large Japanese
companies are operating at the highest level in the above ranking (level5, optimizing),
enabling them to achieve quality and productivity levels more than twice that of their
U.S. counterparts.

ENGINEERING PLAN

The engineering plan includes the schedule and a list of the resources required. The
resources list must cover both the resources for developing the product itself and those
for developing any of the manufacturing and design processes that the product
requires.The importantthing about an engineeringplan is that the schedulebe realistic.
Developinga truly realisticscheduleis almost impossibleif the product has never before
been attempted; it is merely very difficult if the product has been attempted previously
but the team has never before worked together. In the latter case, each team member’s
118 The Technology Balance Sheet

ability to establish a realistic schedule for his or her portion of the work will probably
be untested. In terms of the process-capability levels outlined by Humphrey, it is
unlikely that such a softwareteam in a new venture could get above level 2 (repeatable)
by the time it ships its first product.
Product-gestationtime gets ingrained in the people and the company. Their ideas
regarding product-gestation time are often based directly on the lead times at a larger
firm, which are guaranteed to be much, much longer. One of the most important aspects
of an engineeringculture is to establish an accurate but responsive ability to schedule.
There are four ways to schedule a project:

Optimistically: Put enormous pressure on the team by preparing an aggressive

schedule that the team believes can only be met if everything goes right.

Pessimistically: Build so many delays and contingencies into the schedule that the
schedule will certainly be met (an approach unlikely to be used by a start-up).

Realistically: Allow for an appropriatenumber of contingencies,whch will become

possible when the team is mature enough and understands the project and each
other well enough. However, it often happens that everyone up the chain of
command then adds a contingency, and the net result is a bloated, pessimistic
schedule (again, not typical of start-ups).With realistic scheduling, the company
may end up with two schedules-the optimistic one and the one with the contin-
gencies added.

Running blind: Work on the project until it gets done. The firm that uses this
approachhadbetter startwithlotsof money,beabletoraisemoremoneyeasily,and
have plenty of extra resources.

In the final analysis, schedules really don’t always work. Any critical schedule
milestone must coincide with an immovable deadline such as a demonstration to the
board, a trade show, a funding event, or a customer shipment. If customer shipment
serves as a deadline, quality must always be used to control shipment.

ENGINEERING AND MANUFACTURING SPECIFICATIONS

The engineering and manufacturing specifications describe the product in several

ways. First, they describe its external specification, or the product’s function,
including performance, as seen by a user. Second, they describe its internal specifi-
cation, or the product’s structure and internal function as seen by the engineering
team (i.e., a set of components to be designed). Finally, when the product has been
fully specified both externally and internally, manufacturing requires process and
product specifications describing how the product will actually be built and tested.
 The Technology Balance Sheet 119

CHIEF TECHNICAL OFFICER

The chief technical officer (CTO), or engineeringvice president, is the technical leader
in charge of implementing the product. This person is ultimately responsible for all
products and is the CEO for the engineering organization. Thus, his or her general
qualifications must parallel those of the CEO because the CTO is the "clock and
"standards setter" for engineering.
The company should have selected its CTO by the end of the seed stage, and if it is
tackling a technologically difficult product, the CTO must be on board from the start.
Funding a high-tech venture without a CTO is extremely risky because he or she is the
individual responsible for ensuring that the product is really feasible at the price,
quality, schedule, and resource level specified in the business plan.

ENGINEERING TEAM AND CULTURE

The engineeringteam and cultureare just beginningto formby the end of the seed stage,
since at this point, a complete team has yet to be hired and the head of engineeringmay
not even be on board. The organizationalstructureis quite important because the CTO
may have positioned himself or herself as a bottleneck by assuming responsibility for
all intergroupproblem resolution. As with any organization,theory X, Y, and Z will all
work. I do not favor highly top-down engineering organizationsbecause they do not
bring out the creativity of the people doing the work. Furthermore,top-down structures
eliminate critical intraorganization communication. Worst of all, top-down organiza-
tions usually do not engender commitment to schedule,resources, and product on the
part of the responsible engineers.

ARCHITECTURE

The termarchitecturewas coined in.1964by the IBM System/360design team to describe

a computer's instruction set, or how the computer appeared to a program (or pro-
grammer).Architecture is now used in a broader sense that encompassesboth "external
archtecture" and "internal architecture." The external architecture describes the gen-
eral function of any computer component (i.e., what it does)--such as the instruction-
set architecture, operating system, compiler, a network protocol, or spreadsheet-and
how thiscomponentappearsto anyoneusing it.Theinternalarchitecture(or"realization")
forms the blueprint for how the components that create the external architecture are
implemented;it is what a development team designs and builds.
It is therefore vital to have a product architect who can both define the product
externally and be able to play a major role in decomposing it for realization and then
engineeringit. His or her responsibility for product architectureapplies equally to all
120 The Technology Balance Sheet

levels of hardware and software. Thus, the product architect is likely to be the most
critical person in the engineering group,’ including the CTO.
The architect’skey job is to guide the product‘s implementationand evolutionover
its lifetime. The lifetime of a good architecturewill be considerable, and the company
fortunate enough to have chosen a good architect and architecture will profit immea-
surably.Much of DEC‘s successduring its first three decades (1957-1987)was based on
constant and evolving architectures for its minis, including the VAX. System/360
hardware and software systems and their successors were the basis of most of IBMs
revenues and profits for a similar period. More recently, the Apple I1 and Macintosh
architectures have each prospered for over a decade. In 1990, Sun Microsystems has
been attemptingto repeat the successof these predecessor architecturesby establishing
the SPARC architecture as the standard for workstation-class computers.
Although most of the examples cited above involve hardware engineering, the
same sort of architecturalintegrity must also be maintained for software.At Microsoft,
everyproduct, such as Word or Excel, has a singlearchitectwho maintainsthe product’s
integrity(andis usually its chef implementeras well).When responsibilityfor a product
is diffused, as in the case of Fortran or UNJX, by placing it in the hands of some
amorphous, committeelike group that is pushed around by numerous standards
organizations,the product’s efficacy declines and its ability to evolve may be stymied.
In my view, lack of a good architect,or lack of a suitablearchitecture,is the fatal flaw
in many hgh-tech ventures. Although at first, the product architect may be the CTO or
even the CEO, ultimately, someone within the engineering organizationmust assume
responsibility for maintaining the efficacy of the product’s external specification,
especially with respect to how that product is changed as it is implemented in
succeeding product generations. In some cases, as in the example of Ardent’s Titan
workstation described later in the chapter, several architects may be required as a
product is broken into various parts.
Not having an architect is quite risky, because it leaves the product’s definition to
some nebulous process or to a group that gropes with the product design, as I recently
saw in a company building a multimedia system. Not having a way to manage the
product’sdesign and delegateit to those who must do the design is almost always fatal.
One of the biggest dangers is overcommitment.When a technically difficult project
begins,and one person functionsas CEO,CTO, and architect,the company, engineering
group, and product are all likely to be out of control unless the firm has a sufficiently
strong staff, including a chief operating officer. In a start-up, such a project must have
a full-timearchitect who will also play a major role in the product’s design.

1. My own background and biases as an architect may account for this belief.
 The Technology Balance Sheet 121

In a recent case, the architecturalconcept for a product was superb,but the architect
had four problems that thwarted effective implementation of the architecture.These
problems, which are typical of many architects, were as follows:

The architect lacked an understanding of the specific benchmarks by wluch the

product would be judged in the marketplace.

He had trouble finishing detailed specifications for the product, leaving the
engineering organization with only a fuzzy idea of what it was supposed to be
designing.

He was not knowledgeable about implementation, which meant that the

architecture could not be implemented within a reasonable time and at a
reasonable product cost.

His poor interpersonal skills made it hard to keep the actual implementation in
synchronization with an ambiguous specification. (This was both his problem
and that of the CTO.)

Although not having any architect can be a problem, having too many architects
can also be a problem, as illustrated by the following example.

Venus. On Friday, Augustl3,1982, I went to a design review for Venus (VAX8600).

The review, attended by several hundred people, focused on the schedule and the
risk involved in not getting the chip layouts to the gate array supplier. I asked
whether the design had been simulated or thoroughly reviewed. It hadn’t, since the
group was in such a hurry to meet the schedule that they wanted to skip the checking
stage. On Saturday, I visited the project team to talk with its members and found that
the management didn’t understand the project and that four individuals each
regarded himself as the project’s sole architect and wanted the credit. The project
had about four design styles, because it consisted of four large subsystems.
By Thursday, no one wanted credit. Within six months, the project was brought
under control through many management changes and the introduction of a design
process that required the use of design reviews and simulationto ensure the correctness
of the design prior to building the hardware.The product ultimately shpped two years
later than scheduled,whereas, left on its original course, it would probably never have
shpped. The project ended up with an organizationalstructureconsistingof four archi-
tects and a lead architect to resolve the conflicts among the group and finish the design.
Despite all these dire warnings about architecture,it is possible for a talented team
of architects to work well together and produce an excellent product. The following
122 The Technology Balance Sheet

story about the architectureand development team for the Ardent workstation (Titan)
illustratest h s point and showshow an architecturaltask can successfullybe broken up.

The Ardent Titan architecture and development team. Given the complexity of a
graphics supercomputer, Ardent had to break the definition and responsibility for
various elements of its design into independent parts and architectures. The entire
project worked well when all the roles were defined. The architecturewas the basis for
three products, includingTitan, which could be evolved through three generationsover
a five-year period.
The company’s chief hardware architect, GSM, was responsible both for defining
the instruction-set architecture externally and for defining the internal architecture
(howthe parts of the entirecomputer fitted togetherusing a corebus).GSM alsodefined
the processor’s internal supercomputerarchitectureand took on many of the difficult
processor-design tasks, although he did not lead the hardware project nor was he
directly responsible for implementing the processor. After the first version of the
machine was introduced, GSM led benchmarlung and observed the machine in real
applications.This was critical for the design of the next implementation.
JRA was the architect, leader, and key implementer for the development of the
parallelizing vectorizing compiler. Having a single individualbe responsible both for
the architecture and for leading the implementation thereof was an ideal case. The
languages archtecture and debugger were the responsibility of SCJ.
WT was the architect, leader, and key implementer for the development of UNIX,
which supported the Titan hardware and provided a program environment for the
compiler and other applications programs.
TD was the architect,leader,and key implementerfor the graphicshardware, while
WW designed and built the softwarepipeline to transform and display3-D objects.MK
was the architect and chief implementer for the graphics library.
All architects/implementers had to cooperate on determining each architectural
interface and on the entire design.

TECHNICAL RESOURCES

The next dimensionon the technologybalance sheetis technicalresources.Thisessential

category includes people, equipment (bothcomputers and networks),and softwareto
run the engineering enterprise (i.e., operating systems, languages, computer-aided
design [CADI programs, and softwarelicenses).Of all the resources, the technical staff
is the most important.
The company’shiring abditydeterminesthe qualityof the staff.All firms,regardless
of how well they may be managed, find that hiring grade-A people takes much longer
than anyone had planned. The key to hiring is having the right sources. The most
effective approach is to develop a network of contacts with the best people working in
 The Technology Balance Sheet 123

each area such that recruiting is by word of mouth. A technical advisory board can be
one of the company’s greatest hiring assets.
The organization’s first hires have to be great because really good engineers like to
be involved with other good (or even better) engineersand are intolerant of bozos and
turkeys. Great people hire even better people. Poor people hire even poorer people.
(This is the pygmy theory of hiring.)Furthermore,because the company can expect to
acquire its share of average people merely in the course of making minor hiring errors,
it must never deliberatelyhire average people just to fill slots.
The following storiesillustratesome interestingapproaches to hiring used by start-
ups, along with critical observations on hiring the engineering team.

Ardent and Stellar. Both Ardent and Stellar allowed three months to form their team,
but it took six months before the companieswere fully staffed. Each firm established a
technicaladvisory board to aid in the product definition, and these boards were the key
to finding and recruiting the best people.

WAVETRACER. Richard Fiorintino,CEO of WAVETRACER, recruited engineersby

sending a personal mailing to surrounding towns, at a cost of less than $500. Recruiting
firms (headhunters)are a last resort, because they will be costly and error-prone. The
start-up’s leadership team itself is clearly responsible for staffing, no matter what the
formal hiring channel may be.

Objectivity. Objectivity, a start-up building an object-oriented database, had a

relatively long seed stage, during which it both designed its product and hired its key
engineering leaders. When the product development stage started, Objectivity was
ready to do full-scale recruitingbecause it knew what it was going to build and how it
was going to build it.
It first created a database of everyone working in the field who had experience in
theory, use, and development. It grouped the list according to technology and gave
priority to people who had built specialized databases. Using all available sources,
Objectivity did a forced ranking of everyoneon the list, whether they were available or
not. The process was carried out with peers via phone calls and in direct interviews,
which were scheduled two nights per week and on weekends.
But the hiring process did not end when the candidates made a commitment and
joined the organization. Instead, the process continued for a six-month probationary
period,during whicheachnewhire worked with the teamand was givenanopportunity
to discuss his or her design and engineering philosophy at length. Several candidates
ultimately left during the probationary period.
Objectivity’s scheme had many advantages: the company found a large pool of
engineers, got the best people to head the team, formed the team itself, gained an in-
depth understanding of its potential competitors, and established links with potential
124 The Technology Balance Sheet

buyers. Objectivity‘sapproachdid have one flaw, however:by using the grapevine,the

organization risked disclosingbasic technology and prompting other firms to start up.

Visix Software. Visix built a high-quality, high-performance platform for building

graphical-interface, networked applications. Its desktop for UNIX, Looking Glass,
extends that of the Apple Macintosh to handle networkmg. Visix achieved product
quality by implementing a rigorous hiring process, by managing to keep a small team
together over a five-year implementation period, and by having a single product
architect. A key step in the hiring process was to review the code of each potential
software engineer. Any engineer reluctant to show his or her work to fellow engineers
is a likely loser.

GO Corp. According to Robert Carr, who heads software development at GO Corp.

(Carr,1989),”Allgood softwarethese days gets done through teamwork.”He suggests
the following approach

1. Define the development style. Choices include the collegial model (for
people at a more senior level) versus leadership by a chief programmer
(where the team has less experience).

2. Hire the best first. Others will be turned off if turkeys are present. Worseyet, the
turkeys will want to hire pygmies.

3. Focus on interpersonal skills.Teamwork is number one. Meet with eight to ten

other staff members. Don’t push lukewarm people. Listen to what your troops
are saying.

4. Don’t be afraid to rob the cradle. A twenty-two-year-old may have ten years of
experience.

5. Hire people who have shipped products and been through the cycle, including
support and feedback.

6. Don’t skimp on salaries. Staffmembers should receive stock equal to half their
salaries.It is better to hire the best people and pay them well than to h e a greater
number of poor people and pay them poorly.

Micrografx. Paul Grayson, chairman of Micrografx, recommends that a company

create a development environment that fosters excellence and specifies three types of
reward that can help create such an environment (Grayson, 1989):

1. Cash: A royalty can be paid based on 2 percent of sales, with the lead
programmer getting 1percent.Bonuses can be awarded for project completion.
 The Technology Balance Sheet 125

2. Recognition: An outstandingteam member can be given celebrity status withm

the company.

3. Personal growth: Although team members should have to prove themselves by

working on a low-visibilityproject during their first six months, top achievers
can then be rewarded with the opportunity to ”do something new.”

TECHNOLOGY FUTURE

The technology future dimension measures the new venture’s ability to sustain the
competitive viability of its technology. This dimension includes such factors as an
assessment of the firm’sproductsand architecturesrelative to the state of the art, morale,
process technologies under development, and ability to hire critical people. Like
financeability, the technology future dimension represents an overall look at the
company’s ability to build competitive products in the future.
For example, assume Company N introduces a Motorola 68000 workstationbased
ona CISC (complexinstructionset computer)microprocessor,perhapswithan attached
signal processor, while all the other workstation firms are introducing products based
onRISC (reducedinstruction set computer)microprocessors(suchasSunMicrosystem’s
SPARC, MIPS, Motorola 88000, or Intel 80860). Because the RISC microprocessors
deliverhgher performance,CompanyN s product specificationssufferby comparison,
at least superficially.Company N s ability to respond to the ensuing performance race
by increasing the workstation‘s functionality with voice and video, for example, and
providing a wide range of applications software in the 1990s will determine its
technology future.

OPERATIONAL MANAGEMENT

Operational management is the engineering organization’s ability to manage itself by

meeting its product specification,budget, and schedule commitments. Management
includes all the techniques of managing design reviews, management by objectives,
staff meetings, team building, conflict resolution, etc. Andy Grove, CEO of Intel, has
produced some of the best handbooks on this subject (Grove, 1983,1987).
As a product reaches the final stages of completion, it will become clear that the
team must compromise among the following three indigenous variables:

The schedule, or when the product will be ready

The complete set of resources that is applied toward meeting the schedule,
including computers, consultants, other software, etc.

The characteristics of the product itself, including performance, product cost,

features, etc.
126 The Technology Balance Sheet

The best approach is for the company to pick two out of three, manage those, and
be happy with the outcome. For a start-up, the schedule and resources are really fixed
because of the incredible cost of raising additional funds. Furthermore, it is generally
inadvisableto attemptto add criticaldesign resources to a project that is already running
late, because the firm is then apt to become subject to Brook‘s law: ”Adding resources
to a late project makes it later.”
Therefore, the function of the company’s first product will inevitably be less than
perfect. Faced with the need to cut function in order to meet schedule and resource
constraints, it is best to sacrifice some the product’s features rather than sacrifice
performance. Performance equates to quality in many systems and should not be
sacnficed.Likewise,reliabdity is not a ”feature”;it is a quality constraintthat must never
be sacrificed.

Ardent. At Ardent, Tom Bentley, a former Hewlett-Packard engineer who headed

mechanicaldesign,said it was hard to find contractors who would meet the company’s
standards. ‘We expected a designer to meet both schedule and contract cost [goals],
while also meeting the product cost, quality,and featuresconstraints.SteveJobsexpects
two [of these], and most companies in the valley are happy with just one.”

TECHNOLOGY BALANCE SHEET FOR ARDENT

While working at Ardent, I used a technology balance sheet to analyze the company’s
technology capabilities. Table 6-1 shows the dimensions (and subcategories thereof)
that were analyzed.

TECHNOLOGY AND ENGINEERING FLAWS

Some of the technology and engineering flaws presented in this section are similar to
various people and business plan flaws that were discussed in earlier chapters. The
flaws range from lack of technology, either because extensive research is needed or
because the technology is ubiquitous and trivial, to simplyhaving a poor team. As with
other types of flaws, predicatinga high-techventure on technologythat is flawed in one
or more respects could prove to be fatal.

TACKLING A PRODUCT THAT REQUIRES

SIGNIFICANT RESEARCH TO MAKE IT FEASIBLE

A wonderful product that is clearly needed is just waiting to be developed. Designing

the product, however, will require an unknown amount of basic and applied research.
As of 1990, the estimate of when such a product can be produced ranges from now to
 Technology and Engineering Flaws 127

~~ ~~

Table 6-1. Technology Balance Sheet for Ardent.

-
Technology Base Hardware and software release
Packaging, mechanical design specifications
(includingthermal and acoustic Chief Technical Officer,
analysis) TearnlCulture, Architect(s)
Industrial design
Discussed in an earlier sectionof the
Digital systems design chapter.
Signal propagation,
electromagnetic interference, Technical Resources
and radio frequency
interference Computing environment
General logic design Multisegment Ethernet and
Gate array design Appletalk
Testing Macintoshes for documentation
Sun Workstations (local and
Architecture/implementation windows)
Vector multiprocessing MIPS file and computation
Performance analysisand servers
simulation
Graphics Valid logic for logic design
Mass storage and input/output Verilog for system description/
Image processing simulation
Software Technology Future
Operating and file system
Language and compiler design Plan outline for next products
Graphical user interface
Database Standards
Quality assurance
See Chapter 8, Figure 8-5 for product
Marketing, sales, and product standards.
support
Benchmarking Process Definitions
Mathematics and scientific
progress New products introduction
Signal and image processing
Visualization Plan
Computational chemistry Embodied in master schedule
Computational fluid dynamics
Mechanical CAD and finite Yearly budget with all resources
element modeling Operational Management and Control
Seismic processing
Technical publication Schedule fantasy factor = 1.2 after a
major organizationalchange
Engineering Specifications
Weekly schedule review at each level
Reference manuals for all Staff meetings at each level of
components management, with minutes and
Principles of operation for hardware action items
Eight-corners test Management by objectives
Products committee to track/
Manufacturing Specifications coordinate all products and future
product plans
Test vectors and specifications for all
chips and boards
128 The Technology Balance Sheet

eighteen months from now to never (althoughnever is a word that cannot really be used
when it comes to technology).The following exampleillustrates the slow evolution of
a product whose development has required (and will continue to require) a consider-
able amount of research.

The speech typewriter (speechwriter). Kurzweil AI was formed in 1982 to build a

speech typewriter. Its founder, Ray Kurzweil, has produced an impressive array of
inventions, including the first machine to read to the blind (1972),which does optical
character recognition of variable fonts and is connected to a speech synthesizer. The
company developing the reader was sold to Xerox. A second firm, which was formed
in the late 1970s to build keyboard-controlled music synthesizers for the professional
and home market, is now for sale in 1990.
The aim of speech research, which has been under way since the 1930s, is to
understand speech well enough to permit it to be recognized by a machine. In 1980,at
least one market research firm published a report estimating the market for voice-
activated typewriters at $3.5billion in 1990.Kurzweilbelieved that enough was known
about speechunderstandingto finallybuild a comparativelyelementarybutnonetheless
useful product that would function within certain limited contexts,such as having the
machinerun by a single, trained operator who would use a large,but limited vocabulary
and speak separated words.
Kurzweil’s first task was to advance the art on which to base a product. In order to
bring himself up to the state of the art in speech recognition, Kurzweil put together a
team from the research community at MIT and Harvard to develop technology for
speech understanding. In 1985, the firm introduced its first product and tried to sell a
recognizer to a number of softwarecompanies (whoseproducts included spreadsheets,
word processing, databases, CAD, etc.) as a control mechanism, but the product’s
capability and accuracy were limited and it worked poorly. Furthermore,users had to
”train” the recognizer. The Kurzweil AI product predated a product by Articulate
Systems (using Dragon’s recognizer) to control the Macintosh.
By 1989,the Kurzweil product had evolved into a unique voice editor that runs on
a PC and is capable of recognizing keywords and expanding them using a word
processor database and report generator. The voice editor is tailored to a particular
applicationby its vocabulary and phrases and is then further tuned by the user. In 1990,
the product isbeing successfullysold for writingreportsin internalmedicine,pathology,
radiology, and emergency medicine, since these fields all require reports based on
distinct, limited vocabularies.
In contrast to speech-research laboratories such as Bell Labs, IBM, and university
laboratories, Kurzweil has advanced toward the goal of a typewriter by building and
marketing a product. Other companieshave alsobuilt and marketed speech recognizers
for limited use. Unlike other laboratories, NEC has been marketing limited vocabulary
recognizers for almost a decade in order to really understand their problems and use.
 Technology and Engineering Flaws 129

Thus, for a researcher, a start-up is an interesting alternative to the large company or

government-fundedlaboratory, assuming the firmcan find investorswilling to wait for
their investments to mature. Dragon Systems, Inc. (page282), provides an alternative
role model for how a venture requiring a slow-to-emergetechnology may be formed.
No doubt the hottest product-the one that absolutely everyone will have, need,
and use after 2001-will be the universal speech typewriter!And the next advance will
probably be a speech typewriter that does on-the-fly language translation.

REQUIRING A TRILOGY OF BREAKTHROUGHS

It has been observed that a successful start-up cannot be based on more than two
breakthroughs in the state of the art. And for each of the areas requiring a breakthrough,
an alternative technology should be available as a backup. Clearly, a risk exists when
three or more technologieshave to be understood ke., researched to the point of being
usable) and developed. It is almost assuredly fatal for a start-up to engage in research
whose result cannot be known or scheduled, because the company’s other functions
must all be supported in the meantime, and the fundingrequirementsare uncertain and
often open-ended. The schedule for such a project contains loops, parallel and redun-
dant exploratory paths, and conditionalbranches.
The following example discusses Trilogy, Inc., which attempted to develop a
product requiringmultipletechnologcalbreakthroughs.The “trilogyofbreakthroughs”
flaw is in fact named after Trilogy, since this flaw contributed greatly to the difficulties
the firm encountered.

Trilogy,Inc. Trilogy was started to develop an IBM-compatibleline of computerswith

major subsystems packaged on a single semiconductor wafer. Unisys and Digital
invested in the technology as codevelopers.2The risks included the following:

1. Interconnecting high-density, high-speed semiconductor circuitry on a single

wafer

2. Devising a scheme to ensure defect-freeparts using redundant parts of a wafer

3. Packaging an entire wafer such that power is input, heat is dissipated, and the
wafer is rewired to circumvent inherent wafer defects

4. Developing a CAD system to manage the redundancy-based logic design and

interconnect scheme

2. I made this recommendation. After Trilogy failed, Digital bought rights to all its technology. The power
supply, heat sink, wafer-packaging scheme, and facilities were used as the basis for the VAX 9000.
130 The Technology Balance Sheet

5. Developing a computer design more complex than previous designs

Some observersfelt that Trilogy’spleasant facilities and large staff were fatal flaws.
The real culprit, however, was that the requisite technology could not be developed in
time to implement a product. The five risks listed above had the following outcomes:

1. The circuits were slower than specified, increasing the design’s complexity
while decreasing its competitiveness.

2., 3. Not enough redundancy was available to cover wafer faults.

4. The CAD system was quite slow and decreased productivity.

5. The design was so complex as to increase the design time and adversely
affect product competitiveness.

Althoughthe preceding problems occurred during the productdevelopmentstage,

the issues were known at the concept or seed stage. In hindsight, an analysis of the
situation should have produced an emphatic ”no go” until the required breakthroughs
were reduced to a manageable number.
When it became clear that Trilogy’s technology was inadequate to build the
product, the company acquired Elexsi Computer with its remaining capital and
attemptedto makeit succeed.Unfortunately, minisupersfromAlliant and Convexwere
also being brought to market at that time.

HAVING LITTLE OR NO SUSTAINING TECHNOLOGY

Offering just another commodity product of a particular type (i.e., “brand X ) in a

crowded field is usually a fatal flaw. Starting a company with commodity technology,
such as a new chip, is the opposite of the trilogy-of-breakthroughs flaw. It comes from
the belief that the firm has just a slightly better idea about the product or how to sell it.
The minicomputer, PC, and workstation industries all began as technology companies
to a greater or lesser degree, and the introduction of various components (SSI/MSI, 1 6
bit microprocessors, and 32-bit microprocessors, respectively) allowed dozens of no-
tech companiesto enter the market. In early 1990, the smallest PC electronicsassembly
costs $200, and within five years, just one or two very-high-tech chips (available from
Intel and a memory supplier) will form the entire, minimal PC with 2 megabytes to 8
megabytes of memory. Dell Computer is an excellent example of how a company was
able to get started and grow with PCs despite the low-tech odds, because Dell
considered the whole environment of product, sales, service, and support.
 Technology and Engineering Flaws 131

THE NOT-INVENTED-HERE (NIH) SYNDROME

One of the most dangerous flaws is a form of technical arrogance in which a company
feels compelled to reengineer every part of a hardware or software system because it
believes that it can do a better job than any of its potential suppliers. For a new venture,
inventing every possible component in order to make an ultimate product (instead of
buying everything possible in order to get to market rapidly with a good product at the
lowest development cost) is often fatal.
The other effect of the NIH syndrome is the incompatible-productflaw (page 190).
A company designs a new interface, such as a programming language or a feature for
an existinglanguage,when an old one would have been just fine.In this case, NIH hurts
the buyer, who has to change and adapt to something different. Needless innovations
and changes that have the effect of rendering hardware, programs, and data incompat-
ible are extremely costly for the whole computing enterprise.
The NIH syndrome is endemic among most engineers, especially in the United
States,the United Kingdom, and France. NIH does not necessarilyhave anything to do
with a team’s competence,only its lack of business sawy, although the brightest teams
are often the most unhappy about using less-than-perfect components. The NIH
syndrome’s effects on productivity and on profit and loss are devastating, and this
syndrome may account for why Japanese engineers are at least twice as productive as
American engineers in a field such as automotive engineering.NIH often triggers the
formation of multiple companies in one, a type of business plan flaw that is described
on page 50.
Even well-established and well-respected firms have exhibited this flaw. In the
early 1960s, IBM found that every computer products group was building a
computer based on each group’s own logic circuits, requiring redundancy in design,
manufacturing, and field spares. Gene Amdahl proposed that any group using
components from another group be rewarded and given special recognition.One of
his coworkers squelched the idea, claiming that ”it’s Un-American.”

THE MISSING COMPONENT

Every day that an organizationdepends on a risky part or a marginal vendor, it risks its
life, because if a critical component (or process) fails to materialize as scheduled, the
company may run out of time and, hence, out of money. Selecting poor vendors is a
common and hard-to-avoid error. Only through experiencewill a start-up learn which
firms can be trusted to meet their commitments.
Henry Burkhardt, CEO of Kendall Square Research, described the problems of
selecting the right vendor by offering what might be called a “tale of three cities.” In it,
132 The Technology Balance Sheet

he compared the experienceof dealingwith vendors in a Texascity,a Californiacity,and

a Japanesecity:

Texas: W e have the fastest, biggest, and cheapest parts. If you don‘t believe it, write me
because I’m the president of this new division. (Theydon’t really have a competitive
product. On calling them, the secretary to the president states that you have to
write because theletter goesdirectlyto themarketingVP. I wroteto thepresident
and informed them they lie about their parts and even lie about their willingness
to listen. The letter does go through the company like wildfire, but the division
president is still there, selling the same parts in the same old way.)

California: Everyone knows our parts are the fastest and the biggest. W e started the
industry. (We ask for a delivery commitment. It reads ’We’ll make our best
effort to deliver.” On inspection, the parts fail after a year without special
treatment that’s not part of the specification.The customers all complain about
missed delivery schedules,and manufacturing people scream when they hear
the name of the company. Every transaction with the company requires nego-
tiation.)

Japan: W e h u e fast, large parts as stated in our specs, and we are committed to higk
quality. (Existingcustomersagree, and no one canidentlfy a part ever failing.We
selected them because the contract simply states that they will meet their specs
and deliveries.All specs and delivery dates were met.)

The following story illustrates the type of havoc that can ensue when a company
deals with a poor vendor.

WAVETRACER. In building a signal-processingcomputer,WAVETRACER used an

unreliable printed circuit board vendor to make its prototypeboards. The boards had
numerous errors, costing the firm several months over its plan at a critical time when it
needed a product and credibility with its first customer. Because of this schedule slip,
WAVETRACER was forced to seek additionalfinancingearlier and in a greater amount
than would otherwise have been necessary. The valuation was decreased and the
external ownership increased.

New microprocessors have historically had bugs. New complex microproces-

sorsfrom semiconductor companies-including Intel, Motorola, and National-have
all had bugs. The more complex the part, the more error-prone it is; hence, another
reason for RISC. The first users are able to help find new flaws and often rediscover
flaws that manufacturers forget to address. Apollo, Sequent, and several other
companies have war stories to tell in this regard.
 Technology and Engineering Flaws 133

INABILITY TO HIRE THE ENGINEERS

Hiringis absolutelycritical, yet every high-tech venture I know of has had more trouble
hiring than it ever planned or imagined. This leads to an additionalflaw-lowering the
standards. By reducing its standards,the firm risks producingboth a downward spiral
in quality and a bloated staff that generates no meaningful output. A pygmy heading
engineering will proceed to hire even smaller pygmies.

FAILING TO GET RID OF POOR HIRES AS SOON AS POSSIBLE

If a person is found to be a poor hire, he or she must be dismissed at the earliest feasible
moment. Negative producers3should be terminated immediately, placeholders very
rapidly, and marginal producers as soon as possible.

CompanyX. Iknowofa firm(let'scal1it "CompanyX) thatwashavingtroublestaffing

a new project with good people and made a borderline hire without proper reference
checking. When the team discovered that the borderline individual was in fact a poor
hire, they felt they could manage him by close supervision and checking. However, he
refused to ask for help, chafed at havinghis workreviewed,and waslate-all sure signs
of a bad design(er).Simulation revealed continued bugs with no evidence of progress
toward a correct design. In essence, bugs were just being moved around. When
Company X finally conducted a design walk-through, the engineer quit and went to a
competitor, where he may or may not have greater success. Although Company X did
nothing to influence its former engineer's selection of a new employer, outplacing
negatively productive people with a potential competitor can do wonders for a firm's
competitive lead.

LEAKING TECHNOLOGY AND PRODUCT IDEAS

If a new venture permits its technology and product ideas to leak, it risks giving both
established competitors and other start-ups an opportunity to respond. It is therefore
important that the staff say no more than is absolutely necessary in order to sell new

TNegative productivity is a principlethat I claim is worthy of a Nobel Prize. Normal principles of

productivity assume that workers create positive output. Brooks refined the concept of software
productivity to express it in terms of the "mythical man-month,'' and in software engineering, it is
understood that different programmers vary in their productivity by several orders of magnitude.
According to theprinaple of negative productivity, it is possible for an individual to produce bad results
that others must then rcdo; hence, someone who is very negatively productive can kcep a whole team
busy with damage control, preventing the team from producing any output whatsoever.
134 The Technology Balance Sheet

recruits.They should try to get recruits to tell more about themselves than the company
tells about itself and avoid any mention of costs and schedules.

PREANNOUNCING THE PRODUCT

It is absolutelyfoolhardy to preannounce a product before it has been tested internally

and passed its acceptance tests. At the very least, preannouncement is likely to be an
embarrassment;at worst, there might be legal repercussions.
In no case should a product be officially announced before it is operating well
enough to pass formal tests that are comparable to actual customer use. Ideally, the
product announcement is made at the end of beta testing at customer sites.Anything
less conservative is a flaw.
This is one flaw that is even more painful in large companies than in start-ups. In
1966, IBM preannounced a large computer that would compete with Control Data
Corporation's 6600 in an attempt to get customers to wait for the IBM product, which,
in this particular case, never came. CDC sued IBM and was awarded $600 million in a
consent decree that forbade preannouncement.

TECHNOLOGY BALANCE SHEET RULES

The following is the fundamental rule for evaluating a new venture's technology:

Has the company generated and maintained a complete "technology balance

sheet" that is adequate to develop the product and specifies the information
listed below?

"Buy-out" technology (software and hardware), including semis, etc.

Patentable or unique componentsthat are the basis for the firm's future

Industry and de facto standards that the start-up must "track" or advance

The company's own standards or ways of doing things

Patentable or unique processes, including design

Plan, with schedule and resources

Engineering and manufacturing specifications

Chief technology officer he., the vice president of engineering)

 Technology Balance Sheet Rules 135

Team
Product architects and architectural processes

People (including consultants) who embody the technology

Computer-aided design (CAD) and computer-aided software engineering

(CASE)tools, computer resources, and network environment
Ability to acquire future technology

Operational management control

The following are some specific rules applicable to the technology balance sheet.

Can the team, at the concept stage, show how all the technology will come
together to form a product that will be not only unique but also self-sustaining
(i.e., capable of evolving into future generations)?

The technology balance sheet should be used to account for both uniqueness and
mastery of the technology.Mastering the technology means being able to assemble the
"to be acquired engineering team, consultants, patents, standards, components,
design process, CAD tools, etc. This rule tests whether the organization has a way to
evolve its product and extend it into future generationsor whether it is merely starting
on a one-shot basis.
The same rule should be applied again at the seed stage, continually challengingthe
foundersabout the uniqueness of their technology.It examineswhether the technology
remains sufficientlyunique, yet implementable,to support a self-sustaining company.
The rules in Chapter 8, "The Product," also examine uniqueness.

Can the team, at the concept stage, show how the technology can be developed
while requiring fewer than three breakthroughs or significant advancements
in the state of the art?

This rule tests whether the technology is too high (sometimesreaching infinity),
such that the new venture is engaging in research instead of product development.
Applied research or advanced development is being done if a project schedulecontains
major loops with conditionalbranches or multiple exploratorypaths in its PERT chart.
Such a company is likely to be fatally flawed if it has been funded with the goal of
developing a product, as opposed to being funded as a research and development
partnership. In the latter case, investors are cognizant of the risk, and the goal is to first
master the technology before building a product.
136 The Technology Balance Sheet

Does a simple product development plan, specifying resources and schedule,

exist at the concept stage?

This rule tests whether the start-uphas a plan outliningthe stepsand resources that
will be required to develop the product.

Does the company have a working product or product prototype and people
who understand it?

Ideally, a high-tech venture is based on a working product or product prototype

that has been funded by a public institution together with the people who understand
and embody the technology, even though such products and people may fail the
experience tests required by many financiers.
The next best thing is to base the company on key people who have pioneered in
developing technologicalcomponents. They must have a thorough understanding of
the product development process gained through building products for use by others
and must be committed to engineering design rather than research.
Probably the worst alternativeis to base the firm on the results of military research
and development,because it is likely to be fatally flawed, as described in “Augustine’s
Laws” (Augustine, 1987).Military products are cost- and reliability-insensitive.They
don’t have to work or are rarely tested to ensure that they work. The development
budgets, lead time (measured in decades),and quality of military products are outside
commercial bounds.

Has the company’s proprietary technology been demonstrated during the seed
stage via physical or computer model, breadboard, or some other form of
demonstration that would prove its viability, such that the development
breakthroughs have been reduced to a level of risk that is acceptable for the
product development stage?

This rule verifies that the start-up is in control of its technological destiny by
checking whether the seed stage requirements of reducing risk have been satisfied by
constructingbreadboards, models, or demonstrations of critical technology.Ideally, at
this point, the firm has ideas that may result in copyrightsand patents in order to protect
and enhance its technology.

If the company is depending on a concurrent breakthrough or leading-edge

product from another supplier (e.g., a component or system vendor), have the
risks been clearly identified and factored into the plan?
 Technology Balance Sheet Rules 137

This rule determines whether the start-up's risks have been transferred to an
outside vendor and then assesses the overall risk in using such a vendor. Information
about the vendor's past performance is required, especiallyevidenceof its reliability in
meeting delivery schedules. Founding a company predicated on the availability of a
componentthat a manufacturerhas never beforebuilt is always risky. The new venture
gets no points for picking the best technology or engineeringthe lowest cost if it is then
unable to obtain a key part or unable to obtain it on time or in manufacturingquantities.
The evaluation of vendors and components is an excellent position for a seasoned
engineer, by the way. Such individuals know which components and suppliers are
lugh-qualityand reliable.Newengineers,ontheotherhand, tend tobelievespecifications.

Does the chief technicalofficer have the capabilityand stature to hire, lead, and
manage a superb engineering group?

The general qualificationsof the CTO must parallel those of the CEO, because he or
she is the "clock" and "standards setter" for engineering.The CTO should have a track
record of both technical and managerial accomplishment.The CTOs technical back-
ground must be solid enough to gain the engineers' respect and confidencein his or her
technical decisions. The CTOs managerial slulls must be strong enough to deal with
conflicting egos, limited resources, and all the other trials and tribulations that a
manager faces.This individual should be especially talented at recognizing, selecting,
and encouraging top-notch engineers.

Does the product have an architect with proven experience?

As stated previously, the product architect is likely to be the most critical person
within the engineering function. His or her key job is to guide the product's
introduction and evolution over the course of its lifetime, and a track record of
success in past endeavors is the strongest possible recommendation. In some cases,
several architects may be required as a product is broken into various parts, but the
boundaries of each architect's responsibilitiesmust be clear, and the architects must
be capable of functioning as a cohesive team.

Are key technologists,or avenues for hiring them, available?

In one sense, this rule relates to the question of whether the company has the "right
tech (i.e.,an appropriatelevel of technology).If the technologyupon which the venture
is to be based is so "far-out" that only a handful of technologists skilled in that art are
available, the firm is likely to have serious staffing problems. On the other hand, if the
138 The Technology Balance Sheet

start-up is to be based on an ingenious use of a recently introduced or established

technology, hiring prospects will be much better. Some innovative ways of finding
appropriate personnel were discussed earlier in the chapter.

Does the company have hiring criteria, and is there a systematic recruiting
process?

Thisrule checkswhether the firm has established hiringcriteria,coveringboth work

habits, management ability, and technical skills. Having specificationsfor each person
to be hired is helpful and perhaps essential. In addition, the company needs a first-rate
process for initially identifying potential employees and then bringing them in for an
interview, screening them, and finally selling them. A critical part of the process is
thorough reference checking of all candidates!

Does the job candidates’ prior experience show evidence of operationalman-

agement ability as well as resources- and schedule-planningability?

This question examines the planning and management history of the engineer/
managementteam. History is likely to be the best predictor of a manager’sability to help
people enjoy their work and be productive in it. And with regard to scheduling,if the
candidates have historically been on time, then they will most likely continue to meet
their commitmentsin the future.

Has engineeringoutlined a qualitydesign and product-releaseprocesstogether

with engineering,manufacturing-engineering,and product-releasestandards,
including,for example,coding practices, designrules, code walk-throughs,and
design reviews?

This rule measures the existence and effectiveness of the company’s engineering
design process. For a software team, it would not be unreasonable to ask whether the
process at least satisfies the Software Engineering Institute’s process-capability re-
quirements for level 1 and what plans exist to upgrade the process so it will satisfy the
requirements of increasingly higher levels (Humphrey, 1989).
It is not uncommon for engineers to react negatively to the establishment of
standardsand processes.For example,engineerswho havejust left largeh sfrequently
rebel at anything that might look like bureaucracy or restrictions on their freedom, and
engineerscoming from a research environment are unlikely to understand the need for
any rigor in standards and processes. Object-oriented programming languages and
methods promise to make the task of building softwaresubstantiallyeasierbecause they
enable modules to be built in a more isolated and independent manner and because
more software is likely to be available from other sources and to be reusable.
 Conclusion 139

Is a productdevelopmentschedulein place,and does it specifygross milestones

and resources?

This process question examines whether the start-up has a schedule for the project
with enough intermediatemilestones.Without such a schedule,it is impossibleto make
a meaningful business plan. People experienced in high-tech ventures know that it is
essential for the company to be operating accordingto a detailed schedule,even though
no schedule can be fully validated until the entire team responsible for the project has
been hired and brought on board. An unwillingnessto make a detailed scheduleat this
point is therefore a good early warning indicator that the project will probably be
difficult and unpredictable. A start-up can certainly be financed on an open-ended
schedule,but this approach can be expected to increaseproduct development spending
by at least a factor of 2.

Does the company have a plan for acquiring and operating CAD and CASE
tools, computing resources, and its network?

Developingproducts based on up-to-date technology requires up-to-date engi-

neering tools. Tools represent both a key part of engineering and a large fraction of
product development cost. A CAD program for schematic capture or board layout
can cost several hundred thousand dollars. A simulator to accelerate the testing of
a complex chip may cost half a million dollars. Thus, it is critical for the start-up to
prepare a detailed list of all the tools (both computers and the necessary networks)
it will require for high-tech hardware and software development. In the early stages,
developers often administer their own systems, which may include interfaces with
various national and international wide area networks, but as a company grows, the
expense of systemadministrators and network administrators must alsobe assumed.

CONCLUSION

Chapters5and6 have presented a pictureof highinformationtechnologyandexamined

how a new venture uses technology to engineer products in a timely and predictable
fashion.At each of the development phases described in Chapter 5, the company must
have an adequatetechnologybalance sheet covering the following twelve dimensions:
its technologybase; standards;design,quality,and other processes; plan, with schedule
and resources;engineeringspecifications;manufacturingspecifications;chief technical
officer; team and engineering culture; architecture; technical resources; technology
future; and operational management.
Chapter 7

MANUFACTURING

The manufacturing organization buys materials and converts those materials into
products according to the product and process specificationsdeveloped by the engi-
neering organization. Manufacturing is measured on its ability to do this in a cost-
effective,hgh-quality, and timely fashion.A major portion of this effort is the manage-
ment of raw materialsand finished goods, which is a balancing act.Enough of each must
be on hand to give the company flexibilityin dealing with fluctuationsin the order rate,
but not so much that it feels a financial impact from having excess inventory.
The importance of manufacturing varies with the business in which a particular
high-tech venture is engaged. For example, manufacturing operations in a software
company primarily involve the reproduction of magnetic storage media and manuals.
This process requireslittlecapitalinvestment, and the cashvalue of the work in progress
is low.At the oppositeextremeis semiconductormanufacture,especiallythat involving
an advanced process. Here, the capital investment is huge and the work in progress
more valuable. The following is a spectrum (in ascending order) of manufacturing
complexity and expense for various computer-relatedproducts:

Software-reproduction of magnetic storage media and manuals

Printed circuit boards and/or assembly of small components

Small systems (e.g.,terminals, printers) involving low technology

Systemsinvolving a unique or proprietary technology (e.g.,print heads, scanners)

General-purposecomputers (collections of boards)

140
 The Sanders Guiding Principles for Manufacturing 141

Complex electromechanicaldevices (e.& disks)

Semiconductor manufacturing involving advanced processes

The essence of manufacturing is being able to plan the output. However, many
start-ups go through dramatic changes in their plan during the market development
stage.At first, there may be no demand for the product whatsoever;then, demand may
suddenly increase beyond manufacturing’sproduction capability.The manufacturing
organization is necessarily slow to respond because the typical lead time for materials
(semiconductors,disks, printed circuit boards, etc.) is sixteen weeks, followed by four
weeks of process time for the product. In other words, a total of five months normally
elapse between when the company places orders for materials and when it can deliver
its product. The slowness of manufacturing’s response time may tempt a new firm to
rush headlong into mass production so that it will be able to meet all its orders, but ths
is often foolish. A few guiding principles for start-up manufacturing are therefore in
order.

THE SANDERS GUIDING PRINCIPLES FOR MANUFACTURING

Matt Sanders’offers two general principles for manufacturing:emphasize quality,and

m i n i m i z e the use of the start-up’s resources (capital, time, and space). These two
principles form the basis for eight guidelines, which are examined in detail in the
following subsections.

ONLY BUILD PRODUCTS OF THE HIGHEST POSSIBLE QUALITY

The only acceptable engineering, product, and manufacturing strategy is to build

products of the highest possible quality. Anything less than the highest quality is likely
to prove extremely costly in the long run. At the front end of the manufacturingprocess,
using or accepting poor-quality components is costly in terms of increased inventories
and additional work. Likewise, implementing a poor-quality design is costly because
redesign and rework will continuallybe required while the product is being produced.
Finally, if the product fails in the field, an expensiveserviceorganizationwill be needed
to maintain it. Any product yield of less than 90 percent at customer sites represents a
serious product design and quality problem. A 95 percent to 99 percent yield should be
the target for the initial products, with 99 percent (or better) the target for steady-state
production.

____
1.Matt Sanders is a founder of ConvergentTechnologies and Ardent and was the principal responsiblefor
establishing the manufacturing organization and operations of both companies.
142 Manufacturing

Quality must be designed in right from the start; it cannot be added on by

manufacturing. A start-up should focus on a simple first product, since this
approach allows the company to get to market quickly and increasesthe probability
of its having a really well-designedproduct. Attention should be paid to minimizing
components,not only to decrease cost but also to increasereliability (parts that aren’t
there can’t fail).Quality is a discipline that concernsboth engineering and manufac-
turing, and engineers must understand the manufacturing process by which their
product is fabricated in order for the firm to produce the best product.
An example of a company that has emphasized the relationshipbetween engineer-
ing and manufacturing is Sequent Computer Systems.It has taken a simple step to en-
sure that manufacturabilityis a key part of the design-namely, Sequent permits any-
one to stop the manufacturing line for any reason. The manufacturing and design
engineersresponsible for the product can only restart the line after the problem has been
remedied. By giving so much power to those building the product, Sequent ensures that
engineeringdelivers perfect specifications,and that if it doesn’t,problems are attended
to immediately.Under this system, design engineers quickly become expert manufac-
turing engineers.

ONLY INVEST IN MANUFACTURING IF THE PROCESS IS UNIQUE

AND IS THE ESSENCE OF THE START-UP

Today’s start-upshould invest in manufacturingonly when the manufacturingprocess

is an essential and proprietary part of the company.Examplesare firms manufacturing
complex electromechanicaldevices, semiconductors, and some proprietary parts of a
larger product. In contrast, computer systems ventures should minimize their invest-
ment in manufacturingprocessesand seekhigh-quality subcontractingsourcesinstead.
It is common for a start-up to want to make everything it can in order to have
”control” over its destiny. For most new ventures, however, buying manufacturing
capabilitiesfrom outside sourcesnot only is a better use of resources but also is likely to
yield higher quality and lower costs, because the subcontractors specialize in all the
necessary testing and fabricationsteps. Sincethe volume of products is likely to be low
at first, the best use of resources is to buy as much as possible from outside sources to
avoid investing in new processes.

EXAMINE EVERY MAKEBUY DECISION

In addition to the decision regarding in-house manufacturing versus subcontracting,

additional make/buy decisionsmust be made with respect to all parts of the enterprise,
including product design, design processes, sales, service, and support. The start-up
may find it cost-effective to buy one or more of these capabilities from an external
organization.
 The Sanders Guiding Principles for Manufacturing 143

GET TO MARKET FAST

If the company plans effectively from the outset, it can get to market rapidly and
minimize its investment in manufacturing,whether that be in-housemanufacturingor
subcontracted manufacturing. The key to time-to-market and product quality is for
engineeringand manufacturingto functionwell togetherfromthebeginning.Oftentimes,
engineeringwill build the first product prototypes and then (whenthe engineers have
learned how to build the product) turn the process over to manufacturing. This
approach leads to delays and keeps the manufacturing organization from hitting the
ground running. It is wiser to give manufacturing responsibility for building all the
products, including the prototypes.

USE MINIMAL CASH

A good way to leverage the firm’s cash is to minimize inventory through design and by
using outside suppliers. The start-up should get a subassembly supplier that will fund
the inventory and give favorable payment terms. By making inventory part of the
product cost, the company can convert what would otherwise have been a fixed
manufacturing cost to a variable cost. It can also negotiate flexible terms for varying
quantities in order to reduce the cost due to unpredictably fluctuating volumes.

HAVE ONLY A MINIMAL STAFF, BUT HIRE THE CRITICAL PEOPLE

Farming out everything that it can will save the new venture not only on capital
equipment and inventory costs but also on personnel expense. Once a company hires
someone, it has a commitment to that person. Indirect manufacturing personnel are a
fixed expense, not a variable cost, and the goal of start-up manufacturing must be to
push manufacturing spending into the variable-cost category as much as possible.
Therefore,instead of hiring a staff of specialistsand training them from scratch, the firm
should use subcontractor personnel who have already passed through the learning
curve. This approach is likely to be cheaper and result in the production of better
products. Using a range of subcontractors does require the company to have the
appropriate logistical systems and personnel to handle coordination,however.
Although most of the advice given so far has been to minimize cost and hire as few
people as possible, when people aye hired, it is important to hire the right ones. The head
of manufacturing is one such critical hire. A materials person who understands the
procurement of top-quality components is likely to produce the highest payoff. A
person assigned solely to work on quality will producethe next highestpayoff. The final
members of the team should be responsible for testing and for developing unique
processes (if either of these will be done in-house).
144 Manufacturing

MAKE SURE PRODUCT COST IS PREDICTABLE AND LOW

Predictability comes from understanding product cost. As stated above, the start-up
should make its costs variable rather than fixed, insofar as possible. This means
religiously tracking the parts list during the design process and making sure that
product cost and quality are major design constraints, not afterthoughts. Paying strict
attentionto quality and establishing a cooperativerelationshipwith subcontractorsand
vendors will help ensure that costs are predictable.
The cost of manufacturing a typical computer product, once the assembly process
reaches steady state, is:

80-85 percent Materials

10 percent Indirect labor (fixed) for salaried personnel and supervision

2-3 percent Direct labor

2-3 percent Depreciation of facilities and equipment (for simple products)

Readers should note the importance of the cost of materials and indirect labor.

AVOID THE EVIL OF INVENTORY

As stated earlier, material control is a balancing act. Having too much material means
that all the start-up’scapital may be tied up in inventory.On the other hand, having too
little material means lost sales opportunities. The balancing act is complicated by the
long built-in delay (typically five months) between the time materials are ordered and
the time finished goods are ready to be shipped. Inventory is a very important area on
which to focus management attention, because it is the biggest cash sink for a high-
growth venture-and the place where the company can be lost.

OFFSHORE MANUFACTURING
Offshore manufacturing has enabled numerous U.S. start-ups to follow many of the
guidelines given in the preceding section. The relationship between Stardent and
Kubota is an especially good example of the merits of subcontractingmanufacturing
and paying careful attention to quality. In the Stardent/Kubota relationship,Stardent
is responsible for designing basic hardware and software for its Titan workstation as
well as acquiring softwarein each market area (e.g.,chemistry and imaging).Kubota is
responsible for all manufacturing.
Partnership with a Japanese firm has had an especially significant impact on
product quality. In the late 1940s, American manufacturing expert Edward Demming
 Manufacturing Flaws 145

visited Japan and told the Japanese the importance of making quality the number one
priority. He emphasized that a faulty part had to be either reworked, thrown out, or
(worst of all) used in the product. He proved all three of these alternatives to be more
expensive than making the part correctly the first time. Being a statistician by training,
he also showed the Japanesethe evils of ”tolerancebuildup”-the cumulativeeffect of
using a number of parts that are eachbarely in-specification,which can result in a faulty
final product unless the tolerance specificationsare tightened to prevent this.
Japanese manufacturers such as Kubota have learned this lesson well. Kubota is a
century-oldfirm that manufacturesmechanicalequipmentand alsodesignsmechanical
engineeringsoftware and integratesMCAD (mechanicalcomputer-aided design) and
CAM (computer-aided manufacturing) software. Since the fabrication of computers
fundamentally involves mechanical assembly, Kubota is able to use its manufacturing
skills to produce a high-quality product. Its plant is run by engineers who understand
the fundamentalsof the materials and processes required to form the parts and know
how to combine them into a hgh-quality product. American computer manufacturers,
in contrast, are oftenheaded by eitherMBAs or individualswho have worked their way
up through the ranks without coming to understand the total picture of their operation.
Kubota’s dedication to quality is reflected in the failure rates of the two products
built by Stardent’s predecessor companies, Ardent and Stellar. The failure rate of the
Ardent workstation (built by Kubota) was half that predicted by the parts count, while
the Stellar failure rate was equal to that predicted by the parts count. Once Ardent and
Stellar merged, and Kubota began manufacturingboth products, the failure rate of the
original Stellar product improved toward that of the Ardent product.
It would be unfair to give all the credit to the Japanese,however. In addition to the
excellenceof Kubota’smanufacturing,severalother factorscontributedto the reliability
of Ardent’s Titan workstation.First, Ardent’s mechanicaland electricaldesignerswere
Hewlett-Packard alumni, and HI’ is an ideal training ground for engineers who build
reliable products (albeit expensively). Second, Ardent had stringent standards for
design quality. Third, the engineers tested the design rigorously at all eight corners of
operation (allpermutations of high and low values of voltage,temperature, and speed).
Fourth,Kubotainsistedona”perfect”designin0rderthat theproductbemanufacturable.

MANUFACTURING FLAWS
Nearly all of the flaws discussed in this section result from having a poor plan. Some of
them result from failingto adhere to commonsenserules of good practice to reduce risk.
Many of them affect quality,clearly a criticalfactor in the case of manufacturingoutput.
The first flaw involves a quality problem that manifests itself in manufacturing,
although its root cause may lie in either manufacturing or some other part of the
organization.
146 Manufacturing

A PRODUCT MANUFACTURINGLINE WITH A HIGHLY ERRATIC FLOW

A manufacturing line may flow extremely erratically for a number of reasons:

The line may be poorly designed and may run only rarely.

Its yields may be inadequate because of poor materials or poor training.

The vendors may be unreliable.

The sales forecasts may be erratic.

It is unusual to see a new venture with a really fine manufacturing facility unless
manufacturing is the company’s dominant focus. The assembly line often runs poorly
for one or more of the above-listed reasons. Thus, right from the start, the firm is likely
to get a reputation for unpredictable product quality.

POOR MAKE/BW DECISIONS

A company I know of that started to build a system to eliminate paper in a very large
office provides a good example of the problems that can arise when a firm’s make/buy
decisionsare poorly thought out. The system was supposed to scan everypiece of paper
entering the building and convert it to image format. From then on, all storage,
transmission, and viewing would be via computer. The company began by building
every component of the system: a jukebox to manage the optical disks on which the
informationwould be stored, scannersand viewing computers,all the computers to be
used throughout the network, and all the applications software. Although the firm
could have bought virtually all the computers and workstationsneeded for the system,
it designed every component itself to get the lowest manufacturing cost, even though
the entire system cost several million dollars.
The firm ultimately had to be downsized to supply only the large file systemsusing
the optical storage jukebox that it manufactured. It never got around to building the
software to manage the elimination of paper because it had spent all its resources
attempting to reduce the cost of components that it could have bought off the shelf.

HAVING A MANUFACTURING FACILITY THAT IS TOO BIG

A high-tech venture courts trouble if it builds an extensive manufacturing facility in

anticipation of high volumes before the product has even been introduced to, or
accepted by, the marketplace. Some companies,such as NeXT, have survived this flaw,
but it is nevertheless dangerous. As stated above, most start-upsfind it best to conserve
their cash and use subcontractors.
 Manufacturing Flaws 147

USING A CRITICAL, BUT MARGINAL, COMPONENT OR PROCESS

If the product is based on a critical, but marginal, component or process, the net result
is a product that is poor (e.g.,unreliable, very costly due to work-in-progressdelays, or
not producible in adequate volumes). Successful technology follows only one or two
well-worn paths, not many. Technologyprogressesrapidly when everyonegoes down
the same paths and develops all the understandingrequired to make the process work.
By its nature, a start-upmust strike out in a direction other than that in which larger
or existing companies are going. The trick is to distinguish between potentially
productive directions and foolhardy ones. The greatest temptation is to use a new
semiconductor component even though, at present, it doesn’t quite exist. At the
beginning stage of any new technology,a number of false starts will be made, and only
a few of the paths taken will lead toward success.

ATTEMPTING A PROCESS
THAT REQUIRES SIGNIFICANT BREAKTHROUGHS

A start-up may be predicated on a new process requiring significantbreakthroughsin

manufacturing and yet not be staffed with a leader or the critical process-engineering
design slullsneeded to achievethosebreakthroughs.During the past twenty-fiveyears,
the humble printed circuit board has made new computer classes possible, given birth
to new companies, and caused greatBrief to others.How these eventshave cometo pass
provides an excellent illustration of the importance of selecting the right process
technology.
Printed circuit board technology is measured by yield (hence, cost), size, and
interconnection density, the latter being a combination of line width and number of
layers. Organizations with large and bureaucratic manufacturing and field-service
organizations,SuchasDigitalEquipmentCorporation(DEC)and IBM, have traditionally
taken a very conservativeapproach and favored small printed circuit boards. Manu-
facturing wants tiny boards in order to get perfect yields and to make testers small,
cheap, and simple. Field service wants small boards in order to have compact and
economicalunits that will facilitatefield replacement. In contrast to manufacturingand
field service,system designerswant very largeboards in order to be ableto get the entire
system (e.g., the PC), or at least one major component (e.g., a processor), on a board.
High-tech ventures tend to be founded by system designers, not manufacturing and
field-service people, so start-ups have pushed the size limits upward to build new
computers and new computer classes.
In 1968, examples of the small-board approach included DEC, which was using
small boards suitable for packaging circuits with discrete components and automated
assembly, a technology borrowed from IBM. When ex-DEC personnel founded Data
General and built that company’sfirst mini(theNOVA), they packaged one major com-
148 Manufacturing

ponent (centralprocessingunit, memory, and input /output) per board, endingup with
a three-board computer. DEC increased its board size markedly in the next few years.
Another manufacturer using small boards was Computer Controls Corporation,
which built the first 16-bitintegrated circuit minicomputerusing very smallboards that
had only one or two integrated circuits per board. Had it packaged the computer on
larger boards, resulting in lower cost, it might have survived without the Honeywell
merger, makmg it a competitor today.
In the 1990s, it is very difficult to build a large or cost-effective multiprocessor
system using small boards because the shape of an ideal system is a cube, or simply a
singleprinted circuitboard as in the caseof a PC. By using smallboards,one cannotbuild
a very large cube (or a very large computer).
So far, this discussion has made a case for "large is better," but it is possible to try
manufacturingprinted circuitboards that are too large. An example was Elexsi, which,
in 1982, built a large superminias a multiprocessor, using boards that were beyond the
limits of the standard manufacturing process (photolithography,plating tanks, solder
machines,component inserters, and testers).Elexsi was driven by its engineersto build
very large boards so that it could get its ECL processor on one board. Pressuring the
manufacturingorganizationto do somethng completely contrary to the infrastructure
cost the firm at least a year and a half in entering the market-and probably its life.

HIRING PERSONNEL WITH A BACKGROUND IN MANUFACTURING

AT LARGE COMPANIES

The new venture may be tempted to hire as its manufacturing head an individualwho
has had responsibilityfor settingup and operating a manufacturingplant within a large
organization. This is highly risky, since the large-company person will probably be
unable to function without a big staff. The skills required to succeed in a large
corporation,such as negotiation or managing a big staff, are not especiallyuseful for a
start-up, which, by its nature, is small and focused.

MANUFACTURING RULES

Unless the company is breaking new ground in manufacturing processes, as would be

the case with disks or semiconductors,manufacturingis not stressed as a criticaldimen-
sion.In fact,by thecompletionof theseed stage, the head of manufacturingmay not even
have been hired. In the case of software, manufacturing is almost trivial. Nevertheless,
the start-up should keep the following rules in mind in order to avoid rude surprises.

Does the company have a well-defined organization and processes that will
enable it to produce products at the cost, quality, and schedules required by its
customers?
 ManufacturingRules 149

This is the basic test for whether the firm is disciplined and will be able to survive.
The organizationthat fails to satisfy any of these fundamentalrequirementsis doomed.

Does the company have initial ideas and an outline for a manufacturing
strategy, including the degree of integration (i.e., which components or assem-
blies it will buy and which it will build), plant location, critical processes,
specialized components, and quality control?

Thisrule testswhether the organizationhas givenany real thought tomanufacturing

its product. In the case of software, manufacturing is Straightforward and is usually
done externally;and in the case of systems that require no new processes, the plan may
be quite simple. If the start-up is building a computer component, such as a disk or
semiconductor,in contrast, then manufacturing is its principal reason for existing,and
thus, the firm must have a detailed outline for an extensive plan. For other systems
requiring high volume, such as a terminal or PC, an extensive plan is needed to
demonstrate that the organization can meet cost and quality requirements. This plan
may involve a partnership for offshore or automated manufacturing.

If a strategic manufacturing partner is required, have candidates and contacts

been identified?

This rule tests whether the company is approaching its manufacturing needs by
looking for a partner to share in financing the manufacturing operation or by building
the required expertise from the outset.

Will contractand/or offshoremanufacturing capabilitybe required in order for

the start-up to produce effectively in terms of quality and cost?

In 1990, finding an offshore partner is a straightforward process because all

governments (exceptthat of the United States)understand that manufacturing is vital
to their economy. Manufacturing high-definition computing television and pocket/
wallet computers calls for collaboration with offshore manufacturers because of the
capital and skilled labor needed to satisfy volume, cost, and quality demands. For
complex componentsrequiring skilled assembly (semiconductorsand disks),manufac-
turing has moved to the Pacific Rim, where a higher level of skills is available from a
better-trained work force. A study by InternationalData Corporation (IDC)showsthat
between 1982 and 1989, the manufacturing of nearly all hard disks, floppy disks, and
tape drives moved out of the United States.

By the end of the seed stage, does the start-up have a plan in place (complete
with costs) that identifies critical processes, suppliers, and an approach to
running the manufacturing operation?
150 Manufacturing

By the end of the seed stage, the company should have a good idea about potential
suppliers of parts and processes, including special devices such as test equipment. A
new venture that starts up without even a rudimentary manufacturing plan is quite
likely to require additional funding once it faces equipment ”sticker shock.”

If the company intends to do its own manufacturing, have the plant size and
factory location been figured into the plan?

The manufacturing plan is more than a spreadsheet exercise that relates space,
people, and product output. It must include an initial attempt to define the plant design
in order that the requirements for space and people, including those with special skills,
may be understood.Unlike many of the other resources,acquiringmanufacturingcapa-
city calls for a great deal of careful advance planning. If the start-up is predicated on a
novel manufacturingprocess or will need highly trained individualswho can evolvethe
process, the plan must take into account the location and availability of a work force.

If achieving the planned unit cost and schedule goals is predicated on essential
breakthroughs in the manufacturing process, are the necessary resources
(manufacturingvice president, specialists,time, and money) available?

If the firm’s product technology is embodied in its manufacturing process, as

opposed to its product design, then the manufacturing process must be treated as an
engineering design and managed and measured as such. High-tech ventures are often
predicated on the development of new processes for the manufacture of disks, tapes,
semiconductors, printers, and various display devices. In these cases, starting up
without a seasoned vice president of manufacturing is a flawed approach.
One firm based its business plan on having a highly automated plant. Although no
fundamentally new processes were required to build the plant itself, a total system did
have to be developed to ensure proper coordination of all the process steps in order to
produce the product.

Is the product design planning process predicated on producing a design that

ensures manufacturabilityand the highest quality?

This rule tests whether manufacturabilityand quality have been designed into the
product from the outset. Manufacturability is not always regarded as a criticalaspect of
product design. More typically, the product is ”thrown over the wall to be built” after
the design is done because its manufacturing is thought to require simple and well-
proven processes.However, such an approach is unlikely to yield the lowest cost or the
highest quality. Unless the firmplans to produce a manufacturing-intensiveproduct, it
will probably not have a manufacturing person on board at start-up. The best way to
ensure both manufacturability and quality is to hire people who have manufactured
high-quality products before.
 Conclusion 151

Does the company manage its raw materials and finished goods inventoriesin
an optimal fashion?

As I observed at the startof the chapter,managingraw materials and finished goods

is a balancing act.The start-upmust have enough of each on hand to enableit to respond
flexibly to fluctuationsin the order rate but not so much that it feels a financial impact
from having excess inventory.

Does the company introduce products into manufacturing rapidly, accompa-

nied by clear product and process specifications?

Thebestguaranteeof a speedytime-to-marketis for engineeringand manufacturing

to function well together from the beginning. Although engineers must understand
exactly how their product is built, manufacturingshould be responsible for building all
the products, including the prototypes.

CONCLUSION

In the case of manufacturing (as with every other dimension of an organization),

achieving the highest qualityis the most importantoperatingprinciple. Without quality
in every part of the operation,costs will be high, predictabilitywill be nil,and customers
will be unhappy.
The health of a start-up’s manufacturing dimension will be assured if it follows
these eight guiding principles suggested by Matt Sanders:

1. Only build products of the highest possible quality.

2. Only invest in manufacturing if the process is unique and is the essence of the
start-up.

3. Examine every make/buy decision.

4. Get to market fast.

5. Use minimal cash.

6. Have only a minimal staff, but hire the critical people.

7. Make sure product cost is predictable and low.

8. Avoid the evil of inventory.

Chapter 8

THE PRODUCT

People, product, plan.

-Venture capital adage

The venture capital adage ”people, product, plan” clearly emphasizes the product as
one of the crucial elementsof a start-up.The product, together with various servicesthe
company sells, is the organization’soutput; it is what customersbuy and use. Customer
purchase orders convert the product to revenue, and the quantity of those orders
determines whether the firm is viable.
During the concept stage, the product is representedby a few sketchesand perhaps
a prototype demonstration in a laboratory. After that stage, the product progresses
through various design phases as a series of specificationsand demonstrations until it
can be realized and replicated through a manufacturingprocess,as described in chapter
7. As the companyenters the market development stage,the actualproduct is produced
by the manufacturing organization and shipped to customers for revenue.
A product can be viewed in three ways:

The product specification: Instructions or information (the bit pattern in the case of
software)developedby engineeringto describethe product sothat a manufacturing
organization can replicate it.

The product itself, or reality: The physical product coming from a production line or
replicated from a master software tape; this is what customers ultimately buy.

A market, or the buyer’s image: Specifications,pictures, and brochures that describe

what a product is or how it appears, explain why someone would want it, or cause
prospective customers to believe it can do something significant for them.

152
 Understanding Why Customers Buy 153

In this chapter,we will look at the product in many different ways, starting with the
buying rationale. Products are placed within a product space, ranging from sand
(silicon)and iron (magnetics)to organization-specificuse, with training, learning, and
service.A historicalview of how the computer evolved into classes and gavebirth to the
associated applications softwareprovides a background for the product development
cycle. Important product-design issues, such as evolvability, and common flaws in
developingproductsare also explored to show readers what to watch out for in product
design and product positioning.

UNDERSTANDING WHY CUSTOMERS BUY

Many factors affect whether a customer will buy a product from a particular manufac-
turer. In the case of an established product class, the most obvious rationale is the
product’s relationship to other products in its class in terms of performance and price.
New products, in contrast, may be purchased on a sole-sourcebasis. As the industry
matures and a commodity, high-technologymarket forms,then more factors, including
appearanceand prestige,become important.No matter what product or servicethe new
venture intends to supply, its staff must understand why the customer will buy that
product. The following subsections examine the most significant determinants of the
customer’s purchasing decision.

PERFORMANCE AND PRICE

In the early days of computing, computer pioneer Herb Grosch posited the following
relationshipbetweenperformance and price for computersintroducedat the sametime:

= k X price2
performanceklme

This relationship argues that an economy of scale exists, i.e., for twice the price you get
four times the performance. With the introduction of new classes of computers,
however, it has been shown (Bell, Mudge, and McNamara,l978;Mendelson, 1987)that
this relationship is flawed, and if it ever was valid, it holds no longer. Today, a strong
diseconomy of scale exists, such that:

=k X priceo8

We can observethis phenomenonby looking at various machine classes in the price

versus performance plane of the sketch’ in Figure 8-1, in which different classes of
1,This sketch was used to ”position”Titan, Ardent’sfirst graphics supercomputer against other potential
competitive computers.
154 The Product

scientific and engineering computers are characterized, from supercomputers and

mainframesto personalsupercomputers.The diagonallines are constantperformance/
price lines. If there is no economy of scale, all computers built at a given time should lie
on the same line. If there is economy of scale, hgher-priced computers should lie above
the h e . If there is diseconomy of scale, lower-priced computers should lie above the
line. Sincethere is no economyof scalefor any of the constituentparts of a computer,the
lower-priced machines offer the best performance for the price.
To understand why someone will buy a particular system or software package
requiresa deep understandingof the product applicationand the buyer. Drucker (1985)
devotes much of his book Innovation nnd Entrepreneurship to understanding the buying
decision in an abstract fashion. The two most obvious purchasing rationales include:

Better performance for the same price

Better performance for the price at another price level

100

0.1
1K 1OK lOOK 1M 10M

Figure 8-1. Sketch used in 1986 to "Position"Titan, Ardent'sFirst Graphics

Supercomputer, in the Performance Versus Price Space.
 UnderstandingWhy Customers Buy 155

In the hgh-performance computing world, where performance is one of the main

measures of comparison, having a system five to ten times faster than any other
competing product simplifies decision making for the buyer. However, performance
will fail to be a sufficiently convincing selling point if the user cannot program the
computer or cannot obtain critical software to make the product useful. Furthermore,
in the future, performance is unlikely to be enough to differentiatea product for very
long,sincecomputersevolveso rapidly. Establishedvendorswill evolvetheir hardware
quickly in order to hold their customers. If product evolution still isn’t fast enough to
hold customers, vendors may even sell futures-promises of machines that aren’t quite
ready yet.
The second buying rationale listed above, better performance for the price at
another price level, causes inherently new usage and creates new markets, such as the
computer in the home. Because performance and price are relatively easy to measure,
they are employed in this book as simple, straightforward criteria for segmenting
computer classes and uses.

ECONOMICS, PERSONAL POWER, AND APPEARANCE

Ultimately (in the five- to ten-year time frame),buying decisions are based on econom-
ics, includingpersonaleconomics.Decisioncriteria alsoinclude such unusual attributes
as the feelingof personal power associatedwith having the largest and most prestigious
machine. Even appearance can be used to segment a computer market. Some early
buyers were attracted to NeXT’s first black-and-white workstations because of their
attractive packaging, even though they lacked software and offered only incremental
gains in functionality over the Macintosh or existing workstations. After the first few
buyers, NeXT had to compete on the basis of true functionality and applications.

NEW CAPABILITIES

Yet another rationale for purchasing is that the product offers the user completely new
capabilities.The potentialmarket for new productsis often impossibleto predict, except
by producing the product and building the market for it. For example,as ”multimedia”
becomes available, it‘s difficult to understand exactly who will use it or to make a
compelling argument for why such capability is needed. However, many who have
seen these programs say that comparing them to current products is like comparing
color TV to black-and-whiteTV or comparing a Macintosh to an IBM PC running MS/
DO’S without Windows. ”Multimedia” products reaching the market in 1990 (such as
156 The Product

MACROMIND)are great, and almost everyonewould prefer a dynamic,color presen-

tation to dull, black-and-white overheads. On the other hand, SteveJobshas described
multimedia as ”the AI [artificialintelligence] of the 1990s”-i.e., a great promise that
may not be fulfilled.The market is probably limited by a lack of low-cost playback units.
Readers are invited to start a company to build these, if they think they can beat Sony
and other consumer electronics firms to the market and then survive there.
Some of the factors influencing computer-equipment purchase decisions are
shown in Figure 8-2. In order to determine why someone would buy a hardware
componentor system (ora new softwarepackage or system),weights must be assigned
to these factors for each of the potential customers.Although the figure has a decided
hardware and manufacturing orientation and flavor, it applies equally to software.
The final test of the product is whether it can meet the cost and quality goals set by
the company and (mostimportant)by the customer.Many of the criteria listed in Figure
8-2 can be quantified,either as operationalor as one-timecosts, permitting users to truly
”value” their purchase. Although quality is often synonymous with performance,
reliability, and ease of use, quality cannot be measured as quantitatively as cost. For
example, quality can run the gamut from a poorly performingprogram that ”feels”bad
toa program that hasalookand feel that exceedstheuser’sexpectations.Understanding
and measuring the product is the task of everyone in the firm, but marketing (the
product-management organization)is responsible for the critical accounting function
that measures the company’s competitiveproduct position.
When it comes to product planning, the start-upshould keep in mind the following
three important observations about the behavior and motivations of buyers in the
computer market:

If cost is a significantfactor in buying a computer, such as when the individualwill

,be paying for it, he or she should: (1) wait as long as possible to buy, because
computers evolve rapidly, and (2) always buy the lowest-priced machine that can
perform the task.

If the requirementsfor the computerare unknown,as they are witha centralservice,

or if the buying institution is not the paying institution, as happens with a service
center that is paid forby the governmentor when chargesare billed to someoneelse,
then the customer should buy the largest, most general, and most expensive
computer that can perform the greatest number of tasks.

Readers should always remember that performance and price are not necessarily
the main determinantsof a product’s success. Products must be differentiatedby
additional characteristics, including the way the buying organization, such as a
company, functions (e.g.,conducts its business).
 Market Considerations for a New Product 157

Purchase price, cost of ownership, return on the investment, and

apparent lifetime based on the rate of technological change (obsoles-
cence)

Peak performance or response time, and work throughput

Unique features or functions that differentiate the product from

competitive and potentially competitive products

Availability of the appropriate applications software or other parts of

the infrastructure needed to carry out the buyer’s mission

Adherence to former and future standards, past compatibility (knowl-

edge of how to use the product), future compatibility (or growth path),
and cost of converting any data or software from a current standard
or system

Level of comfort with the vendor and individual sales/service person,

including support; ease of purchase, installation, and use; machine
appearance; brand prestige (e.g., %ray aura,” “IBM feel,” or “Macin-
tosh cult”); and ability to associate with other users

Need for specialized programming, knowledge, and training

Personal control over the allocation and management of the buyer’s

resources

Figure 8-2. Buying Criteria for Computer Systems (Including Software).

MARKET CONSIDERATIONS FOR A NEW PRODUCT

A key part of understanding the buying rationale for a product is to precisely under-
stand the overaII market considerations, including the dynamics of introducing a new
or existing product into a new or existing market and the issue of nichemanship.

NEW OR EXISTING PRODUCT IN NEW OR EXISTING MARKET

Table 8-1 shows four quadrants (existingproduct into existingmarket, existingproduct

into new market, etc.). Each quadrant presents a unique set of problems and opportu-
nities and raises some fundamental questions that must be answered before the
start-up attempts to introduce a product into that product/market quadrant.
158 The Product

Table 8-1. Example of Market/Product Opportunitiesand Key Problems.

~-
Product Market
Existing New Customers and/or Use
Existing High-tech commodity Distribution or application
(e.g., disk) pioneering (e.g.,mail order or
home PC)
Can you achieve the Can you establish
level of cost and quality infrastructure?
required to beat the
established suppliers ?
New Substitution (e.g., work- Pioneering to an emerging
stations replacing minis or market (e.g.,first spreadsheet,
mainframes),applications voice or handwritten
software (e.g.,derived from control of PC)
users)
Can you find the customers? Are you first and right?

Introducing a new product into a new market (the lower-right quadrant of Table
8-1)is a very difficult task, since the start-up essentially has to ”make the market.” The
difficulty of making a market is usually proportional to the distance between the new
productand other known productsand markets,because the more unfamiliar a product
is,the harderit istoestablishamarketforit.Attheotherendof thespectrum,introducing
a new version of an established product (the upper-left quadrant) poses a different set
of problems. Established products (such as disks) are high-tech commoditiesthat are
usually differentiated by cost and quality. If a company starts up to enter a well-
established market with a well-established product, it is fundamentally betting that it
has a unique approach to the product design or a special method of manufacture or a
lock-in feature.
In the case of software, where all products cost virtually nothing to produce, the
organizationis betting that it has a unique way to distributeits product so as to address
a fundamentallynew set of users, thereby creatinga new market.If patent and copyright
laws continue to support ”look and feel,” then software products will continue to be
high-priced. However, if it becomes legal to clone sofhvare so that it looks exactly the
same to a user and carries out operations on exactly the same data, then softwareprices
will fall to near zero, since softwarecloning is almost always possible, given enough
time.
Introducinga new product into an existingmarket (thelower-leftquadrant of Table
8-1) involves the process of substitution-getting buyers to switch from the product
 Market Considerations for a New Product 159

they are currently using to the new product. Chapter 12, "Technical Workstations,"
includesa descriptionof how Apollo successfullyreplaced time-shared minicomputers
by introducing a new product into an existing market. Building on this initial impetus,
Apollo enjoyed a second, significantlylarger gain with fulfillment of new applications
that onlybecamepossible throughbetteruser interactionandlarger,mdtiple-windowed
screens. Among the applications were software engineering, office automation, and
electrical and mechanical computer-aided design (CAD). Stellar was subsequently
formedtodoitagaininasimilarway,only thistimebybuildingagraphicssupercomputer
to replace fast workstations, including those connected to a shared supercomputer.
The real key to introducing a new product is to develop one whose obvious
superiority to existing products justifies a high margin. The following rule applies:

A new venturemust maintain high operatingmargins to fuel its growth.In order to justify
thehigher margins,itsproductmustbesignificantlybetterthanproductsfromestablished
suppliers. Predicating a start-up on a product that is only slightly better than existing
products is an approach doomed from the start!

Although the phrase "significantly better" usually implies a clear advantage in

performance,price, or quality,another attribute,uniqueness (discussedbelow),can also
serve to differentiate a product and thereby become the basis for profitable sales.

MARKET NICHEMANSHIP:UNIQUENESS
AS A MEANS OF JUSTIFYINGHIGHER MARGINS

Nichemanship is the art of introducing a new product or service that will serve a well-
defined segment (a "niche") of an existing,larger market that is being poorly served by
currentsuppliers.Niches are created by developinga product with unique featuresthat
appeal to a select segment of buyers. Niche products are distinct from newly invented
productsthat create entirelynew markets,such as a new computer price class or the first
spreadsheet. Successful niches include:

The minisupercomputers from Alliant and Convex, created to meet the needs of
scientificand engineeringusers who were working with superminicomputersand
who required power for computation

Military computers or components, such as those available from Performance

Semiconductors

Fault-tolerant transaction-processing computers from Stratus Computer that at-

tacked a small segment of Tandem Computer's market
160 The Product

High-performanceand high-feature word processors targeted toward the docu-

ment-preparationandtypesettinguserswho formed thebasis of desktoppublishing

More powerful spreadsheetsable to display more complex data for technicalusers

Nichemanship is one way that a small company can play on the same field as the
giants of the computer industry without getting trampled to death.Jeff Tarter, editor of
Soft *letter (1989),offered the followingadvicetoa firm that had just entered thedesktop-
publishing market with no way to differentiate itself

Be a Goliath; Davids rarely win. In the software business small market shares are rarely
profitable.One or two companiesget 80-90% of profit dollars. Davids get fringe customers
that the large companies cannot serve through normal sales channels or make extravagant
demands for hand holding, advanced features and pricing concessions. Some alterna-
tives: 1. become a mini-Goliath in a niche e.g. home, academe or 2. get an alternative
channel through a mass merchandiser or a private label e.g. Tandy.

VIEWING THE NEW PRODUCT AS PART OF THE BIG PICTURE

A new product in the computer field-be it hardware, software, or a combination

thereof-should be examined from a number of different perspectives as objectivelyas
possible. Although this is difficult for the product’s inventors or investors to do, the
rewards of such a multifacetedexaminationare great,because it can reveal the product’s
weak points, if any, and predict the likelihood of its market success. In the following
subsections,readersare asked to contemplatethe functionaluse of computers,consider
where a new product falls within the ”computer product space,” understand the basic
nature of computer classes, decide whether the product creates a new computer class
or application,and view the new product from a historical perspective.

COMPUTER FUNCTIONS

Figure 8-3 shows a taxonomy of computer use within various organizations.Note that
the need for, and resulting economics of, computer use vary considerably from one
organization to another. To a great extent, the functional use determines the configu-
rations, software, and performance/price requirements.

COMPUTER PRODUCT SPACE

A new computer product will be part of a system that is the sum of all its hardware and
software components, including the highest-level programs required to perform a
 Viewing the New Product as Part of the Big Picture 161

Commercial organizations
Financial accounting and control, with record storage and batch processingfor the firm
Billing, inventory, accounts receivable/payable, payroll
Transaction processing for sales and intrafirm/interfirm transactions
Business analysis

--
Technical organizations(science and engineering)
Numbers, algorithms, text, graphs, storage, and processing
Data acquisition and real-time experimentation
Interactive problem solving using computer simulation rather than experiments to
model for science, engineering, and product and process design, including computer-
aided manufacturing (CAM)
Communication, databases (notebooks)

Manufacturing
Record storage and batch processing
Continuous and discrete real-time control
Plant scheduling and process optimization

Communication
Message switching and organizationwide electronic mail
Computer networking, including all local area networks (LANs)
Voice and speech, teleconferencing

Office automation, electronic (desktop)publishing, and word processing

Image processing for the transduction, storage, and transmission of documents

Education
Reading, writing, communication
Mathematics
Computer-assisted instruction via simulation models
Database and network access

Home
Entertainment (e.g., games), instruction (including simulation), database, network
access

Figure 8-3. Taxonomy of Computer Use Within Several Major Organizational

Categories.

given application.The system and its component parts can be characterized in terms of
the three dimensions of the computer product space shown in Figure 8-4:

Class: The class dimension is characterized by price (and the dependent variable,
performance).
162 The Product
Level of integration

Person-spdIlc
applications
1
Systems Integrators
Organization-sprlk
applications

Proresslonal PKlCerS
appllcatlons control P m f d o n a l applkatlons
(e& accounting,
CAD) I Bask sdta.reappUcatlons

Generic
applications
(e.&, word
pmesslng)

Lnngwes,
operating
systems, etc.
Robotics and
manufacturing
automation
=- t Computer system
Database

Communlcatlon
Hardware/mltw.n
and common
carrkrs
Hardware
systems

PiVCegFOrs,
CRn, disks,
computer
components
Tl- Peripheral equlpment
1
Chips, dlsk
substrates,
etc. Materials and process qnlpmrnt

Sand, Iron
ore, materials

Information-Roedng Function

Robots Communlcatlon Communlcatlon Processlng Memories S w i t c h and

$1' (amweyes, w~th wlth links =
legs, etc.) mrehanlsms humans eommunkations

Pocket

Slo0,lmO

supermlni

Mainframe
$10 mllllon
Supercomputer
Computer prlce and claw

Figure 8-4. The Information-ProcessingProduct Space Formed by the Level-of-

Integration, Function, and Price/Class Dimensions.

Function: The function dimension involves processing or control, memory (e.g.,

database),switching (communications),transduction (interfacingto humans),and
other information processes (such as robots).
 Viewing the New Product as Part of the Big Picture 163

Level of integration: In the level-of-integration dimension, each level carries out a

particular function, building on a lower level and producing a function for use by
the next higher level.

Figure 8-4 shows a plane within the three-dimensional space and characterizes
some of the products within the computer and communications industries. The
computer product space can be used in various ways to plot historical trends; a
component product (point or line), such as a disk or spreadsheet within a given level of
integration;a system product or product line covering a volume of price and function
in the space and based on a set of levels of integration; all the products of a given firm
or industry segment; and trajectories of products and companies.
Figure 8-5 shows the standards that define the interface at each level of integration
for a particular system in the workstation computer class, the Stardent 3000 graphics
supercomputer.Each computer class has a particular set of standards that define the
interface at each level of integration.

Application Chemistry (e+, Gaussian), FEM (e.&, Anysys, Nastran), CFD, AVS
imaging, and general technical applications (visualization)

Libraryhtility NAG, Linpack, Matlab, Mathmatica, etc.

Fortran with
Programming DCLEDT DEC and C Looking
language and Cray ext. Glass
user interface
Debugger
Perf. analyze

Operating U n k File System NFS Motif

system Ingress, Unify x11
PEX
MIPS AB1
TCPIIP DECnet

Hardware I I I I I I
r
.-~ ,~ __
, l l r-
p e- . MIPS CPUI Mass. Mass. Networking Comm. Graphics arch.
IEEE F.P. store store
with vector 5 1/4" disk 8mm tape
ext. 1/2" tape

Ethernet FDDI RS232 RGB, PAL,

Interconnect Titan bus SCSI
RS170
VME HPPI Centronics
SMD
DR 11W

Chip
MIPS micro. Standard chips and peripherals Custom graphics
Custom Rendering chips
Vector prw.

Figure 8-5. Standards Used for the Stardent 3000 Graphics Supercomputer at Each
Level of Integration.
164 The Product

COMPUTER CLASSES

Computers can be classified in a number of different ways. The following subsections

show how they can be classified in terms of evolutionary stage, price, both price and
weight, and according to several other criteria.

Classes by Evolution

Figure 8-6 shows the evolution of computing styles, applications (whethercommercial

or technical),and operating systems.

Classes by Price

A simple view of the computer classes is given in Figure 8-7, which lists what buyers
believe to be almost twenty distinct kinds of computers, ranging in price from $10 toys
to $20 million supercomputers. Given the plethora of computers, together with the
consolidation talung place through corporate acquisitionsand mergers, one might be
skeptical of a start-up aimed at building yet another general-purpose computer.

note: m w s denote I change from s u t e x to state y

Figure 8-6. Evolution of Computing Styles with Time.

 c
- 0

z.-
Y

F 5
F
0
0
P
0
0
0
F
0
0
0
0
-
3
-.
5.
I

5' $
0 0 0 0 a 3
0 (D

I I I II I I II I I II I I II I I II 1 I I1 I
h
w
v

Hand-heldgame - 1 I I I I I I

-
3
+5 Video game I II I I I I
5

-
I -1 I I I I I
D Home PC + ---+--- + - - - + - - - + - - - + - - - + - - -
8 Wallet
I I I I I
.a-.
v)
I I
Pocket I I I I I f
ka! Hand-held programmable I
-_-_---- I I
-I --- -I --- -I ------- I i
5-
5' Notebook
I I - i - I I I I 09

a
i--L I I 1- I I I I
a Laptop
P t--- t--- t----t-- -t--- t--- t---
Desktop PC
I I I I I I I
Professional workstation I I I 1 1 I I

6
166 The Product

Classes by Price and Weight

Dave Nelson (an Apollo founder)and I posited an interestingmodel for characterizing

computer classes, which is given in Table 8-2. Note that the relationshipbetween price
and weight is:

price = 10 X lW'ms' weight = 0.05 X 10class#

Other products and services can also be characterized by classes:

Prices of cars = $6,000 X 1.5c'ass#

Classes by Application

Class is determined quite subjectivelyby the buyers who regard one or more products
as equivalent.In a similar fashion, computing is a commodityfor which other products
and servicescan be substituted.For example,users could employ a serviceto get a result,
or they may lease or purchase any number of computers from different classes-such
as a collection of PCs, a shared microcomputer, or a variety of minicomputers-to
produce the same result. Figure 8-8 plots the size of the computer market, in terms of
computer size (i.e., class) and application,at several points between 1960 and 1987.

Computer-Class Consolidation

After examining the preceding charts, readers might conclude that the number of
computers and computing styles within a class always grows with time, but this is not
necessarily the case. It is quite possible for a computer class to be consolidated when a
companymakes a singlearchitectureavailablefor a largemarket range.That firmis thus
able to dominatea singlemarket and become the leader.As a result, all other firmsmust
then become compatiblewith this form of computing.The followingare someexamples
of class dominations:

Mamframes: IBM 360 (circa 1964)for centralized computing.

Minis: DEC VAX (circa 1978)for a wide range of computing styles and the IBM
AS/400 series, which evolved from earlier business minis.
 ‘Viewingthe New Product as Part of the Big Picture 167

Table 8-2. Nelson-Bell Computer Classes.

Class Weight
No. Where Used Class Price ($) a (Lbs.)
0 Wallet Calculator, personal data cardb 10 0.05
1 Pocket/palm Calculator, personal database 100 0.50

2 Briefcase Notebook and laptop portable 1,000 5.00

3 Office PC and workstation 10,000 50.00

Persalnal supercomputer 50,000 150.00
4 Project, group Micro, graphics super, mini 100,000 500.00
Department minisuper, supermini 500,000 1,500.00

5 Center Mainframe 1 million 5,000.00

6 Center, region Supercomputer 10 million 50,000.00

aAsof 1990, the price of each class may extend upward to encompass the price of the next class and
downward by a factor of 2 or 3.
b”Smartcard” with on-board micro storing thousands of characters to a million characters.

PCs: IBM-compatible PCs controlled by Intel’s 80x86 architecture and Microsoft’s

DOS operating system.

Workstations: Sun Microsystems, UNIX, or a n evolution of MS/DOS and OS/2.By

7 995, it should be possible to determine whether workstations will remain a distinct
computer class or merge ,and become competitive with PCs.

Some Closing Thoughts on Computer Classes

Only computer classes that are available on a ubiquitous basis from competitive sources
will survive and thrive. All proprietary systems, though not competitive, will continue
to be available to serve a declining installed base at premium prices in 2001. These
suppliers will not be significant.
As noted above, the rang,e of computer classes extends from the few-dollar card
computer to the $20 million supercomputer. It is unlikely that a company can be formed
to exploit the over-$50-million computer market.
168 The Product

Figure 8-8. Market Size for Various Computer Classes and Applications with Time.
(Courtesy of the IEEE Scientific Supercomputer Subcommittee, from "The
Computer Spectrum: A Perspective on the Evolution of Computing.")
 Viewing the New Product as Part of the Big Picture 169

There are ample opportunitiesfor new computer products that do not fit into the
computer-class taxonomy given above.For example, there exists a potential market for
a variety of stationary and mobilie robots to move things and carry out tasks remotely
(e.g.,surveillance).

CREATING NEW COMPUTER CLASSES AND APPLICATIONS

As semiconductor and magnetic density evolve to offer greater performance and

greater functionality,as discussedin Chapter 5, there are severalapproachesthat a high-
tech venture can use and several forms that new products can take:

1. Creation ofa new class: The start-upcan use the increase in density made possible
by the new technology to build a less costly version of the previous generation.

This strategy creates a new, lower price class. A new company will be required
to exploit this new product form. Once the class forms, new firms will also be
required to build appropriate software.

2. Evolution: The start-up can use the increase in density to build a system with
increased performance having roughly the same price as the previous genera-
tion.

The result is the next, evolutionary model of a computer in the same price class,
but with the power of a previous generation’shigher-classmodel. For example,
every three years, such a system provides four times the memory, requiring two
more bits of addressing.

3. hention: Thestart-upcaiipushthenewtechnologytothelimit inanunorthodox

fashion to gain performance, although at an increased cost.

This strategy allows a given manufacturer to enter the next higher price class. It
is also the strategy required when pushing technology to build the next
supercomputer.As noted above, however, it is unlikely that a company can be
formed to exploit the over-$50-millioncomputer market.

Figure 8-9 shows how new classes form and how old classes are reimplemented
with new technology as a function of time. A company that has a fixed organization,
existing user base, and established cost structure always tends to adopt the second
strategy in the preceding list, evolution. This satisfies the firm’s current users and the
marketing and sales departments, as well as those engineers who feel safer with
evolutionthan with settingout on a new path. A new organizationis required if the first
or third strategies are to be adopted.
170 The Product

\I\ evolutionary
paths

Constant

T (t - 1) T (t) T (t + 1)
Technology time
Figure 8-9. Computer Price Versus Time for Each of the Computer Classes.

The process of evolution in software is almost identical to the process of evolution

in hardware classes, whereby a new release or version of a program can:

Become smaller and simpler,thereby addressinga new market that doesn't require
the advanced features of the old product

Increasefunctionalityby using a more powerful platform that has more processing

and memory with which to operate

Be recast to employ another computing style (e.g., graphics) so as to permit a new

type of use and address a larger base of users
 Viewing the New Product as Part of the Big Picture 171

VIEWING NEW PRODUCTS FROM A HISTORICAL PERSPECTIVE

The brief hstory of the computer industry presented in t h s subsection will enable
readers to appreciatethe industiry’sdynamicsand gain insightinto the opportunitiesfor
future products. Two observationsare critical

All new forms of computing (from supercomputers to pocket calculators) and

constituent software (from computer-aided design of molecules to databases and
human user-interface paradigms) require new companies to introduce the new
computer class or invent the new way of computing.

IBM has dominated the entire history of computing, beginning in 1950. It is

incredibly robust and dynamic and has been able to retain its leadership during the
whole period.While rarely being the first to introduce a new product or leading in
a given area (except disk memories), IBM ultimately leads in market share.

Although the following overview characterizes computer generations in terms of

hardware technologies, it should be realized that with changes in hardware (for
example, the Macintosh),a new industrialstructure forms around the new computing
paradigm. Just as the old-line hlardware suppliers are irrelevant to advancing modem
computing, the traditional suppliers of software for mainframes are irrelevant to the
new distributed and interactive workstation computing environments.Several orders
of magnitude more softwarecompanies form to exploit each new computer class than
there are firms actually producing computers within the class. Similarly,for each new
computerclass, the number of componentsuppliersthat form to support the computers
in that class is at least equal to the number of computer-producingfirms within the class.
For example, twenty-nine companies started up during the period 1977-1987 to build
Winchester-technology disk drives, with a combined investment from private and
public sources of nearly $1billion.
In the beginning, when Eckertand Mauchly established UNIVACand built the first
commercialcomputer, they designed and built every part of the computer (exceptfor
the vacuum tubes),including its power supplies, logic and memories, and tapes. They
also wrote all the software.Today,a high-tech venturecan simplyassemblea computer,
complete with software, from component suppliers.Table 8-3 shows how the structure
of the industry, including the use of manufactured components from other industries
(e.g., semiconductors, disks, cathode-raytubes [CRTs], and all software components),
has evolved.
Table 8-4 shows the top circuit component suppliersranked by revenue at various
times. It is interesting to note that the suppliers of circuits rarely make many of the
technology transitionsfrom geineration to generation.On the other hand, old computer
172 The Product

Table 8-3. Sources of Computer Components in Each Generation.

Generation
Component 1 s t b 2d 3d 4th 5th
~~

Technology Vacuum tubes, SSI, MSP LSI, VLSI" ULSI"

transistors
Power supply mfg. rnfg., cs mfg., cs, std. cs

Logic mfg. semico. semico. std.,

semico. std.,
mfg.-custom

Memory mfg. mfg., cs, semico., std. semico., std.

semico.

Packaging mfg. mfg. mfg., std., cs mfg., std., cs

Mass storage mfg. cs cs, std. cs

Terminal mfg. cs, std. cs cs

Communications/
LAN mfg. semico. semico. std. semico. std

Operating system mfg. mfg. mfg., AT&T, AT&T, OSF,

Microsoft Microsoft, IBM

Databases - mfg. cs, std. cs, std.

Languages mfg. cs cs. std. cs

Generic
applications mfg. mfg. 3d, std. cs

Professional
applications user 3d 3d cs

Abbreviations:
mfg. = manufactured by the computer manufacturer
cs = obtained from a component supplier
semico. = a general product of the semiconductor industry
std. = a standard product available as a commodity from numerous sources
3d = software written by a third-party software house

"Integrated circuit sizes measured in transistors per chip:

SSI 2-64 Small-scaleintegration-circuits suitable for logic
MSI 64-2K Medium-scale integration-circuits suitable for arithmetic units and register arrays
LSI 2K-64K Large-scaleintegration-circuits suitable for small microprocessors and memories
VLSI 64K-2M Very-large-scaleintegration-circuits suitable for all processors and large
memories
ULSI 2M-64M Ultra-large-scale integration-circuits suitable for complete systems on a chip
 Viewing the New Product as Part of the Big Picture 173

Table 8-4. Leading Tube, Transistor, and Integrated Circuit Manufacturers

at Various Times.
Integrated
Tube Transistor 1955 Transistor 1960 Circuit 1978 1989
RCA Hughes Texas Texas NEC
Instruments Instruments

Sylvania Transitron Transitron Motorola Toshiba

GE Philco Philco Fairchild” Hitachi

Raytheon Sylvania GE National Motorola

Westinghouse Texas RCA Intel Fujitsu

Instruments
Amperex GE Motorola RCAb Texas
Instruments

Tungsol RCA Clevite Signetics‘ Mitsubishi

Ranland Westinghouse Fairchild General Intel

Instruments

Eimac Motorola Hughes AMD Matsushita

Philco Clevite Sylvania Mostek‘ Philips

SGS-Thom

Samsu n g

Sharp

Siemens
Sanyo
Oki
AMD
Sony
AT&T
a>National >GE >Philips
note. >went to or was acquired by
174 The Product

companiestend to continueexistingbecause of their establishedbase of customersthat

have installed their computers and have softwareto use on them. These customersare
buying “code museums.”

Products of the First Generation: 1950-1959

In the first generation, large computers operating in batch mode were the dominant
form of computing. They typically cost between $250,000 and $10 million. Using
vacuum-tube technology, the major vendors, IBM and Remington Rand Univac, had a
combined market share of about 90 percent.The total market was small,however, since
primary memorieswere small (lessthan 64kilobytes)and costly, disk memories did not
yet exist, and programming was very difficult. Matters improved near the end of the
decade, however, as IBM introduced its first disk, the M A C , in 1957.Fortran (1959)
and Cobol (1960) became available to assist programmers and to aid in transporting
programs between different manufacturers’ machines. In addition, a few desk-size
computers, costing about $50,000, were introduced.

Products of the Second Generation: 1960-1968

The second generationwas based on transistorcircuitry,and the cost of computers was

reduced enough that largeorganizationscould aff ord computingforroutine commercial
applications. Smaller computersbecame possible, and new companies formed to build
them, includingDigital EquipmentCorporation (DEC;1957),which introduced its first
true minicomputer (the PDP-8) in 1965. Computersbegan to take on a broader role in
control and communications outside of computation and data processing.

Products of the Third Generation: 1969-1977

With Kdby’s and Noyce‘s invention of the integrated circuit (1958),the evolutionary
basis of all subsequent computing classes was established. Whereas in the first two
generations, only a few companies started up, the integrated circuit allowed nearly a
hundred companiesto form to build minicomputers, because the cost and difficulty of
designing circuitry was almost eliminated.Computer-componentindustries emerged
to supply peripherals, memory subsystems, and various types of software, ranging
from languages to applications.
Minicomputers costing $10,000 to $100,000 were developed and embedded into
larger systems-such as process controllers, telephone systems, and mainframes for
small organizationsand groups-establishing the notion of departmentalcomputing.
The mini made a particularly strong impact in technical applications,including facto-
ries, engineering, and scientific applications.Out of necessity,the mini had to commu-
nicate with other systems,because it was accessibleand affordable.The result was that
 Viewing the New Product as Part of the Big Picture 175

it pioneered distributed processing. In 1978, DEC introduced the VAX 11/780 as an

extension of the earlier PDP-11 series and established the superminicomputer class
($250,000-$1million), which had the power and capability of mainframes.
Figure 8-10, which summarizes the outcome of the ninety-two minicomputer
companies founded during the 1970s(Bell, 1984), illustrates the odds of establishing a
viable business in a new product class.

Fifty companies started up and retained autonomy for a while.

Prime continued to grow by acquiring Computervision and in 1989
was in significant debt through a leveraged buyout.
Data General continues, practicing the UNlX religion.
Tandem started in 1975 and has remained successful.
Nine stalled or found niches to support their customer base.
Thirty-eight ceased to exist.

Nine companies merged with larger firms.

Concurrent (formerly Pekin Elmer, formerly Interdata) acquired
Masscomp.
Two continued with nic:he products.
Six ceased to exist.

Eight existing companies built minis.

DEC and IBM continued to build aggressive products.
Control Data Corporation continued as a distributor.
Five ceased manufacturing.

Twenty-five existing nonccomputer firms built minis for special use.

9Hewlett-Packard (HP) ,acquired a company and became successful.
Hughes, Raytheon, anld Texas Instruments (TI) still build special
computers.
Twenty-one ceased manufacturing
Note: This figure does not include struggling Wang Laboratories or the plethora of compa-
nies that started up in the 1980s to make minis using multiple microprocessors.

Figure 8-10. Outcome of Ninety-two 1970s US.Minicomputer Start-ups.

176 The Product

Figure 8-10 gives rise to the following related observations:

IBM always has a large market share, no matter when or what part of the market
it enters, provided that it enters the market.

Few companies can enter fundamentally new businesses. Only one firm, HI’,was
able to survive the transition from its instrument business into computing.

Only four companies can be considered leaders after twenty-five years: DEC, HI’,
IBM, and Tandem.

Sevenwinners in 1980became struggling companies against microprocessorsand

distributed workstations.

Of all the organizations,about 25 percent were successful,in that they survived as

an operational entity at a site.

Entering a well-defined niche is a way of surviving, but not of leading.

The probability of survival for even a few years after a merger was about fifty-fifty.

Products of the Fourth Generation: 1978-1990

The fourth generation could be called ”the age of microprocessors and the dawn of
distributed computing.”Althoughthedatesshownforeachgenerationinthesubsection
headings imply a clear-cut boundary between generations, this is not in fact the case,
because each generationincludesthe maturingproductsof the previous generationand
the seeds of the next generation.In particular, the seeds for the fourth generation were
sown during the third generation, with the production of the Intel 4004, the first and
widely used microprocessor.
Although the computers of preceding generations spawned very few computer
classes, the 4004 and its successors gave rise to many different computer classes, as
shown in Table 8-5. By 1978, single-user personal computers were being configured
around micros with CRTs, keyboards, and floppy and large hard disks (circa 1982)
produced by a high-technologycomponentindustry.Thesepersonalcomputersappeared
in every form, from the home computer, which sold for a few hundred dollars; to
personal computers; to powerful workstations,which initially sold for $50,000.
One of the most important standard components to evolve was the operating
system.Once UNIX and MS/DOS became availableby 1981,the problem of developing
an operating system was reduced to licensing a fully developed product, including all
documentation for reproduction, from AT&T and Microsoft, respectively.
 Viewing the New Product as Part of the Big Picture 177

Table 8-5. Computer Classes Formed from New Microprocessor Introductions.

Companies (Year
Year Class Components Product Introduced)
1971 Calculator Intel 4004 Busicom (1971)
1973 Business terminal Intel 8008 Datapoint (1973)
1973-1977 Personal computer 8080,6502 Micra1 (19731,
SCELBI (1974), Altair
(19751, Commodore
(19771, Radio Shack
(19771, Apple (1977)

1981 IBM PC 8088; MS/DOS IBM (1981)

1981 Workstation 68000; UNIX Apollo (19811, Sun (1982)
1981 Micro" 68000; UNIX Onyx (19801, Altos (19821,
NCR (1982), Plexus
(1983)

1982 FT multi", 68000 Stratus (1982)

1982-1985 Multi", 68000; 32x32 Synapse (1983),
Arix (1984), Sequent
(1985), Encore (1986)

aSubstitutionfor minicomputer-technology-based computers.

bMulti= multiple-microprocessor computer; FI = fault-tolerant multiprocessor

New companiesusing microprocessortechnologyformed to establish the personal

computer, workstation, micro, multi, and other computer classes. As each of the new
classes began to capture a noticeable proportion of an established firm's revenue, that
firm was forced to respond to the new mode of computing. Usually, the new classes
proved unsuccessful, and sales started to decline. IBM is the exception; the IBM PC
(1981) was designed and built in reaction to Apple's success. In 1989, DEC and IBM
entered the workstation market with RISC-based products designed to compete with
the market leader, Sun Microsystems.
The followingobservationscan be made about the introduction of new computers:

Established computer companies4riven externally by their existing customers

and internally by their marketing and engineering groups, organization,and cost
structures-tend to build computers a certain way with a previous design (arclu-
tecture) to service an estabslished customer base. In effect, computer firms usually
178 T h e Product

build code museums to hold programs created on earlier machines and to serve
their present customers.

New computer classes must be created by new ventures that are neither bound by
traditionaluse nor locked into providing compatible code-museum environments.
Early in the developmentof the class, traditionalsupplierstend to aid the new firm’s
formation by buying computers to serve their company.

When a computer start-upbegins to take a noticeable portion of potential revenue

(e.g., about $1 billion) from large, established vendors, they respond by designing
products appropriate to the new style of computing.Thus, the new firm begins to
be limited by competition rather than being assisted by it through distribution
agreements.

Products of the Fifth Generation: 1988-20012

Products of the fifth generationare characterized by a ubiquitous network, ultra-large-

scale integration (ULSI), reduced instruction set computers (RISCs),and parallelism.
Personal computers and their workstation cousins, wluch many feel characterized the
fourth generation, were interconnected via local area networks. In the fifth generation,
all computers are interconnected locally and globally to form a distributed computing
environment.All of the local area networks are finallyinterconnected via a hierarchy of
fast networks, includingmetropolitannetworks,wide areanetworks,andthe integrated
services digital network (ISDN),provided that it ever exists.
If the connection is very tight and a group of individualscan carry out their work
together by operating on the same problems and share the same information, on an
instantaneous basis, then the fifth generation will have arrived. Steve Jobs calls this
”interpersonalcomputing.”At this point, the facility will be as capable and robust as the
original time-sharing systems introduced in 1970!
The fifth generation is also characterized by the use of RISC,which has enabled
performance to evolve at the rate of 60 percent per year (or a factor of 2 every eighteen
months) since 1985. RISC uses a somewhat simpler architecture that permits a single
processor and the important supporting circuitry to be placed on a single chip. The
supporting circuitry includes floating point arithmetic, virtual memory translation
hardware, and fast cache memories. As with previous generations,the fifth generation
will have fully arrived when the RISC architecture has replaced virtually all other
computer classes.
_ _ ~
2. It is especially difficult to tell when one generation ends and another begins when writing a book
near a boundary line. Therefore, the dates for the fourth and fifth generations have been shown with
a two-year overlap.
 Viewing the New Product as Part of the Big Picture 179

(It is important to keep in mind that the start of a new generation can only be
identified after the fact, when history shows that a new technology has replaced
virtually all other forms of computing that characterized the earlier generation.)
The RISC notion dates back to Cocke at IBM Research in the late 1970s. The
chronology of companies adopting RISC for computer products includes:
1983 Introductionof RISC-based computers by Ridge and Pyramid
1986 MIPS Computer Systems (based on Hennesey, Stanford)
1987 IBMs RT Workstation,based on the original 801 project

HP’s Precision (IBM alumni Joel Birnbaum and Bill Worley)

Sun’s SPARC (based on Patterson,University of California at

Berkelley)

1988 Apollo’s Prism

1989 Motorola’s 88000 nnicroprocessor

Intel‘s i860, ii numeric acceleratingprocessor for the 80386/80486

The significance of the various technology factors can be seen in Figure 8-11. The
two bipolar technologies that formed the basis of supercomputers, mainframes, and
minicomputersevolvedataconstantrateof14percentperyear,doublinginperformance
every five years. The CMOS microprocessor,based on traditional architectural ideas,
evolved at roughly 40 percent per year. These microsbegan to overtake the ECL-based
technology in the late 1980s.
However, just as the crossover in performance began to happen, the RISC micro-
processor was introduced by MIPS Computer Systems and created a discontinuity in
performanceby a factor of 2 to 4 over traditional architectures, such as the Intel 80x86
and Motorola 68000. The shift to a RISC architecture, which is inherentlyeasier to build,
provided an additionalstimulusthat acceleratedthe evolution.With RISC, the first chip
to be implemented in a new semiconductorprocess, such as VLSI ECL (circa 1989)and
eventually VLSI gallium arsenide (circa 19951,becomes the newest-generationmicro-
processor.
Unlike previous processor evolution, which averaged product-gestation times of
three to six years, the microprocessor ”tracks” the semiconductor process with a new
designatdouble the performance approximatelyeveryeighteenmonths,as shown inTable
8-6.
Parallel computing,an equally important idea whose time has come, will mark the
fifth generationas clearly as disbributedprocessing and RISC. With parallelprocessing,
performance can be increased almost infinitely by interconnecting a number of
180 The Product

- I I
I I
- I
iips
i mhz
I !CL)

I
I

- I I
I
I

i i
I

I I I
0.1
I ~ I I I I ~ I I I I ~ I

1980 1985 1990

Notes:
Circuit Technologies
ECL = emitter coupled logic (bipolar semiconductor)
TTL = transistor-transistor logic (bipolar semiconductor)
CMOS = complementary metal oxide semiconductor (field effect)

Compufcr Architectures
CISC = complex (or complete) instructor-set computer
RISC = reduced instruction-set computer

Figure 8-11. Performance Versus Time for Various Computer Architectures.

microprocessors. As with all other concepts that mark a generation, nearly every
computer will be built with some form of parallel processing beyond concurrent
compute and input/output. The alternativecomputer structuresof the 1990s,based on
parallel processing, are shown in Table 8-7.
In effect, the monoprogrammed, massively parallel computers become the 1990s
counterpart of the attached array processors that were introduced in the 1970s.

HINTS FOR DESIGNING GREAT PRODUCTS

Designing a great product means more than just having a great idea for the product’s
architecture.It also means doingeverythingright and not committingirreparableflaws.
 Hints for Designing Great Products 181

Table 8-6. Performance Measures for Leading-Edge, RISC, One-Chip

Microprocessors.

Mflopsb
Year Clock (Mhz) PkMipsa Mflopsb with Vector Unit
1986 8 5 1 -
1987 16 10 2 16
1988 25 16 5 25
1989 40 25 - -

ECL Shift Permits:

1990 80 50 10 100

1992 160 100 20 200

"PkMips = millions of instructions per second (peak).
"Mflops = millions of floating point operations per second for the linear algebra package, Linpack,
solving a 100 X 100 matrix.

With shorter design cycles, it is of paramount importance to design products (i.e.,have

an architecture)that can be evolved over several generations, each of which may last
only eighteen to twenty-four months. The following subsectionsexplore how a high-
tech venture can achieve these goals.

MAKING COST AND QUALITY PART OF THE START-UPPLAN

BY DESIGNING THEM IN

As indicated in Chapter 7, quallity plays a critical role in the start-up manufacturing

process. If the level of quality is inadequate, few sales will be made in today's quality-
conscious marketplace, and the product will be doomed. Without incorporating cost
and quality goals into the product design, it is impossibleto create an accuratebusiness
plan and model for the company, because quality affects sales volumes and product
support costs as well as warranty and staffing costs. The worst thing about making a
poor product is that engineersmust fly all over the world ina chaoticfashion to diagnose
and repair it.
I believe the best way for a start-up to address the quality issue is simply to sell its
products with a money-back guarantee. This approach dramatically simplifies the
contractingprocess,becauseit avoidscustomeracceptancetests,performanceguarantees,
and all the legal mumbojumbo of the contractsthat customersformerlyused to protect
themselves against nonoperational systems. Here's how two firms successfully em-
ployed such guarantees to promote the sale of their products.
182 The Product

Table 8-7.Parallel Computer Types, Style of Use, and Application

(in Increasing Order of Specificity).

Computer Type Form Application

General-Purpose
Simple uniprocessor A processor and its memory Simplest PC, workstation,
microprocessor

Multiprocessor All processors sharing memory All mainframes, minis,

workstations, large-
transaction processing

Multivector processor Multi, using vector processors All types of

supercomputers

Monoprogrammed for a Single Task at a Given Time, Multiple Uses

Multicomputer Interconnected computers Large area networks, high
reliability, technical high
performance, distributed
database
SIMD (single Massive (1M) data parallelism Supercomputing, signal
instruction multiple processing
data)

Bound to a Single Application by Hardware and Software

Array processor 4-8 processing elements Digital signal processing
Special SIMD Graphics and image processors CAD, visualization
Neural net Pattern recognition Speech, signal processing
Systolic processor Pipelined processing Specialized signal
processing
Data-flow processor Research phase Unlikely multicomputer

Silicon Compilers’ and Ardent’s guarantees. The chip CAD softwaresold by Silicon
Compilers was first offered with the guaranteethat any chip produced by the compiler
would operate according to the specifications.This eliminated the fear of buymg the
compiler and still not being able to get a working chip.
At Ardent, we offered a thirty-day money-back guarantee that allowed buyers to
return computers in the event they were unsatisfied with their purchase.
Both these firms found that the contracting process was considerably simplified.In
Ardent’scase,there wasno need for lengthyandcomplexacceptancecriteria.Customers
simply tried the product, with any and all programs they chose, and decided whether
or not to keep it. Out of a thousand computers delivered, only two were returned.
 Hints for Designing Great Products 183

THE BUILDER-IS-THE-USER PRODUCTS

The easiest and best kind of products to build are those originally intended for use
by the engineer or his or her close friends. In this situation, the designer knows the
users and gets immediate feedback about the product. The ultimate developer-is-
the-user product is a C language compiler, which is written in its own C language
that the product developer must constantly use and enhance. This approach has
been employed to develop operating systems (especially UNIX); most computer
languages, from Algol to Zetai-LISP; generic programs such as word-processing
systems;and operating environments, including windows and desktops. Computer
languages inherently originate with a user who needs to perform operations (the
verbs that form the syntax of the language) on special data structures.
Nearly all computers-including minis, the first time-shared computers,PCs, and
workstations-were developed by people who needed them for an application. Sun
Microsystemsclaimed that a major factor in its successwas hardware engineersselling
to software engineers, and HI' called this the "next bench syndrome (i.e., the product
will be used by the person at the bench next to the developer).Declared Jobs at Apple:
"Never build a computer you wouldn't want to own."

The minicomputer, Cray, and spreadsheets. Nearly all the computerswith which I've
been involved were first designed for a particular and personal use. Digital's PDP-5,
progenitor of the first minicomputer, was developed in 1965.It originated with a need
to build a special-purpose data-acquisition device for an Atomic Energy of Canada
reactor. Instead of creating a one-shot front end, we took the opportunity to build a
component(tool)that could beused to build this and many other applications.Ted Hoff
credits the PDP-5's successor, the PDP-8, as being the design model for developing
Intel's first microprocessor.In elfect, the first mini was the model for using embedded
computers for control.
Seymour Cray claims that he is merely building computers for his friends who do
numerical simulation. And Dan Bricklin has described conceptualizing the first
spreadsheet,Visicalc,as a student at Harvard Business Schooland personally working
with financial spreadsheets.
When a companydesignsproductsaccordingto the "builder-is-the-usef'paradigm,
it will have no trouble identifyrng the users and their needs. However, it is less clear
whether a very large body of customers will respond the same way as the initial
designer/user group and buy these products. Thus, even a well-designed product may
only appeal to a select market, which, when satisfied, is saturated.As an example,Gold
Hills made a LISP operating environment for the PC and grew very rapidly to satisfy
its market.Unfortunately, themarket (in 1990)consisted of a slowlygrowingpopulation
of about ten thousand programmers who were LISP users and evangelists. Once that
market was satisfied, sales dropped sharply, and the firm had to downsize itself to fit
184 The Product

the true market. The advantagesand dangers of forming a company this way-i.e., by
picking atypical users to satisfy-should be obvious.

EVOLVABILITY AS A REQUIREMENT

As the new venture starts up, it should be clear about whether its technology position
will enable it to develop subsequent products based on the first, core product. With
shorter product-gestationtimes for hardware, it is crucial to build systems that can be
evolved easily and over a range of products using the same basic technology or
components.Without evolvability,the companywon't be able to get large enough, fast
enough to become self-sustaining, since each new product will be disjointed from the
preceding ones. Software evolvability implies building a product that can be either
modified to include new features with every release or, alternatively,downsized and
simplified in order to address markets that do not require all the features of the initial
version.
An excellentway to meet the need for computer-systemevolvabilityis to design the
product as a "multi." The multi is a type of computer introduced in the early 1980swhen
the first powerful microprocessorsbecame available.It is built by connectinga number
of processors, memories, and input/output modules to a single bus whereby any
modulecancommunicatewithany other.Althoughtheuseof auniversalbus tosupport
multiple processors is a relatively new idea, the concept of a universal bus to support a
processor, memories, and input/output devices dates back to DEC's PDP-11 Unibus,3
introduced in 1970. Multis "work because each processor has local cache memory,
reducing the bus traffic.
The multi offers a number of advantagesover computers built as discrete projects
(see Figure 8-12), including lower product and development costs, greater reliability,
and improved interconnections to other systems.Furthermore, a multi can be evolved
as new processor and memory componentsbecome available. Only a few mainframes
and supercomputersdelivercomparableperformance,and none equalsthe performance
per dollar, performance, and price range or is able to be evolved automatically.
Figure 8-12 shows the Encore Multimax product line (computer family A in the
figure) and contrasts it with computers built as a series of discrete projects. The
Multimaxevolved over time in two ways: (1)the processormodulewas simplyreplaced
with the next-generation microprocessor, and (2) a version was developed that could
only grow to half the maximum size of the original model (ten modules versus twenty
modules).Table 8-8 summarizes this evolution.

3. The Unibus operated at 2 megabytes per second and was the principal method of interconnecting
options to the PDP-11 and VAX computers for almost fifteen years. The VME bus (operating at 20
megabytes per second) is an industry-standard bus related to the Unibus, but it may not have as long a
life due to rapid technological change.
 Hints for Designing Great Products 185

15

13

11

- I I I I I 1 I
50 150 250 350 450

List price of central processing units

(thousands of dollars)
Figure 8-12. Performance Plotted Against Price for Two "Multis" and Four Conventional
Computer Families Available in 1985.

Becauseof the compellingadvantages,allbut the most trivial and largest computers

are built in this fashion as multiprocessors, including those made by Apollo, Arete
186 The Product

Table 8-8. Evolution of Encore Multimax Performance Versus Time.

Performance Range
Model Date Cache (VAX 780 Units)
120 1985 32K 13-15

310 1987 64K 4.0-20

320 1987 64K 4.0-40

510 1989 256K 17.0-85

500 1989 256K 17.0-170

&Theprocessing module has two processors. Up to twenty processors can exist in one system. The
bus that interconnects processors and memory transfers 64-bit words at a 12.5 megahertz rate (100
megabytes per second).

(Ark), Compaq, DEC, Data General, Encore, Masscomp, MIPS, Motorola, NCR,
Sequent, SiliconGraphics, Solborne, Stardent,and Stratus. PCs of the 1990s will evolve
this way, and by 2001, many microprocessors will be packaged on a single chip.
Designing and building computers as multis dramatically changes the industry and
workforce,sincea team of fifty engineerscan produce a better productthan three or four
teams of five hundred building a range of "point" products. Similarly,a factory needs
to produce only a few board types, resulting in drastically reduced product costs due
to manufacturing economies and design improvements.

PRODUCT FLAWS

It is often difficult to separate flaws involving the product per se from those involving
flawed technology or a flawed business plan. The lack of a concrete and producible
product that can be delivered for a customer application is clearly a product flaw. In a
world of standards, an incompatible product that requires a whole new infrastructure
of support is almost certainly doomed. F a h g to get all the details of the product
specificationcorrect when a company is attemptingto establish a new product class or
new product niche can also be fatal. The followingsubsectionsexamine these and other
product flaws that can derail the unwary start-up.

BUILDING A SPECIALIZED PRODUCT

WHEN A PLAIN OLD COMPUTER WILL DO

An ordinary computer,with appropriatesoftware,will usually fdl a buyer's need better

than a specialized product. Thus, any companybuilding a special system is likely to be
a software supplier or a value-added reseller (VAR) with a computer application.
 Product Flaws 187

Often, as a new applicationis discovered, the temptation is to respond to that need

by making a special product or even a unique computer. Obviously,if the product is
trivial or intended for another Junction, a general-purpose computer won't be able to
compete. For example, all clocks and watches have a speciallyprogrammed computer
as their basis, and a general-purposecomputer cannot perform this functionbecause of
cost and packaging. Clocksand watches ultimatelytake on many of the functionsof on-
board human information-processing devices like calculators, calendars, and phone
books. Furthermore, these timekeeping devices can communicate by display and by
voice. In the very long run,when computers will cost virtually nothing and occupy
virtually no volume, all compui:erswill perform the same infinite number of functions.
Table 8-9 lists firms that all1 started building unique computers but subsequently
discovered that they were conclentratingon what they bought, not what they planned
to sell, and eventually had to focus on their application.The scenario is invariably the
same:

A high-tech venture starts by building its own proprietary computer in order to sell
hardwareandget higher list prices and margins.The firm's first fallbackis toresellanother
supplier'sproduct in order to keep its revenue high. Finally, the company has to resort to
selling only software, since customers already have a computer or a channel through
which to buy one and merely want another applications program.

HAVING TECHNOLOGY BUT NO PRODUCT

Somestart-upsfail to definean actualproduct concept and have only a set of techniques

or technology (e.g.,artificialintelligence [AI]or rulebased programming) for building
products. In the case of products such as those built using the emerging multimedia
technology, it is difficult to distinguish between a flashy demo and a complete system
that more than a few users might buy. Thus, testing for the existenceof a product versus
a technology can be quite difficult.
Artificial intelligence product companies of the early 1980s clearly illustrate the
difference between having technology and having a product. In 1980, the Japanese
issued a challenge to the world by initiating a fifth-generation computer research
program based on nonprocedural programming, whereby highly parallel computers
would perform functions in a "humanlike" fasluon based on artificial intelligence
research. This stimulated the birth of a number of "AI and rule-based systems
companies as well as increased funding for computer-scienceresearch, including the
formation of research consortia such as Microelectronics and Computer Technology
Corporation (MCC) and Semiconductor Research Corporation (SRC). The case of
BREIT Internationalillustrates the problem of focusing on technology as opposed to a
product.
188 The Product

Table 8-9. Products Unable to Compete with General-PurposeComputers.

Era Product Company Examples
1960s Process control Foxboro, Leads and Northrup,
Honeywell
1965 Nuclear instrumentation Chicago Nuclear

1970s Industrial controller DEC, Gould

Mechanical design stations Computervision
Typesetting stations Compugraphic
1980s ECAD (electronic Daisy, Valid Logic
computer-aided design)
stations

Electronic typewriters Brother, Smith Corona

Word Processor Wang
1985 Document processing Cygnet, Filenet
1990s Handy pocket minders Casio, Sharp
All Military computers All military vendors

BREIT International. In August 1984, BREIT's CEO, a former computer salesperson,

came together with a Martin-Marietta programmer working on military AI programs.
BREIT started with eight people (six of them programmers), riding the AI wave. The
productive and creative programmers were versed in AI and system programming,
including natural language recognition,rudimentary databases and knowledgebases,
and graphical interface design. A hardware person with knowledge of videodisk
technology also joined. BREIT defined "knowledge transfer" as its product area, thus
separating itself from the unsuccessful computer-aided instruction product area.
BREIT's staff-which had grown to include the president, a marketing person and
a consultant, two salespeople, two administrators, two secretaries, and ten software
engineers-worked togetherfor twenty-twomonths.After the first fifteen months, they
alpha-testedan intelligentcomputer-aidedinstructionprogram with a superbgraphical
interface.Based on the alpha tests, the firm concluded that the product was too general
and couldn't solve a real need. The last seven months were aimed at further generali-
zation and, hence, less emphasis on a specific user.
BREIT received two rounds of financing (spending about $1.5 million), which
resulted in the production of interestingdemonstrationsof what might ultimately have
become a product. Had the company been funded for a seed stage, instead of being
 Product Flaws 189

directly formed, time and money could have been saved. Alternatively, given that
BREW had invested a small amount in developing a technology, additional funding
might have produced a product that users would buy, if the firm had had sufficient
maturity and savvy to direct the conversion of its technology into a useful product.
BREIT may simply have quit too early and had the wrong product target.

FALLING IN LOVE WITH AN IMAGE OR MOCK-UP OF A PRODUCT

An attractive,but superficial,mock-up of a product conceptor demo may be built to sell

investors and future employees. Although the product might be flawed in some way
(e.g., a commodity, unbuildable, too expensive), the model can override rationality.
Being ableto see or touch a product concept means an almostcertain saleto product
or company funders. At DEC, practically any physical model was a key to selling the
development of a product. Many notable firms got their start with a beautifully
designedmock-up.As the follovving examplesillustrate,however, countlessill-founded
ventures and bad product ideas got their start the same way.

Viatron, Zilog, DEC, and ”Company X”. In 1970, Viatron had working models of its
MOS (metaloxidesemiconductor)-baseddesktopcomputers,whichit offered at $49per
month. Although the models played a key role in the marketing to investors and
prospective customers, they were driven by real, working computers. The company
couldn’tbuild theMOSchips.Beforeit went bankrupt, Viatron,which had gone public,
set a 1970srecord for start-up losses. The president indicated that the key to his fund-
raising successwas promisingincrediblereturnsbased on ambitioustechnologyclaims.
In 1981,Zilog built a wonderfulwooden model of one of its first UNIX systems, the
S8000. A female reporter took one of Zilog’s executives to dinner and began querying
him on the computer and when it would be ready to ship. Wanting to seem powerful
and important,he replied, “What do you mean ship? It’s a *+#@! wooden model!” The
quote was printed verbatim. The company eventually left the systems business.
One of the best and most creative industrial designers I know is Ken Olsen.
Although customers rarely saw his prototypes, the engineering model shops at DEC
could turn out models of a design in metal or plastic in one to three days, depending on
the complexity and material. IJnfortunately, the boxes’ appeal often oversold Ken,
especiallyif the general concept was h s idea, and going fromproto to product often cost
the firm tens of millions of dollars.What went into the boxes was often less than useful
to a customer. An example of this occurred in the early 1980s,when DEC’s disastrous
foray into PCs cost the company nearly a billion dollars.
The worst case of the model’sbeing the product occurred at a firmI’ll call Company
X, which built a box with blinking lights to satisfy government tests so that it could
receive a progress payment on its contract. The people involved went to jail.
190 T h e Product

Despite these horror stories, models do have their place. ConvergentTechnologies

foundoutthatAT&TwantedaUNIXPCandbuiltacardboardandplasticmodelof such
a device. AT&T bought a hundred thousand computers. When it sold Megaframes to
computer companiesas a brand relabeler, it could color, logo, and repackage a model
for a new box even before a firm asked. The model became the salespeople's tool kit.

BUILDING AN INCOMPATIBLE PRODUCT

Engineers are especiallyprone to ill-directed creativity when it comes to inventing new

and incompatible, but not necessarilybetter, systems.
A hgh-tech venture must either createthe standardsor follow them. If the company
failsto make its new product a standard, then it gets to build the product t w i c e t h e first
time its way and the second time according to the market's requirements.
One of the most serious flaws is for a firm to (relinventa new architecture, protocol,
language,human interface, file format, etc., when an existing one is just fine or could be
evolved to do the job. Incompatibility is the most costly form of the not-invented-here
syndrome. An excellent case in point is the first computer I designed at DEC, the
PDP-4, whch was incompatible with its predecessor. Fortunately, DEC survived.
Today, no start-up making such a dumb mistake on such a grand scale would survive.
Yet almost every start-up invents some new interface that should in fact have been
compatible with an established or de facto standard. The following examples illustrate
the dangers of blazing a new trail-especially if that trail parallels an existing super-
highway.

DEC's first 18-bit computer. DEC's first computer, the PDP-I, was introduced in 1960
and had an 18-bit word. Three years later, I designed a second 18-bit computer, the
PDP-4, which cost roughly half as much and used different components,including a
new input/output scheme oriented toward process control. The PDP-4 was somewhat
simpler to build and solved a few problems better than the PDP-I, but as noted above,
it was incompatiblewith the earliermodel, and all its softwarehad tobe redesignedfrom
scratch. We made no attempt to have the two use the same languages. A ll subsequent
18-bit computers were PDP-4-compatible. It was a tremendous and silly waste of
resources for a smallcompany.At the time,none of us had an understandingof the costs
and importanceof software,both to DEC and to its customers.The right solutionwould
have been to have evolved the PDP-I, using the new ideas.

Ardent, Stellar, and Stardent architectures. Titan, Ardent's graphics supercomputer,

used multiple MIPS chips as the basic computing engine. Ardent bought the manu-
facturing rights to Raster Technologies chips for coloring 3-D polygons. The team
started by extending the MIPS scalar architecture to connect a fast vector processor
 Product Flaws 191

according to the Cray supercomputer design formula. In the process, it dropped

compatibilitywith the MIPS system.This meant that system software, includingUNIX
and compilers, was unique. Pirograms such as third-party databases, compilers, and
other applicationsrunning on tlhe MIPS system could not be used on Titanwithout being
recompiled and retested. A year later, Ardent realized its mistake as customers
demanded software, much of which was available from MIPS, and a MIPScompatible
interfacewasbuilt. Becauseof Ardent’sfailureto design a compatibleproduct in the first
place, the Titan project required more work than would otherwisehave been necessary,
the product got to market later, and the company was jeopardized.
Stellar started at an even lower level by inventing a unique architecture for both
computation and graphics and had to do even more design work. When Ardent and
Stellar merged to form Stardent,the company settled on an architecturethat is upward-
compatible with MIPS yet exploits the Ardent multiple-processor technology of
supercomputing(architectureand compilers that understand parallelization).

Amdahl Corporation. In the early 1970s,Amdahl was created to build very-high-end

IBM 360-compatibleproducts. Just as Amdahl’s first product was being designed, IBM
modified the archtecture to provide a significant virtual memory function, necessitat-
ing a major redesign. The resulting delay in the product essentially cost Amdahl its
ownership by forcing it to obtain additional funds from its Japanese partner, Fujitsu.
It is essential for a start-ulp to understand, and have a rational policy about, the
. standardization or proprietariness of various interfaces that it maintains. When an
existing standard is adequate, then it should be followed exactly. The decisionbecomes
more difficultwhen a new standard is needed and is in the process of being developed,
either on a de facto basis by a large supplier or by a committee.

Stardent’s DorC graphics library and AVS visualization architectures. Stardent at-
tempted to establish two Standards for 3-D graphics and for visualization-Dore and
AVS. Dore was created during the time when a standards body from various manufac-
turers was worlung on Phigs+. Phigs+ became the standard, although Dore was
superior by almost all measures. Dore was freely licensed to all comers, and it runs as
a second standard on many different computingplatforms. Stardent must support both
Dore and Phigs+ on its platforms.
AVS (ApplicationVisualizationSystem)is a high-levelprogramming environment
that enables the user to take scientific data and view it in a flexible and interactive
fashion. Stardent salespeople regard AVS as providing a significant competitive
advantage. AVS is licensed to other companies, but not to firms that Stardent views as
competitors, such as Silicon Graphcs. However, SiliconGraphicsnow employsAVS’s
inventor and architect and may build a more competitiveproduct, which it would then
licenseto other companies,but not to firmsthat it views as competitors,such as Stardent.
192 The Product

If this occurs,the small but growing community of users who require visualizationwill
have to learn, and choose between, two similar but incompatible systems.Stay tuned.

STRATEGICPARTNERING THAT
GIVES AWAY PRODUCT AND TECHNOLOGY

In this age of strategicpartnerships,it is not uncommonfor a start-upto be workinghand

in hand with a partner that produces, resells, builds on, or otherwise significantly
contributes to the start-up’s product. In such cases, it is very easy for the fledgling
company to end up giving away its product and/or the technology on which the
product is built.

Lockheed and Dessault; Design Power and Intellicorp. Lockheed developed a

mechanical design package, CADAM, that was based on the standard 2-D represen-
tational structure of traditionalblueprints. It gave the sourcecode to the French aircraft
maker Dessault, which turned the package into a 3-D design program, called CATIA.
Today, CATIA is one of the dominant CAD/CAM programs.
When Design Power started up, it developed a design program called D++that was
written in KEE, a language created by Intellicorp. Soon, Design Power had an oppor-
tunity to raise quick cash by selling D++to Intellicorp.The sale looked too easy and too
good to be true. Fortunately,Design Power wrote a very strong nondisclosurestatement
into the agreement to protect D++,making it very difficult for anyone seeing the source
code to write a design program. What Design Power was attempting to do was to get
Intellicorp to become its VAR and service provider for the program. However, if no
strings had been attached to the sale, Intellicorp would have received a direct infusion
of technologyand product, thereby enablingit to take D++ and sell it in exactly the same
fashionas Design Power.Withoutthe restrictions,in short, sellingD++ would havebeen
equivalent to selling Design Power itself.

DEVELOPING A PRODUCT SPECIFICATIONTHAT

IGNORES A CRITICAL APPLICATION REQUIREMENT

Ignoring the details often dooms a high-tech venture, as a host of firms building
technical computers have discovered (Chopp,Culler Scientific,Cydrome,E&S, Elexsi,
ETA, etc.).The Ardent story below points out how focusingon just one or two attributes
and neglecting the really important ones can spell disaster. The start-up must have a
detailed understanding of the requirements when it begins the project; otherwise, the
resulting product may be fatally deficient.
 Product Flaws 193

Positioning the first graphics supercomputer at Ardent. In 1986, Ardent's goal at

start-up was to produce a new class of computer, the graphics supercomputer, to solve
the following computing-visualizationparadox:

Supercomputers provide excellent computational capability but have no graphics capa-

bility.Graphics workstations provide the ability to visualize 3-D objects (eg.,molecules,
mechanical structures,physical :;ystem modeling)but have littlecomputational ability. A
high-performancegraphicsworkstation cannot be achieved throughnetworking because
both local and national networks are inadequate to connect the workstations to
supercomputers.

When Ardent began, visualization was just starting to be recognized as a critical

need for supercomputing. Visualization simply means displaying the results of a
computation graphically in order to provide users with insight.Interactive visualization
extends this notion to let users interact with the computation by visualizing its state in
order to guide future computationand thereby enable them to participatein analysisor
design. Figure 8-13 shows how Ardent positioned its product as the supercomputing
equivalent of a workstation.
Ardent decided to use a Cray-style supercomputer architecture with a vector
multiprocessor and high-performance 3-D graphicshardware.Supercomputersobtain
their very high speeds by operating on a string or set of numbers (a vector) all at once.
Vector-processingrates are measured in millions or billions (and eventuallytrillions)of
floating point numbers per second. Linpack, for solving a set of linear equations,is one
benchmark for measuring this. I sketched the performance (in Linpack megaflops)
versus price curve, as shown in Figure 8-1, to initially position Ardent's first product,
Titan.
Unfortunately, no single dimension can be used to characterize a computer. The
other performancedimensions for the product that we recognize in hindsight (shown
in Table 8-10) include integers,rneasured in millions of instructionsper second (mips),
which is typical of systemprograms;scalarsor operationson singlenumbers, measured
by the Whetstones benchmark; vectors, measured in floating point operations per
second;graphics,measuredby shaded polygonsand linesper second;images, measured
by pixels per second; and input /output, measured as a data rate to disk and network.
The requirements for each of these performance dimensions for various markets and
attributes of the computer range from nonexistent or very low to very high.
Titan was originallydesigned to achievehigh performanceby using a RISC chip set
manufacturedby MIPSComputerSystems.One chipwas a high-speed microprocessor,
and the other was a coprocessor to accelerate floating point operations. As the design
progressed, Ardent focused almost exclusively on vector processing and radically
194 The Product

The Vector Equivalent of Technical Workstations

IntegerlScalar Computers Vector Computers

Multiple User
Textual Interface

Departmental

1
Workstations Supers Ardent
Deptmersonal
Single User
Graphic Interface

Figure 8-13. Ardent’sGraphics Supercomputer Positioned as a New Computer Class

Formed as the Vector Equivalent of Technical Workstations.

reduced the scalarperformanceby removingthe MIPS floatingpoint chip to cut product

cost and simplifydevelopment.Unfortunately,almosthalf the market (generalpurpose
and chemistry) used extensive scalar operations, since their computations weren’t
suitablefor vectors.The problem posed by the questionablescalarperformancebecame
apparent in the middle of the product development stage but was not remedied.
When the Titan ran the first customer benchmarks,it barely beat a lowly worksta-
tion, sincethe scalar performance was low and a few other features (suchas a fast divide
instruction)were not as fast as the market required.The price, targeted to be comparable
tothat of otherworkstations($50,000),endedupclosertotwiceasmuch.Titan’sstrength
and weakness were the same: it was a supercomputerwith excellent performance for
its price, but its price was higher than that of a standard workstation. For ordinary
programs, it performed similarly to $50,000 high-performanceworkstations, but for
programs written to be used on a supercomputer, it performed better than $300,000
minisupercomputers.
 Product Flaws 195

Table 8-10. Requirements for Performance Dimensions and Markets”.

Market
Perfo rman ce Fluid lmage General
Dimension Chemistry Dynamics M C A D Petroleum Processing Animation Purpose
Integers m S m m S m 1
Scalars lb S m S S m lb
Vectors 1 V 1 V V 1 m
Graphics v S 1 m m V -
Images - S - V V m -
Input / m m 1 1 V 1 m
output
%=very large, l=large, m=medium, s=small, -=negligible
bFailureto provide good scalar performance cost the chemistry and general purpose markets

Thus, depending on the user, Titan looked either like: (1)a fast, but very expensive,
workstation or (2) a bargain-pi-iced minisupercomputer. Workstation buyers felt that
the price was too high, and price has a tremendous effect on volume. (I believe that
demand increases by five to ten times every time the price is halved.) At a price of
$100,000, demand was estimated to be one thousand units, whereas at $50,000, the
volumewould have been at least ten thousand units. Minisupercomputerbuyers found
Titan’s price very attractive,but these customers buy only one unit at a time and put
vendors through a rigorous beiichmark screening process, which meant that the cost of
each sale was roughly half the price of the machine.
Furthermore, an extraordinary sales and marketing effort was required to select
pretrained users. Ardent’s salesforce failed in this regard, especiallysinceArdent never
had a head of sales for very lon,g.An excellent benchmarkinggroup was ultimately put
in place to identlfy prospects until the next product came out, remedying the problem
by having very fast scalar speeds.
Ardent was able to produce two next models-one that was cheaper and one that
was more expensive. This allowed a higher average selling price, since the company
could more fully satisfy the users’ needs. The low-priced model performed more like a
personal supercomputer and could compete with ordinary workstations across the
board and still offer supercomputingcapability. The more expensive model was sold
both as the original and rarefied low-volume graphics supercomputerand as a higher-
margin, larger, and more expensive minisupercomputer. Having a faultless first
196 The Product

productwould clearlyhavechangedthe courseof the firm,but virtuallyallsupercomputer

builders have made errors on their h s t machine,includingAlliant, CDC, Convex,ETA,
NEC, and Stellar.

PRODUCT RULES

At the concept stage, no attempt is made to determine a product’s efficacy, only that
there exists a product or service concept that a customer could evaluate and decide to
buy. The product rules in this section test the rationale for why customers will buy the
product, the market size, and how the product will be sold.

At the concept stage, has the company translated its technologyuniquenessinto

a relatively concrete product concept (what) that is also self-sustaining-
i.e., that provides for the evolution of future generations of the product?

This rule requires that all aspects of the product designbe traceableto technological
roots. Before the new venture proceeds into the seed stage, it must have an outline for
a possible product architecture capable of sustaining itself over several product gen-
erations.The product concept (i.e.,what) can be tested by looking at its specificationor
a product mock-up or an operating breadboard that a potential user can examine and
react to. Product mock-ups are invariablykey props for sellingpotentialcustomersand,
hence, investors.In the case of a hardware or software system that is sirmlar to existing
products, a user, buyer, or buyer surrogate (marketing person) can simply look at
specificationsof the product’s features and functions.

Have rough goals been formulated for product cost as well as quality and
compatibility?

For hardware, the most importantgoalis manufacturingcost;for software,the most

important goals are size, performance, and operating-environmentcompatibility.This
rule emphasizes reality by forcing the company to attempt to put a cost on its product
right from the concept stage. Only in this way is there any hope of sketching a financial
model for the firm and, hence, testing its financial viability. By the seed stage, the
start-up must have a good cost estimate in order to be able to create a realisticbusiness
plan.

By the seed stage, does a product definitionor functional specificationexist for

the product being designed?
 Product Rules 197

Having a detailed understanding of the specific customers and their requirements

is essential in order for a start-up to focus on the product design. The functional
specificationshould includegoids and constraintsfor the product and project, functions
to be performed,user characteristicsand operating environments,a proposed solution,
design trade-offs, and acceptarice criteria.

Does there exist a data sheet that spells out the planned features, functions,and
benefits?

By the time the seed stage is finished, the start-up should have a pretty firm idea of
what it will be building. The simplest way to test for this is to examine the product’s
preliminaryspecificationsin the form of a data sheetor brochure.In its “test marketing,”
the company usually uses the (datasheet to prepare a set of overheads describing the
product. This presentation should enable prospectivebuyers to understand and react
favorably to the product (their response should include pinpointing missing features
and describing their own buying requirements). The product should be receiving
favorable reactions from all prospective customers at this point, since it’s unlikely that
a product will ever get any better than t h s initial conceptual view.

If the company is producing a software product, does a preliminary user‘s

manual exist?

In the case of software, the start-up must make the product quite specific by
developing a user’s manual that fully describes the program and how to use it.

By the end of the seed stage, does the product continue to show a minimum of
a one-year product lead?

In order for a lasting company to be formed, the product must still be viable at the
time the venture starts up. During the seed stage, new technology and products will
have become available,and the competitivepicture may be entirelydifferent from what
it was during the concept stage.The company should ask itself whether its originalidea
really remainsviable in view of the currentcompetitivescene.Whether the organization
continues to maintain a comfortable lead in product development at the beginning of
the product development stage should almost be the determining factor in deciding to
proceed with the start-up.

Is the product architecture capable of evolving into multiple products or lines

rather than yielding only a single, ”point” product?
198 The Product

This rule examines whether the product is based on a lasting architecture or is

merely a one-shot phenomenon. A lasting technology and a lasting product base are
required in order to produce a lastingcompany.In the process of examining the market,
other alternatives may come to light. The start-up’s goal should be to use either
hardware or software technology as aggressively as possible, while not risking the
schedule.

When positioning a product in a multidimensional space, does the company

understand and attend to all the dimensions, not just the most glamorous?

As an example, I wrote the following eleven rules of supercomputerdesign based

on the Stardent experience. Many of these rules apply to other products as well.

1. Performance,performance, and performanceare the top three objectivecriteria

for a supercomputer.

2. Amdahl’s law: generalized, implies that everything matters-a variant of the

proposition that ”no chain is stronger than its weakest link.” Or as the architect
Mies van der Rohe stated: ”God is in the details.”

3. The scalar speed matters most, and a new super must be the fastest of
comparablecomputers in its class. If it cannot do all the mundane calculations
fast enough, a computer is doomed to a niche and is likely to be unsuccessful.
Furthermore, it will not be able to replace its earlier predecessors7

4. The vector speed-i.e., the computer‘s advertised s p e e d 4 a n be as arbitrarily

high as costs allow. The past rule of thumb was to have a vector unit that would
produce two results per clock tick. Large increases (i.e., of more than one
hundred) over the scalar speed provide only a smallbenefitexcept for selected
applications, making the computer special-purpose (such as the Connection
Machine).The NEC SX-3 has a peak speed of sixteentimes the clock.The vector
(peak),or advertised, speed is the speed that the manufacturer guarantees the
computer will not exceed on any application.
4. Amdahl’s law for building high-performance computers states that every computation is composed
of a serial part that must be run sequentially and a part that can be run in parallel at the highest speed using
multiple processing elements. The overall speed of the computation is controlled by the slow, or serial,
part.
5. This is the CDC Star rule. Star was designed in the late 1960s, and two machines were installed at
Livermore. Neither of them worked well, nor were they able to take on a significant workload because
the new computer could not compute as fast as the CDC 7600, its predecessor, for scalar problems. Users
weren’t able to optimize the computer for vector applications quickly enough, since the compilers were
inadequate. Star was the basis of the CDC 205 and later ETA architectures. None of these computers ran
fast enough on a broad set of applications to be commercially viable.
 Conclusion 199

5. Allow no holes in the performance space (e.g., arithmetic function, input/

output, mass storage) into which a benchmark can step, resulting in large
performance losses.

6. Provide peaks in the performance space in order to produce extraordinary

performance for a benchmark. Use this single number to advertise (character-
ize) the machine and to challenge other machines.

7. Obey computer-design law number one: provide enough address bits for a
decade of constant architectureimplementation.

8. Build at least two generationsof the architecture.No supercomputer designed

by a team has ever been perfect the first time around. Do it again after the first
generation has been used and is really understood.

9. Build on the work of others. Designing a super is hard, so understand exactly

why and how every existing machine works and move forward using t h s
knowledge. Make sure the machine can run as much existing software as
possible.

10. Make the machineeasy to use. Have a great compilerand diagnostictools to aid
users in vectorization imd parallelization.Training for supers is nonexistent in
academe, since computer-sciencedepartments are not oriented toward train-
ing people to use computersor to operatecomputersthat produce numbers.No
computer-sciencetexts exist, or are likely to exist, dealing with how to program
a parallel, vector processor (Le., supercomputer).

11. Have an abundance of resources when embarking on the design of a

supercomputer.The fatality rate for companiesmaking machines is at least 50
percent, even though the design may be good. Building a new super costs a
minimum of $200 millionin 1990(or$50millionfora minisuper)just to produce
a breadboard or simulation of the machine.

CONCLUSION

At the root of having a great product is first conceptualizing a product that is complete
and unique and then understanding the rationale for why customers will buy the
product. Since the introduction of the first commercialmainframe in 1950, computers
have evolved,based on the evolutionof semiconductorand magnetic density increases
(Chapter 5), to form established price classes: supercomputers, mainframes, minicom-
puters (also superminicomputers and minisupercomputers), workstations, and
200 The Product

various other personal computers, including notebook-sizecomputers. By 2001, there

will emerge powerful, useful, and ubiquitous pocket- and wallet-size computers that
everyone will carry on their person. Each of these computerclasses requires and attracts
unique application software products. In the fifth generation of computing,beginning
in 1988,productsarebased on networks,standards,and distributed,personalcomputing,
which Steve Jobscalls “interpersonalcomputing.”
Chapter 9

MARKETING AND SALES

We look for market first when deciding to invest.

-Don Valentine, Sequoia Capital

Marketing and sales are so interdependent that sales personnel often walk around
with titles like ”marketing representative.” When the product is ready to be sold,
marketing is responsible for clearly segmenting the market, providing initial sales
leads, and supplying the righk product information for the sales personnel and
customers. Sales is responsible for identifying specific buyers, closing sales, and
when the product has been delivered, ensuring that the customers are happy.

MARKETING

In the past, I have argued that marketing organizations provide little in the way of
value. From an engineer’s perspective, marketing facilitates the birthing of a
product by initially establishing the requirements that help define the product and
by creating a wonderful image of the product in the minds of potential buyers,
making them want it. Finally,when the actual product becomes available, marketing
helps a sales organization sell it. In the 1970s and 1980s, I believed that obtaining
market input during the formation of a product was usually a waste of time, based
on my own experienceat Digit,alin driving system and generic computer products.’

1. For example, when I led the team that defined the VAX architecture and when the VAX computing
environments were put in place, the only input we sought was from Ken Thompson, one of the UNIX
developers. One marketing person attended biweekly status meetings, at which he was assigned data-
gathering tasks.
201
202 Marketing and Sales

I took this stand because most engineers like to hide behind the cloak of “marketing
says” instead of really understanding what a product must do and who will use it.
In the 1990s, however, I have come to believe that it is critical for marketing to be
involved in the formationof a product from the very beginning in order to increase the
probability of successin the marketplace. In today’s market, competition is guaranteed
to be greater than in the past, products are steadily becoming more complex, and any
new product must be both right and substantiallybetter than average from the outset.
All these factors argue in favor of a strong marketing effort. I still finnly believe,
however, that engineers must define the product and take responsibility for its efficacy.
The preceding definitionof the engineeringfunction‘s responsibility almost com-
pletely overlaps Davidow’s (1986) definition of the strategic principle of marketing:
”Marketingmust invent complete products and drive them to commanding positions
in defensiblemarket segments.”If the engineeringorganizationlacks an understanding
of the product and its applications and fails to make a commitment to meeting the
requirementsof real users, the product is almost certain to grow in an unlimited fashion
in response to “marketing input.”
TheCEOisresponsibleforresolvingtheinevitableconflictbetweenwhat engineering
can build in the small amount of time available and what marketing believes will sell.
In reality, the product’s goodness (i.e.,its effectiveness,uniqueness, and quality)is
probably the determining factor in a high-tech venture’s success. A company with a
really poor product is most likely doomed, regardless of how great its marketing and
sales efforts might be. Given a better-than-averageproduct, a firm with an outstanding
sales force may do well despite a lack of marketing. Having a great product, a driven
sales force, and wonderful marketing is the ideal and should be strived for, although
new ventures almost never live up to the ideal, in my experience.Unfortunately,most
start-ups don’t come out with great products at first, and it takes them longer than
anticipatedto build a sales organizationand learn to sell the product. To put marketing
into proper perspective, it could be said that poor marketing can be the number two
company killer?
It is often difficult to separate the concept of the product from other equally
important views of the market, including:

The product itself that is often characterized as the market+.g., the disk or
spreadsheet markets

Customerswhouse and/orbuy the product-.g., thegovernment,academic,home,

small-business, and large-company markets

2. I believe that a poor CEO is the number one killer of start-up companies. Venture capitalist John
Shoch rates lack of team as the number one killer.
 Marketing 203

Various ways of distributing the product, called the channels of distribution-

e.g., the distributor, OEM (originalequipment manufacturer), and VAR (value-
added reseller) markets

It is criticalto understand that if a product is to be useful for a professional or other

information-processingsystem,.it must be complete.That is, it must include every level
of integrationbetweena computingplatform(hardware)and any necessary"application"
software-in short, all the levels of integration that are required for the product's final
application in the target markets. In order to achieve completeness, the channel of
distribution must often participate "actively" in the product development process by
supplyingapplicationsoftware.In many cases, the applicationsoftwareis so important
that a program such as Autocatl may, by definition, be the complete product from the
user's perspective. The hardware platform on which it runs is regarded as a mere
component!
The creation and distribution of generic computers with generic software and
profession-specificapplications,programs must be clearlyunderstoodby all companies
involved in the distribution chain.

THE SIX BIG QUESTIONS THAT GUIDE MARKETING

This subsection of the chapter focuses on the mechanics of marketing and is

organized around the following six questions that the start-up must answer:

1. What is the product, and is it complete and ready for use by the potential
customers?

2. Who will buy the product?

3. How will the product be used-Le., for what application?

4. Why will customers buy the product, in terms of its features, functions, and
benefits?

5. Where will the product be sold-i.e., through what distribution channels?

6. When will orders be received and filled-i.e., how long will the process take?

The product itself (question1)is primarily the domain of engineering, as discussed

in Chapter 8, but marketing shares the responsibility for the answer. I d e n w n g the
customer groups initially (question 2) is a marketing responsibility of paramount
importancefor successfullydesigning the product and then selling it. How the product
204 Marketing and Sales

will be used and why customers are likely to buy it (questions 3 and 4) are really
fundamental and lie at the heart of market planning. Where the product will be sold
(question5) is the responsibility of both sales and marketing, and how long it will take
to get orders (question 6) is the domain of the selling organization, but with strong
marketing involvement.

Question 1: What Is the Product?

Chapter 8 was devoted to an extensive discussion of the product. However, a few

additional comments about the marketing-related aspects of this topic will be found
in the following subsections.

Question 2 Who Will Buy the Product?

More than half the US. population is engaged in service industries that have
information technology as their base. The Computer and Business Equipment
Manufacturers Association projects that, by the year 2000, the worldwide revenue
of the U.S. computer and business-equipment industry will almost triple from
1990’s level, to reach over $700 billion. The computer industry consists of roughly
one-half equipment, one-third software and services, and one-sixth business
equipment and forms. In addition, the telecommunications industry is becoming
closely related to the computer industry and is about half its size. Thus, the
combined industries are projected at over $1 trillion.
Harvard University’s Program on Information Resources Policy has mapped the
more than eighty serviceand product industriesthat composethe informationbusiness
(Figure9-1).The computer industry includescomputers,software, modems, terminals,
time-sharing, and service bureaus. Other industries that make up the information
business include telecommunications, communication (radio, TV,newspapers), and
almost all consumer electronics.Although it does not appear on the map, computers
alsoplay a key supportingrole in many largeserviceindustries,includingfinancial(e.g.,
banking, stock markets, and insurance)and travel, where informationrepresentstoken
transactionsinvolvingmoney, stocks,futurerisk,and futuretransportation,respectively.
The map was designed to show the historical evolution of the industry and its
companies,the effect of governmentregulation,strategicpositioning of companies,and
the ability of companies to migrate from products to services.Since the computer is the
enabling technologyfor virtuallyevery product and servicein the informationbusiness,
the map’s purpose in the present contextis to showthe diversityof productsand services
witlun this business and, hence, the scope of opportunitiesfor new high-tech ventures.
Figure 9-2 shows International Data Corporation’s estimate of the number of
worldwide information-technologyusers versus time. Note that the time periods are in
 GOVT MAIL MAILGRAM INTERNATLTEL SVCS VANS BROADCASTNETWORKS DATABASES AND PROFESSIONALSVCS
PARCELSVCS TELEX LONG DlST TEL SVCS BROADCASTSTATIONS VIDEOTEX
COURIER SVCS EMS LDCAL TEL SVCS DBS CABLE NETWORKS NEWS SVCS FINANCIALSVCS
OTHER DELIVERY CABLE OPERATORS ADVERTISING SVCS
svcs TELETEXT
MULTIPOINTDISTRIBUTION SVCS
DIGITAL TERMINATION SVCS
MOBILE SVCS FM SUBCARRIERS
PRINTINGCOS PAGING SVCS BILLINGAND TIME-SHARING SERVICE BUREAUS
LIBRARIES METERING SVCS ON-LINE DIRECTORIES
MULTIPLEXINGSVCS
SOFTWARE SVCS
RETAILERS BUM TRANSMISSIONSVCS
NEWSSTANDS INDUSTRY NETWORKS SYNDICATORS AND
PROGRAM PACKAGERS
LOOSE-LEAFSVCS
DEFENSE TELECOM SYSTEMS

SECURITV SVCS

css svcs

PAEXS

SOFWARE PACKAGES
RADIOS DIRECTORIES
TELEPHONE SWITCHING E W I P
Tv SETS NEWSPAPERS
PRINTING AND TELEPHONES MODEMS
GRAPHICS E W I P TERMINALS CONCENTRATORS NEWSLETrERS
COPIERS PRINTERS MULTIPLEXERS MAGAZINES
FACSIMILE
ATMs
CASH REGISTERS POS E W I P SHOPPERS
BROADCAST AND
INSTRUMENTS TRANSMISSION M U l P AUDIO RECORDS
TVPEWRITERS WORD PROCESSORS ANDTAPES
DICTATIONE W I P VIDEO TAPE RECORDERS
BLANK TAPE PHONOS. VIDEO DISC PLAYERS FILMS AND
AND F I N VIDEO PROGRAMS

CALCULATORS

FILE CABINETS MICROFILM, MICROFICHE

PAPER BUSINESS FORMS GREETINGCARDS BOOKS

+ FORM ATM - Aumatffi teller machine DBS - Dlrect broadcastsatellite W S - Point-ol-sale SUBSTANCE -b
COS - Cwnpantes EMS - Elecbonic message servtm svcs - BMC8.3
CSS - Carrler 'srnarr switch PABX. Private autornaoc branch exchange VAN - Valueadded network

Figure 9-1.A Map of the Information Business. (Courtesy of the Program on Information Resources Policy,
Harvard University. John F. McLaughlin and Ann Louise Antonoff, Mapping the Information Business.
Reprinted with Permission.)
206 Marketing and Sales

1m
1. Number of Information
Technology users
versus time
2
m L
bJg
2.u-
4:

1
1950 1957 1964 1971 1978 1985 1992

~o.ofusers
Figure 9-2. Estimate of the Number of Information-TechnologyUsers Versus Time.
(Courtesy of International Data Corporation. Reprinted with permission.)

seven-year increments,whereas the computer generationshave occurred roughly each

decade. By the late 1970s, terminal-based, centralized computing was limited to less
than 10 million users. With the advent of PCs, the number increased about an order of
magnitude. The 1992estimateis that there will exist a ubiquitous network with over 300
millionusers.Thisis probablyslightlyoptimistic,due to the lack of good internetworking,
but may occur by the year 2000.

Market Segmentation-The Key to Determining Who Will Buy. The most im-
portant part of market segmentation begins with the identification of potential
customers who might buy and use a particular product. Ideally, in performing this
segmentation, the company comes to understand precisely who the buyers/users
are-including their educational background, demographics, buying motivation
and patterns, etc.-so that a product can be designed to really meet their needs.
 Marketing 207

Three different segmentation schemes are helpful in pinpointing the users:

1. The buying organization-e.g., a collection of companies, a governmental

group, a household, an educational institution

2. The function of the user's department or organization-eg, finance, engi-

neering, manufacturing-including its physical environment

3. The user's profession--e.g., secretary, VLSI (very-large-scale integration)

design engineer, actuary

The first segmentation scheme simply identifies which doors to knock on; the second
identdies the relevant departmentswithin the specifiedorganizationsthat are potential
buyers; and the third identifies the characteristics of the final and actual users.
The variety of alternative product solutions,from centralized mainframes to fully
distributed personal computers, makes customer segmentation quite subtle, because
oftentimes,the user is not necessarilythe buyer. The only thing that is clear about buyer
segmentationis that, for a particular profession-based application,it is usually possible
to idenhfy the individualb) within an organization who will ultimately use a given
product. This person (or group),.which is referred to as a customer or user performing
a given application,is the only place to start in constructinga market.
What's unclear is whether such individualshave any influence over the purchase
of what they use. Frequently, a strong central service determines the computing
environment for an entire corporation. In more enlightened large organizations, this
absolute power is moderated to take into account the needs of the actual users. For
example, with G M s purchase of EDS to be the computing czar for the corporation,it is
difficult to find anyone, including GM Research (where maximum freedom should
exist), who claims responsibility for how computers are purchased or used. Thus, a
buyer may be an individual engineer or financial analyst who needs the product or a
group entrusted with the respoi~sibility for finding and acquiring tools for others.
In short, it is essential to keep in mind that computer use and computer users are
often segmented from the purchase and operation of computing. Hence, all facets of
buying,using, and operationmust be understoodin order for marketingto be successful
in any organization, including governments, industry, academe, and the home.

Segmentation by the Buying Organization-The SIC Code. The simplest seg-

mentation is by Standard Industrial Classification (SIC) code-that is, how the or-
ganization is classified. (Figure9-3 shows the categoriesemployed in the SIC coding
system.) The SIC code can be employed to identify companies and general trends
that may be helpful in marketing, but it does not begin to identify specific users of
208 Marketing and Sales

Agriculture, forestry, and fishing

Mining, including petroleum
Construction
Manufacturing (discrete,continuous, drug, etc.)
Transportation (air,rail, highway, pipeline, communication, radio, TV)
Wholesale trade
Retail trade
Finance (banks and brokerages),insurance, and real estate
Service, including medical institutions and universities
Public administration-i.e., all levels of government, including the military

Figure 9-3. Standard Industrial Classification Code Groups.

a computing product because the application of computers is more closely related

to profession (e.g.,cost accountant, computational fluid dynamicist) than to orga-
nization.

Segmentationby the User’s Organization. A taxonomy of corporate users is given

in Figure 9-4. Although identifying the tasks of various departments may be only
marginally useful in segmenting the market, it does help the start-up determine the
size and importance of each department within a larger organization.

Segmentation by the User’s Profession. The best segmentation,by the profession

of the individual who will be using the computer, can be obtained by examining the
structure of academic disciplines within a modern university that aligns itself to
solve technologicalproblems. Figure 9-5 shows a partial list of academicdisciplines,
together with the applications that each one often requires. Since many of the
disciplines have mathematics as their base language, generic programs that un-
derstand mathematicsare often used across many of the disciplines.Similarly,word
processing for text and spreadsheets for numbers, along with electronic mail, are the
tools for all communication within an organization.
The balance of this subsection contains several important observationsabout how
computer applications form in response both to generic tasks, such as preparing
 Marketing 209

Commercial department: Includes users who do financial accounting and control,

billing, inventory, accounts receivable/payable, payroll, transaction processing, and
business analysis.

Technical department (science and engineering): Includes users who deal with
numbers; algorithms; text; graphs; data acquisition; real-time control; simulation;
product and process design, including CAM (computer-aided manufacturing); and
communication.

Manufacturingdepartmeni’:Includes users who do record storage and processingfor

inventory, continuous and discrete real-time control, and plant scheduling and
process optimization.

Communicationsandcomputation (managementinformationsystems): Includesusers

who manage or use message switching, electronic mail, voice-messaging systems,
teleconferencing, centralized computers, and computer networking.

Office automation, electronic (desktop) publishing, and word processing: Includes

anyone involved in image processing for the transduction, storage, and transmission
of documents.

Education: Includes teachers, students, course planners, administrators, and others

involved in training and computer-assisted instruction.

Figure 9-4. Market Segmentation by Department.

documents, and to those tasks required by users in specific professions, such as

petroleum engineers or actuaries.
Given the large number of quite big software companies (e.g.,Computer Associ-
ates, Lotus, Microsoft, and Oracle) that supply generic applicationprograms for word
processing,spreadsheets,datakiases,etc.,there existsonly a relativelysmallopportunity
for new generic programs except for programs predicated on a new, broad-based
programming paradigm, such as the fundamental understanding of mathematics.3
In terms of the professional disciplines, the largest opportunity for start-ups is for
enumerated applicationsbased on a potential market of at least every person (or other
information-processingsystem)who would use computers to supplementor supplant
his or her own activities in information processing.
Computer manufacturers exist to design, build, and sell computers. Applying
computers to the innumerable generic and profession-specificapplicationsrequires an

3. This market, created in the 1980s, is well on its way to becoming established.
210 Marketing and Sales

Arts and humanities:

Animation
Computer-assisted art, including sculpture
Music

Business, economics, and finance:

See Figure 9-4, “Market Segmentation by Department.”

Computer and computational science

Engineering:
Aeronautical and astronautical
Biotechnology of all kinds
Chemical
Civil and structural
Computer
Electric power
Electronic, including digital systems, signal, and image processing
Environmental
Manufacturing associated with various industries and/or technologies
Mechanical
Nuclear
Petroleum

Mathematics:
Linear and nonlinear analysis
Numerical analysis
Statistics
Linear, nonlinear, dynamic programming, operations research

Medical and biological science (and biotechnology, shared with engineering)

Science:
Astronomy
Atmospheric
Astrophysics
Chemistry
Geology and geophysics
Oceanography
Physics
Acoustics
Computational fluid dynamics
High-energy physics
Psychology and social science

Figure 9-5. Academic (Intellectual) Disciplines and the Applications They Require.
 Marketing 211

equally large number of professionals who understand both computers and the
particular professions. Generic (andprofession-based applications lie outside the pur-
view of computer manufactureirs simply because the plethora of skilled professionals
who might use computers usually have nothing in common with those who manufac-
ture computers.
In the unllkelyevent that an applicationsprogram is developed within a computer-
manufacturing organization,the likelihood of successfulexploitation of the program is
very small, unless the applications organization can be separated from the manu-
facturing organization. That is, producing hardware and developing applications
software is virtually impossible under the same corporate "roof." If a hardware
company has a significant software product that could be a potential "standard," the
best way to exploit that standard is to spin it off into another firm.
The emergence of new computer classes has created opportunitiesfor start-ups to
formin order to developapplicationsprogramsbased on the new computingparadigms.
Because existing firms operate within the parameters of their established corporate
cultures, costs, and customers, manufacturers in one computer class have usually been
unsuccessful in entering a new (class.Similarly, suppliers of applications have usually
been unsuccessfulin adaptingexistingprograms to a new computingenvironment(e.g.,
translatingbatch-orientedmainframesto interactivevisualizingworkstations).Autodesk
is an example of a high-tech venture that grew to dominatea large mechanicaldesigner
market using the PC, despite the existence of decade-old companies such as
Computervision that used the minicomputer. In electrical computer-aided design
(CAD),Daisy, Mentor, and Valid started up based on the workstation.As PCs evolved,
new firms entered the market a s challengers.

The Industry X Profession Table(s)-Locating the Customers. Given that most

products serve professionals by making them more productive in some way, the
start-up must locate these potential users precisely within their organizations. If the
start-up has a broad range of products, then a good strategy to minimize its
marketing, selling,and support costs is to build products designed for use within the
same organization but by other, related professionals. For example, selling pro-
grams to cost accountants in a rnanufacturing organization might lead the company
into materials-requirements-planning programs.
Constructing a basic, but exhaustive table of the professions (and subprofessions)
within each relevant industry segment is the best starting point for identifymg all the
potential customers for a product. Many different tables can then be generated to
analyze and address all the aspects of the potential market, segmented by customer
groups. A sample of this forma.t, giving the industry and prospective customers Coy
212 Marketing and Sales

Industry Industry m
(e.g.,aerospace) ...
Profession 1
(e.g., aerospace
engineer)

profession and subprofession),is shown in Figure9-6.Somecommon tables of data that

must be prepared in order to fully understand the characteristics of each market
segment include:

Demographics: Number of professionals within each industry (SIC)group

Key customers: For product introduction, feedback,and setting priorities

Ability to buy: Budgets and resources for computing

Product needs: Product (e.g., software) that is relevant for each group
Communication forums: Shows, magazines, and professional societies

Sales channels: How each group buys-eg., VAR, manufacturer, ISV (inde-
pendent software vendor), system integrator

Note that alternativetables using this format (e.g.,locationX profession)are helpful

for segmenting customers within a geographic region instead of by the SIC code.

Question 3: How Will the Product Be Used?

In addition to defining the product itself, ensuring that it is complete for use by the
intended customers, and identifying potential buyers/users, the start-up must
 Marketing 213

determine the actual applicaticln for the product. In the case of most professionals,
the product is used to enhance their ability to process information. Thus, implicit in
the product is an underlying assumption about its application. For example, let’s
look at the evolution of features in a generic word-processing program and see some
of the different ways in which various users might apply such a program:

Text only: Letters, memos, office communication

Tables: Simple reports-e.g., for expenses

List processing: Forms and mailing-list processing

Indexes, etc.: Producing reports, manuals, books
Scientific and math notation: ‘Technical documents

High-quality text layout: Desktop publishing for brochures

It is also important to note that spreadsheet programs like Excel or Lotus 1-2-3are
used for innumerable applications beyond their intended function, such as making
slides or generating relational databases and reports. In addition, of course, there exists
a plethora of templates to transform a spreadsheet into a generic business-planning
document, for doing profit and loss statements and performing other accounting
functions.
The concept of the level of integrationplays an importantrole in a product. It might
be expected that, as the product’s level of integrationdecreases,the number of products
sold would increase.However, the number of buyers/users is not at all correlated with
the level of integration.The following levels of integration and the customerswho buy
products at each level illustrate the point:

Platform--e.g., language, operating system: systems programmers and builders

Mail, word processing: every user

Application program X : professionals needing X

At thelowestleve1,systemsprogrammersusethe platformtobuildtheirownapplications,
but this is a relatively small class of users. At the application level for a series of
professions, the size of the professions and the functionality of the product (i.e.,what it
can do for the user) determine the market size. Thus, although a given application
program used by a particular sei:of professionals may be sold in very low volume, the
collection of all the professional applicationsthat share a common platform is usually
quite large.
214 Marketing and Sales

Customer/Application Profile Table. The main tool for understanding how the
product is to be used is a detailed customer/application profile (CAP) for a set of
possible products. This takes the form of a table of all the customers (derived from
the organization table) versus the applications for the products. Table 9-1 provides
a qualitative view of the size of the overall chemistry market for application
programs (the rows) broken down into four markets: pharmaceutical chemicals
(e.g., medicinal and agricultural), biotechnology (e.g., protein engineering), poly-
mers (e.g., films and plastics),and materials (e.g., surfaces and composites).Within
each industry, theoretical, synthetic, and analytic chemists constitute three kinds of
professional users. There are different classes of application program categories. If
we proceed further, each application domain must list the programs that are critical
for each of the professionals. In some cases, the same program might be used by a
given professional in each industry, whereas other programs are specific to a
particular industry. Eventually, the table has to be filled in with guesses about the
actual market sizes.

The Role of Influential Users and Early Adopters. Selectingthe earliest users, and
especially the beta-site customers, is one of the most important decisions that a
start-up can make. Ideally, the company is able to persuade the most influential
members of an intellectual community to test and embrace its product. Once this
elite is won over, then early adopters in large corporations start to buy, followed by
the beginnings of a “mass market.”
Universities are the most important customers that a high-technologyventure can
have. Putting computers into academe guarantees that the brightest, hardest-working,
and most motivatedpeople will fearlesslytest the product and give an honest, no-holds-
barred opinion.Therefore,universities should be thefirst beta-test sites for all new computers
and applications.
In the earliest days of computing, IBM gave its largest computers to prestigious
universitiesto stimulateearly computer use. It slacked off in the 1960s,1970s,and 1980s
because it did not feel the need for or forgot the value of university interaction and
training of next generation users . By going back to academic users, IBM gets a true
competitive picture of its products. Apple has a very strong university gifts program,
and the whole NeXT venture was predicated on serving the university market.

DECand universities: When DEC was just founded and barely profitable, it gave one
of its first PDP-1s to MIT. The researchers (i.e., the faculty and graduate students)
used the machine to generate much of the early software, such as editors, compilers,
debuggers, operating systems, and applications programs.
In 1978,when the VAX 11/780 was introduced, John Pople, who has been a leader
in computational chemistry, got serial number 1 in order to test VAX’s efficacy for
 Marketing 215

Table 9-1. A Computational Clhemistry Customer/ApplicationProfile.

Application Industry
Pharmaceutical Biotechnology Polymers Materials
364; 18% 22 % 24%
Mode1ing s l m l s s m m s m m s
Molecular static 1 1 1 ms s l s s 1 ms
Molecular dynamics v m s m s s l s s mms
48 %

Semiempirical v m s ms s v s s v l s
Ab initio v m s --- v s s v ms
36%

X ray s s l mmv s m v s m l
Nuclear magnetic s 111 1 s l v s m v s s m
resonance (NMR)
Instrument control -m v -1 v -ml -s m
Sequenceanalysis - __ - 1 m l --- ---
Database s l 1 v l v s ms -s s
16%

Notes: v = very large; 1 = large; s = small; m = medium; - = nonexistent. For each application, a
group of three estimated values is shown within each industry. The first value in each group is for
the theoretical chemists in that industry; the second is for the synthetic chemists; and the third for the
analytic chemists.

chemistry codes. After he found that the machine worked better than all the IBM and
Univac mainframes he was then using, Pople told his friends, and VAX became the
standard for departmental and project-level technical computing until 1985. Further-
more, even in 1990, the VAX 11/780 is the unity benchmark against which all technical
computers are compared. About ten thousand 11/780s were delivered into this
technical community.
Through the years, DEC provided much hardware for research, includingthe VAX
11/750s used by Bill Joy to do the UNIX Berkeley extensions that were the basis of both
DEC's and Sun's versions of UNIX. Collaboration with Carnegie-Mellon University
216 Marketing and Sales

generated someof theearliestmultiprocessors.In thelate 1980s,DECand IBMprovided

hardware to MIT as part of Project Athena, which was the co-originator (with DEC) of
the X Window standard.

Question 4 Why Will CustomersBuy the Product?

The start-up must fundamentally understand the buying rationale; otherwise, it is

unlikely to be successful in convincing users to buy its product. Chapter 8 described
various categories of product-buying criteria, and this chapter will develop some of
those criteria further.

Features, Functions, and Benefits (FFB) Lists and Competitive Tables. The best
way to understand the buying criteria is to start by creating an exhaustive list of the
product's features, functions, and benefits, including its price and performance
characteristics. Independent of whether the product is completely novel, the FFB list
is the fundamental basis for all product and market analysis.
The most common use for the FFl3 list is in creating a table that compares the start-
up's product with the nearest competitor's products or collection of products. Often, a
simplified version of such a table is the basis for an advertisement for the product. For
example, IBM once ran an un-IBM-style ad showing a table that compared its laser
printer with that of the market leader, Hewlett-Packard, in terms of price, performance
(speed),and features (numberof fonts, ability to feed envelopes, etc.).Of course, theIBM
printer won in every category shown in the table. I love these ads.

The ApplicationsVersus FFB Table. The second use for the FFB list is in comparing
the requirements for the product with the applications or customer needs. Table
9-2 shows the basic functions required on a hardware platform: scalar processing,
vector processing, computer graphics, and image processing. The first row of the
table, labeled "Product," is an honest analysis of how well the proposed product
does on each of these functions.
The applications listed serve as a "filtef' when testing the product's suitability to
carry out the application. Regular typeface has been used in the body of the table to
indicate a good fit between an application and the base product. Underlining indicates
a product weakness, andboldfacetypeindicatesacompetitiveadvantage. For example,
not having very high graphicsperformance may make a product deficient for modeling.
The comparative weakness of not having the highest scalar speed is offset by vector
performance in two of the applicationclasses. Having high vector performance is useful
in five of the application classes.
 Marketing 217

Table 9-2. Base Product Requirementsfor Each ApplicationsArea.

Scalar Vectors Graphics lmage
Product 1 V 1 1

Applications
Modeling m S -
V S

Molecular static 1 m 1 S

Molecular dynamic 1 V m S

Semiempirical 1 1 m 1
Ab initio -
V V m 1
X ray v V 1 1
NMR 1 m 1 S

Instrument control m m 1 S

Sequence analysis 1 V m S

Database 1 V S S

Notes. v = very large; 1 = large (or high); m = medium; s = small (or low). Underlining indicates
weakness, and bold indicates significant strength.

Question 5 Where Will the Product Be Sold?

In the early days of computing, the computer manufacturer supplied broad-based

application programs or programming templates for an organization, such as the
corporate-accountingand general-ledgerprograms. In other cases, large companies
(such as GM) wrote many of their own programs for specific profession-based
applications, ranging from accounting to computer-aided design.
With time, computer manufacturers have produced fewer programs for generic
and profession-specific applications, principally because the manufacturers are ori-
ented toward building various classes of computers for as large an audienceas possible.
It would take an enormous staiff, in addition to computer engineers and marketers, to
produce the range of necessary applications.Since, as noted earlier, a single company
218 Marketing and Sales

is unlikely to be able to do more than one thing well, applicationsprograms had to be

produced outside of computer-manufacturingfirms.
Thus, creating a ”complete” product requires a large collection of application
programs (i.e.,more component products). In this respect, the distribution channels for
high-techproducts are unlike the channelsfor other productsbecause they are “active,”
providing added value (includingservice)so that the final buyer/user can perform the
intended function.Of course,thelast stageinany distributionchannelisthebuyer/user,
who ultimately integrates all the components and becomes knowledgeableabout the
system.
Figure9-7, which showsthe channels of product developmentand/or distribution,
can be used to characterize the very broad industrial structure of building products
within the componentand systemframeworkanddeliveringausefulproductto theend
user. One of the least understoodand most underappreciated aspectsof marketingis the
”active-channels”concept, wluch produces applicationsprograms and, in effect, ”fin-
ishes” the product, making it suitablefor use by an actualbuyer. Without this finishing,
whch is accomplished by adding an applications program (a program that is in fact
often more complex than the underlying base system),the product is entirelyuseless to
a buyer. The levels-of-integrationdimension of the computer product space (discussed
in Chapter 8) provides a map of the various levels required to form the complete
product.
In some cases, products require specialization beyond that of the professional
target, suchasthelogicdesignerortaxaccountant,andtheuser may requireconsultants.
Specializationis usually done by the distributing organization or user‘s organization.
Although users do customize certain products, it is becoming a rarity (in terms of both
percentages and actual numbers) for users to write their own applications programs,
because the existence of massive quantities of PCs provides a common programming
environment and a very large target market for a relatively small team of professionals
(sometimesconsisting of no more than one person) to write a successful profession-
specific application. For instance, if software written by a single programmer can be
priced at $50 per copy, selling only a hundred thousand copies per year will generate
$5 million in revenue.
Figure 9-7 shows various alternatives for distribution and applications-
development channels, according to whether each channel is passive (involving
distribution only) or active (providingadded value). Passive distribution implies that
a single manufacturer takes responsibility for a complete product, with all the levels of
integration required for successfuluse.
In the direct-sales and distributor-based channels, pricing and product respon-
sibilityrest solelywith the start-up.In caseswhere another companyis activelyinvolved
in the product’s development, pricing and product responsibility are less clear. Simi-
larly, the distribution of costs and profit is often unclear when two firms share
 Marketing 219

Direct sales: Passive development in the distribution channel.

Direct sales: The manufacturerasells the product with its own sales force or
through company stores
Direct sales, with resale of application products from other suppliers.

Mail order and telemarketing: End users buy without salesperson contact.
Users buy direct from the manufacturer.
Mail-order house buys the product from the manufacturer for sale/distribution.

Distributors: The link between hardware/software manufacturers and end users.

Component distributors: Direct or mail-order component sales to companies or
individuals for use in building larger systems.
Resellers (dealers), including leasing and installation: Aform of distribution whereby
another, usually “geographically local,” company buys equipment from one or more
manufacturers for resale. Such a form is potentially unstable because the manufac-
turer is likely to take over the distribution.
Wholesalers: A stage for distributing goods to retail stores.
Warehouse sales: Distributionthat eliminates one stage in the wholesale/retail chain
by having customers go to more central locations rather than to local retail stores.
Retail sales (dealers): A final stage of distribution to reach an end user with a
combination of product, training, and service.

Brand relabeling for resale by another company: Often, a traditional computer

company distributes under its own label a product that has been manufactured in
whole or in part by another organization. This is also erroneously called an OEM
(original equipment manufacturer) relationship. For example, in this relationship,
either firm may assemble and test the hardware and software components that form
the system.

Value-addedchannel with “activeproduct deve/oipment”anddistribution: A third

party performs a major portion o i the development or support to form a complete
product that is ready for the end user.
Third-party developer or independent software vendor: A company with an applica-
tion program for a particular application market segment to be distributed either via
the equipment manufacturer or as an OEM, VAR, system integrator, jointly with the
manufacturer, or as a manufacturer.
Original equipment manufacturer: Basic equipment is supplied to another manufac-
turer, which adds hardware and software to form a complete application for a
particular, usually narrower, use. The purchased component is generally a small
fraction of the complete system.
Value-added reseller: A reseller that adds something of value (e.g., advice and
training, customization of particular software, or unique software for a profession)
and resells the completed system. Typically, a VAR designs and produces unique
application software such as CAD/CAM (computer-aided design/computer-aided
manufacturing).
System integrator: A company that builds products, using any necessary hardware
(continued)
220 Marketing and Sales

(continued)

and software systems, to meet a particular buyer’s requirements. System integrators

usually serve special government and military computing needs. In some cases, the
system integrators simply supply the product and its required papetwork to satisfy
the government user in a completely parasitic fashion, thereby increasing the price
with little or no attendant benefit, except full employment.

Joint marketing of equipment and applications products: For highly technical or

complex software, the seller of a hardware system and the developer of a critical
application product (an ISV) jointly market and sell the resulting product.

“The term manufacturer refers to the producer of a hardware component, a computer

system, or a software product.

Figure 9-7. Channels of Product Development and /or Distribution.

responsibility(cost)for product efficacy, sales,and product support. Jointrelationships

in which two organizations share such responsibilities are inherently unstable, unless
they are managed very skillfully by both parties.
In the case of an “active” distribution channel, the principal activity is that of
application development, producing either a generic or profession-specificsystem that
utilizes one or more basic hardware platforms. Thus, what at first glance appears to be
a channel of distributionis really the dominant system supplier for a product. In effect,
an application-developmentcompany, or third-party application developer, could use
any hardware, but the product that it produces is unique and based on the firm’s
knowledge of a specific application field that aids a professional in a given discipline.
To the user, the application is the product, and the hardware platform is irrelevant.

The DEC Market Map-The Many Paths from Product to Customers. The
challenge in creating a product is to invent and assemble a collection of components
that, used together, will solve a particular customer problem. Figure 9-8 is a flow
graph that illustrates the many pathsby whicha large hardware company’sproduct
can find its way to the end user. The example shown is from Digital Equipment
Corporation (DEC), as it was organized circa 1982 to address a large number of
different and varied markets, products, and channels of distribution. In order for a
product to reach the market in a ”complete”form, it must make a full circuit through
all the levels of the graph. For example, a simple path is user-written, tailored
applications, using base hardware and software from DEC. The computer could be
operated or serviced by any of three a1ternative organizations: DEC, thc buyer, or a
third party.
 Marketing 221

Products and Third-party applications developers, Final user:

services service, hardware OEMs, system geography,
supplied by I integrators, VARs, etc. I profession and
Digital I organization
I
I Use
I
Market I
and sales
channels
Operate,
service service

Digital product line marketing and sales

channel or alternative distribution channel

Figure 9-8. Product-Distributionand Final Development Paths for Digital Equipment

Corporation (Circa 1082).

In the 1980s, it was comrnon for applications products to be developed by a

company and then resold as a system using DEC's hardware. In this fashion, the
reselling firm was able to chargea percentage on the hardware system as well as charge
for its software. Of course, customers, the software company, and DEC were all
unhappy with the situationbecause responsibilitywas unclear and there were too many
markups in the chain. By the end of the 1980s, when no single hardware vendor
222 Marketing and Sales

appeared to be competitive for all time, application-specific product organizations

stopped being involved in the distribution of hardware.
The DEC map is useful, and even essential, in identifyingthe multiplicity of paths
by which a complete product may be sold. Due to this multiplicity of paths, a “channels
conflict” can occur when two or more sellers appear on a customer’sdoorstep with the
identical product but with different deals. Usually, one seller is the original manufac-
turer, and the other is a VAR reselling the same manufacturer’s product to take
advantage of a quantity discount, but with an applications software program that it
bought from a software supplier. A large user is often able to purchase the hardware
more cheaply directly from DEC and buy the applications software from the same
supplier that the VAR uses.
Although the map is complex, DEC was most successfulbecause it maximized the
number of distribution channels, even though doing so made life more difficult for the
selling organization. In 1983, DEC reorganized and began to eliminate some of the
channel conflict by eliminating various distributors that it thought did not add value.
In the process, it no doubt eliminatedsome channels that were adding value. By the late
1980s, DEC‘s revenue began to decline, and in 1990, DEC was unprofitable for the first
time in its history.Meanwhile,IBM and Sun Microsystemsadopted the DEC-pioneered
multiple-channelsapproach to obtain applicationssoftwareand to resell computers.In
1990, both IBM and Sun have grown. DEC is currently reinstalling a market structure
not unlike the one it had destroyed by 1985.

Knowledge-EngineeringMarket Map. In order to examine the “expert-system”

or ”knowledge-engineering”product world, it is necessary to start with the hard-
ware platform and work all the way up through generic and profession-based
applications to arrive at a product that can be used by a professional to carry out a
particular task. Figure 9-9 shows the various levels of integration, beginning with a
basic hardware platform such as a mainframe, workstation, or personal computer
that uses traditional operating systems and programming languages. More modern
systems, such as workstations, have a human-interfacelayer for controllingmultiple
processes via windows. The LISP language environment hosts generic knowledge-
engineering tools, such as ART or KEE. In the early 1990s, dozens of “expert-system
shells” based on unique and proprietary languages, or based on extensions to the
LISP language, allow users to write a specific expert system. A database environ-
ment usually exists to hold the knowledge and database for the application.
The following are examples of some of the functions that expert systems can be
structured to perform:

Advice: A system that helps determine whether a particular customer should be

given credit is in use at American Express.
 Marketing 223

engineering
language/
environment

UNIX DEC VMS APPLE MAC IBM PC DOS, OS/2 IBM VM, MVS Hardwarel
operating-
system
platforms

Figure 9-9. Market-Map Template for Knowledge-Engineering Products.

Diagnosis: A diagnosticsystem used by GM aids in pinpointing faults in a car that

result in various noises. Medical-diagnosis systems help determine whether a
patient has a disease, given his or her history, observations, and test results.

Configuration design: The DEC configuration programs enable sales and manu-
facturing personnel to put together a plan for building a particular model of a
computer, such as a VA?i:/9010, to meet various customer requirements.
224 Marketing and Sales

True system design: Design Power sells a program for designing steam plants that
includes electrical, structural, and mechanical equipment. The program takes
customer requirements and produces a physical plant design, which is repre-
sented as specifications of equipment and 3-D drawings.

A program can be constructedbased on certain needs to look ahead or to work with

imprecise data. Such a program can often be employed by users who work in different
portions of an industry but have similar application requirements. For example, a
drilling-adviserprogram may advisepetroleum engineersabout the geologic subtleties
found while drillingfor oil. Theseengineersand/or geologistsare within the petroleum
industry segment. Similarly, within the consulting or service segment are petroleum
engineers who would also use such a program. However, in a case such as this, the
buying patterns, use, and SIC codes of the various groups of professionals are often
radically different.It is critical for the start-up to build and understand the market map
before fully designing the product.
In 1990,the &ant (lessthan$lOO million) ”knowledgeindustry”suppliesproprietary
shells,applicationtool kits, and consultingto companiesthat need to build special tools
to tackle problems that can only be solved using the knowledge-engineeringapproach.
By 2001, this industry will be completely absorbed by a new and traditional software
industrythat uses ”expert-systems”technologyto provide toolsfor structuralengineers,
tax and investmentadvisers,and other professionals.The ”expert-system-shell”business
will become commoditylikecompilers that implement a few standard language shells.

Being Global. Perhaps the most important decision a new venture faces in
distributing its product is whether to attempt a global marketing strategy or simply
find foreign distributors as the firm evolves.JimMorgan, CEO of Applied Materials,
argues that a high-tech company must start up with the idea of being a worldwide
enterprise. If the organization does not compete globally, it will remain small, be
unable to grow and prosper, and rapidly lose market share to offshore competitors
that dominate the worldwide market. In Applied Material’s case, its largest market
is outside the United States because other suppliers dominate the U.S. market.
Morgan believes that Japan is the best place to train a company to be a global, high-
quality manufacturer. The Japanese culture teaches a firm the importance of
relationships,includingbeing closeand open with your customers and understanding
them. It also teaches patience and a concern for the long term.

Question 6 When Will Orders Be Received and Filled?

As was indicated earlier in the chapter, how long it will take for orders to be received
and filled is basically a sales question rather than a marketing question. Practically
 Marketing 225

every new venture is overly optimistic about the time that will be required to get
orders. In the beginning, all selling is ”missionaryselling” and takes a lot longer than
a start-up would initially estimate.

THE MARKETING ORGANIZATIONAND ITS FUNCTION

Both engineering and marketing employ models as part of their work in building the
company. Engineering begins by creating some sort of model that describes the
product’s structure and behavior. It then creates a working product that a manu-
facturing organization can replicate. Marketing begins by creating a model of a
marketplace for the ”model product” in terms of specific buyers and ways in which
the product can be sold. When the final product is ready, the model is tested by a
sales channel that offers the real product for sale. The validity of the market model
can only be tested once the actual product is available.
Marketing is first responsible for defining the complete product for the user,
including securing any components, such as applications software, from outside the
company.A market map is drawn to depict the myriad paths for completing and selling
the product. Finally, marketing must ensure that product-revenue projections are met
by producing dormation that a sales organization can use in convincing customersto
buy the product. Marketing is the collective mouthpiece for the firm and the guidance
system for the sales organization; it accomplishes this latter task by outlining which
customersto visit. It must createleads.Furthermore,marketingmust arm the sales force
with various s e l h g tools. Marketing’s job, in essence, is to make selling easy.
Marketing designs and implementsthe tactical plan (i.e.,the T-shirts, testimonials,
trade shows, seminars, news events, and advertising),in accordance with an evolving
market model, which it creates, .tests,and recalibrates. Recalibration is done each time
the product is presented. It begins with the concept, proceeds through product
introduction, and ends when the product is retired.

The Marketing Balance Sheet

Themarketingbalancesheet playsthe samerole in evaluatinga start-up’smarketing

as the technology balance sheet plays for engineering.The dimensions to be evaluated
include:

The marketing processes

The marketing plan

The marketing-support output (eg., literature, public relations)

226 Marketing and Sales

Tactical sales support, including lists of targeted customers

The head of marketing

The top-level marketing team

A customer and/or technical advisory board

The marketing resources, tools, and people

A system for control, with MBOs and output measured against the marketing
plan

Three of the more important dimensionsthat characterize marketing-the head of

marketing, the marketing plan, and marketing processes--are discussed in the fol-
lowing subsections.

The Head of Marketing. The head of marketing for any high-tech venture needs
to be an artist and an inventor. He or she is part wizard, part technologist, part street
fighter, and part strategist. This key individual possesses a powerful imagination
but is able to balance a checkbook,and goes through life with head continually in the
clouds but with feet planted firmly on the ground. His or her charter is to invent a
product in the minds of buyers, produce both the strategic and tactical marketing
plans, build an organization, and behave in a professional,organized fashion from
day one. The head of marketing is an idea-driven artisan who feeds on creative
opportunity and tends to think in terms of trade shows, testimonials, T-shirts, ads,
and news releases that will attract attention to the company and its product.
Marketeerswithextensiveexperienceinlarge,establishedfirmsina well-established
marketplace (e.g., minicomputer, mainframe) are generally not especially useful in a
high-information-technology start-up that plans to enter an emerging marketplace.
Marketing in large organizations is concerned with the creation of slowly evolving
productsto support a bureaucratic,arcanecompanyand its existingcustomerbase. The
process is institutionaland is executed through a fill-in-the-blanksprocess,using tactics
designed to satisfyall the various marketing-supportorganizationsand functions, such
as competitive analysis, pricing, sales training, advertising, public relations, product
testing, and product support.Marketing people with this type of background tend to be
bureaucratic and lack both imagination and fundamentalmarketing skills. With luck,
they are capable of budgeting and managing expenses, and possess good supervisory
skills.
The greatest drawback in workmg with marketing personnel experienced in
timeworn markets is not that they are likely to produce negligible output, though, but
 Marketing 227

that they tend to prevent the development of anything new or creative. At best, these
people bring a textbook, conservative approach to new products that invariably
squashes innovation and ensures that no new kinds of products reach the market. In
sharp contrastto high-information-technologymarketeers, traditionalmarketeersfilter
out product innovationthrough a series of testing phases taken directlyfrom textbooks
that deal with marketing toothpaste and soft drinks. It is ironic that many regard the
Japanese as the worlds best marketeers, because they do not have MBA programs or
formal marketing training. Instead, they spend a great deal of time living with the
potential users of a product, more time building products for the users to try,and the
most time refining the next-generation products.
What are the ideal requirementsfor the marketinghead? At one extreme,creativity
is essential if he or she is to grasp the myriad ideas for applicationsthat are expressed
during presentations to customers. Analytical ability is equally important. Thus, the
head of marketing must have a rare blend of intuitive and analytical skills. As if this
weren’t hard enough, a high-tech venture also needs someone who can develop a
marketing plan, lead a department and team, and manage according to the plan.
Although, as noted above, marketing people with extensive experience in large
organizationshave probably been corrupted beyond hope, people who have been in a
large-company environment only long enough for initial training may make a contri-
bution to a start-up. Two kinds of technicians who come from large firms could be
useful: the supervisor who can hire, plan, and manage a staff and the product manager
who can gather and synthesize product requirements and cany out the functions
associated with a successful product introduction.
A marketing person who is product-oriented may come from an engineering
background. However,this type of person may be overly bureaucraticand less creative
than candidates from other backgrounds. Lee Iacocca-an individual with an engi-
neeringbackgroundwhocannot,orwillnot,deal withthedetailand rigor of engineering
but wants to be part of the creative process of defining and building products-is a case
in point. Alternatively, a candidate coming from sales may focus strictly on sales-
organization support and never raise questions about the product or whether sales
representativesare callingon the right customers.A salesrepresentative,having to deal
with customer requests, becomes quite creative and is highly tactical, concentrating
almost exclusively on the sales process rather than the product.

Kvamme. One of the most creative and perceptive marketers in Silicon Valley, Floyd
Kvamme, entered marketing with an engineeringbackground at Fairchild Semicon-
ductor. The head of sales at Fairchild, Don Valentine, required all marketing
personnel-including Floyd, Jerry Sanders (founder of AMD), Mike Scott, Mike
Markula, and Gene Carter (key executives in the Apple start-up)-to spend at least
228 Marketing and Sales

a year selling before progressing up the marketing ladder. Floyd led National
Semiconductor’s marketing when it started up, ran its plug-compatible business,
went on to head marketing at Apple, and is a venture capitalist at Kleiner, Perkins,
Caufield, and Byers.

The Marketing Plan. Davidow (1986)differentiatesthree types of marketing plans:

for a new business, for a fundamentally new product within an existing business,
and for a new device (or the next product release with minor enhancements). By
definition, all start-up marketing plans are plans for a new business. At a later stage,
when new products or enhancements are added to the existing business, the initial
plan is updated accordingly. These subsequent marketing plans usually do not
require or receive the rigor of the initial plan, created when the company first begins
operating.
Insofar as the scope of this book is concerned, the key activity for a start-up
marketing person is constructing a marketing plan. Later, in the product and market
developmentstages,the company’sfocusistactica1,withthe administrationof marketing
processes to support selling. But in the beginning, there must be a solid marketing plan
based ontheorganization’sview of themarketplace.Themarketingplanhasthreeparts:
a market map, which describes the entire marketplace of interest; the tactics (i.e., the
expenses)for supporting the selling of the product; and the financial cost models for
addressing various market segments.
The basis of the strategic part of the marketing plan is the market map, which
segments the market into identifiable customers. Each segment must indicate the size
of the market, barriers to entry, competitors, market-share goals, and the total cost to
enter the market (including, for example, any companies that must supply critical
softwareorothercomponents).Asstated at thebeginning of thechapter,the planshould
describe what the product is, who will buy it, how the product will be sold, and why it
can be sold.The plan should also specify where or through which distribution channels
the product will be sold and when it will be sold, in terms of order-gestationtime and
effort.
The plan should indicate the identities and quantities of customers, which can be
derived using a competitive market filter that starts with the profile for each customer
and is matched to the product’s features, functions, and benefits (see Table 9-2) The
tactical part of the plan includes the detailed support literature and educationalaspects
taken from the relevant key marketing processes.
Finally, and most important, the plan includes a forecast of units and revenue for
each of the market segments versus expenses that will be incurred in obtaining the
projected revenue (in essence, a pro forma profit and loss statement for each market
segment).In order to make such a detailed plan, the start-up must have a sales model,
which is created by the person responsible for selling.
 Marketing 229

The Marketing Processes. Figure 9-10 describes the processes that are necessary
to support the selling of the product. Each of these processes yields a measurable
output, such as a manual; a news release and the resulting news article; a sales
brochure; a seminar; an application note; a new application that is, in itself, a
product; the training of sales personnel; the plethora of information returning to the
company in the form of customer feedback; and finally, support for high-level
selling when customers visit the home office. When the marketing effort is in full
swing, the head of marketing must be concerned about maintaining the integrity of
the department's output, in terms of quality and productivity. This may take the
form of independent reviews by an outside advisory board. However, responsibility
for the integrity of the output must be fully delegated within the department.

Product user information (e.g., manuals).

Public relations information: news releases, advertising, and technical or other authori-
tative papers that support the company's initial concept and products as they are
released and attain significant positions.

Sales-support literature-i.e., collateral material: brochures, data sheets, price lists,

reports, overheads, and slides and videotape presentations.

Direct-mail programs with telemarketing support.

Trade shows, regional seminars, and technical presentations.

Sample application notes and/or new application products from engineering, an

applications product development organization within marketing, or third parties.

Competitive analysis and product pricing.

Benchmarking, including the distribution of results based on tests conducted during

actual use.

The training of sales and product-support personnel.

Focus groups, consisting of eight to ten potential users with similar backgrounds, that
are run by avery knowledgeable moderator to initially validate the product concept and
provide timely feedbackon the product's features and specifications. This is not a sales
situation. Rather, the purpose is to listen to feedback, and the aim should be to elicit
actual feelings as opposed to polite comments.

Customer and product-testing visits to the home office (or factory), with high-level
selling of the company and its products.

Figure 9-10. Marketing-Process Outputs.

230 Marketing and Sales

The Press

It is essential for a high-tech venture to have a really good relationship with the
technical press from the day the company first opens its doors. If it had to rely strictly
on advertising, no firm could afford to generate the amount of exposure it needs in
order to attract customers, so getting as much free press as possible can be very
beneficial. Ronna Alintuck, formerly of Gateway and a Regis McKenna alumna,
offers the following advice about dealing with the press:

1. Limit access of the press to your best spokespersons. Top analysts are likely
to be very bright but have short attention spans. You rarely recover from
putting the wrong person in front of the press.

2. Don’t try to impress the media. They’ve probably met more successful and
more famous people already. Be genuine.

3. Don’t waste their (and your) time. Don’t call them if you have nothing to say.

4. Once you know them well, ask their opinion and take their insights into
account. Respect the guidance they give to you.

My own recommendationis for the start-up’s spokespersonsto be incredibly direct,

honest, and open. This even means being frank about stiff competition and difficulties
that the new venture may face. If spokespersons avoid discussing problems, or try to
gloss them over, the press is very likely to search out other views of the firm at its
competitors. When the spokespersonsare not in a position to be candid about a certain
topic, they shouldn’t be coy about answering but instead should simply declare the
subject area off limits.

MARKETING FLAWS

As was the case with technology flaws, marketing flaws overlap with flaws in other
dimensions, such as technology, product, business plan, and sales. At the heart of a
high-tech venture’sbusiness plan is the followingquestion: ”Willcustomers buy the
product at a given price and rate?” Many of the flaws involve not understanding this
question. Other flaws involve trying to attack markets already held by strong
competitors. As with all of the organizational dimensions, not having the right
people can cost the company a great deal of time-and maybe its life.

Being Too Early with a Pioneer Product

The first product of a new class is inherently more expensive because of new
technology, the learning curve, and lack of competition. Most entrepreneurs argue
 Marketing 231

for fast introduction, followed by evolution in order to get the product finished and
introduced into the marketplace. Thus, the first product may be barely usable. A
pioneering product usually attracts only a small number of users unless the need for
the product is extraordinarily obvious. In cases where the market is fueled only by
early adopters, it’s critical for the start-up to have the right price and cost. However,
a small market is unlikely to support a high price. The only real solution is simply
to wait until the technology has advanced to the point where it becomes feasible to
offer the product.
Pocket PCs are an interestingcase in point. One might logically expect that a large
company such as IBM should be producing pocket-PC prototypes, complete with cel-
lular radios and fax, and having the devices tested by its work force in order to explore
truly personal computing (let’s call it ”intimate computing”) for the twenty-first
century. However, to be really effective, such a device may have to understand voice
and/or handwriting. The following story will shed some light on why large firms such
as IBM have not yet tried to market pocket PCs.

Workslate, the first pocket PC. In December 1983,Convergent Technologies intro-

duced Workslate at a price of $895. First presented in the American Express
Christmas catalog, Workslate was sold almost as a high-tech novelty product
through several different distribution channels instead of being marketed as a
compatible and integral part of Convergent’s office computer line. The device’s
small, hard-to-use keyboard was a significant drawback, since it made text entry
difficult.On the one hand, Workslate was an expensivephone book, memo pad, and
calculator; but on the other hand, it was a forerunner of palm-top helpers such as
dictionaries, bibles, and the phone directory, calculator, calendar, and memo pad
from Casio and Sharp. In short, Workslate was about six years ahead of its time.
Anyone connectedwith Workslate’sdevelopment,marketing, or purchasingcould
have asked two questions:Who would buy the product for any application that could
be named? Who would buy it at the specified price? The answer to both questions was
”almost no one.” Ultimately, only five thousand units were sold rather than the two
hundred thousand units that had been planned for in the materials purchase. The
product was discontinued in July 1984, and the company wrote off $15 million.
In 1990, Atari and Poquet introduced IBM-compatible pocket PCs for $500 and
$2000, with a small and full screen, respectively. Since both have small keyboards, the
utility (and hence, the market) is likely to be somewhat limited unless it can be shown
that the keyboard doesn’t hamper input or until a use niche can be established.

Failing to Realize That Emerging Markets Take Time, Patience, and Capital

Emerging markets take time. If the start-up is predicated on putting a new product
into a new and undeveloped market, the process is likely to take longer than
232 Marketing and Sales

planned. It is practically impossibleto construct a model that can tell how long it will
take, or how expensive it will be, to develop an emerging market.
Several high-tech ventures started up in the mid-1980s using rule-based systems
technology developed by computer science’s artificial intelligence (AI) community.
These firms have evolved rather slowly and together have annual sales of less than $50
million. The technology-to-producttransition has been slow to occur, with a relatively
small number of applications using the rule-based programming approach. The
organizationsthat have remained small and operated in a controlled fashion have been
successful.

The first AI companies. Teknowledgewas founded in 1981,along with three other

firms-AI Corp., Intellicorp,and the Carnegie Group (based on Carnegie-Mellon’s
work in AI). Teknowledge’s mission was to become known as the premier AI
company. Its strategy was to hire all the best people so that no one else could start
up. It rented expensive space in Palo Alto, hiring AI researchers from Stanford’s AI
laboratory who had little or no product experience.
Teknowledge made ”strategic alliances” with several large U.S. and European
companies,includingGM, by sellingstock in exchangefor a close working relationslup
on applications.It initiallytrained other organizationsin rule-based programming,built
special systems, and did government research in order to develop its next-generation
technology. All three of these activities were potentially profitable, and the firm could
have run profitably from the outset.Sincesuch a venture was service-and labor-intense,
Teknowledge attempted to build a high-standard product in order to obtain higher
operating margins and higher valuation.
New, small, low-overhead competitors, such as Neuron Data, began to introduce
standard rule-based products for a small fraction of Teknowledge’s price. They had
small staffs and equally bright people, but unlike Teknowledge, their people had
product development and start-up experience.
Teknowledgeraised tens of millions of dollars through private and public funding.
In 1988,with money stillin the bank and salesof its expensiveproduct rapidly declining,
Teknowledge merged with ailing Cimflex, a Pittsburgh-based company building
custom programsfor manufacturing.The Teknowledgeportion of the organizationwas
reduced to a small operating division doing contract research. Those at Teknowledge
who were responsible for the merger (and who were also founding shareholders)did
not suffer financially.
In contrast, Intellicorp (which began as IntelliGenetics) started from the same
Stanford core technology to do gene sequencing.After the firm went public, the gene-
sequencing business turned out to be less robust than it had originally thought.
Intellicorp went on to develop and market its proprietary language, KEE, for building
rule-based systems.
 Marketing 233

Chapter 11includes the story of Gensym, which succeeded by understanding and

focusing exclusively on process-control applications.

Trying to Establish a Technology Monopoly

The case of Teknowledgealso demonstrates the flawed approach of starting up with

the intention of creating a technology monopoly by cornering the market on all the
bright people. It is impossibleto achievea technology monopoly. Attempting to hire
all the “smartest” people in a new area in order to prevent the formation of
competitive companies is a dream derived from biotechnology start-up strategies.
The supply of top-quality individuals, albeit finite, is large, and all the ones who
weren’t asked to join the new venture are natural, highly motivated candidates for
starting competitive firms.

Attempting to Establish an Always-EmergingMarket

An emerging market is unable to support a very long, slow market development

with accompanying incremental product tuning. As such, taken over a decade or
longer, the market is still emerging.Market development may be limited by the need
to change basic institutions or processes or to create a generation of potential users.
The ultimate product cannot be built; instead, over time, partial products are
introduced that chip away, niche by niche, at what is perceived to be the true market.
One of the greatest temptations is to attempt to define an obvious, previously
untapped market that the availability of compelling new technology (e.g.,multimedia
for computers) would appear to make possible. Computer-aided instruction (CAI)
epitomizesa market that has been emergingfor a quarter of a century. Computershave
made incredible progress in aiding learning.One of the simplest,yet least obvious, uses
is a ”help” menu that enables a user to learn about a system. Other forms of training
include industrial simulators for power plants, aircraft, military-game simulation,
computer simulation of industrial firms and cities, educational games, industrial
courses, and computerized instruction in the classroom. However, a general program
to provideordinaryclassroominstruction(orevenreplacetheinstructor)at allelementary,
high-school, and college levels still remains elusive, even though the need for such
assistance would appear evident, in view of American students’ poor ability to learn
such subjects as mathematics.

Plato and computer-aided instruction. In 1965, Professor Don Bitzer at the

University of Illinois and Control Data Corporation built a system, Plato, for
supplying computer-aided instruction. Bill Norris, the president of CDC, was its
chief proponent and salesman. CDC invested several hundred million dollars to
234 Marketing and Sales

build an instruction network using Plato and its large 6600 computers. The Plato
system used the first multimedia terminals,with computer interaction,slides,voice,
and audio output. Although Plato has been successful with thousands of courses
and millions of course hours in university training and applications, including
teaching basic skills to prisoners, it and newer PC-based CAI programs have yet to
deliver the promised revolution. Clearly, improving education is an important goal
of all countries. Perhaps the CAI revolution awaits the revolution in ubiquitous,
zero-cost, multimedia capability foreseen by some for the 1990s.
If Plato were located in Silicon Valley, some company would no doubt start up to
develop a low-cost computer platform to utilize the vast array of courses. Corms,
located in the valley, is still waiting for the market surge.Apple servesthe market, albeit
in an ad hoc fashion. By making computers fun-and-game-oriented,Nintendo may
have found the true pathway into the market.

Attacking Walled Cities

A classic marketing flaw is to attack a large company’s customer base with a

competitivereplacement product. Rarely is this approach successful,since customers
would prefer to buy from a few suppliers that are also the leaders. The new product
typically attacks a strongly held market by using a different or incrementally
improved next-generation technology. Existing suppliers, particularly start-ups,
are unwilling to give up their market position and can hold their share of the market
by enhancing their products through evolution. Attacking the customer base (e.g.,
IBM, Lotus 1-2-3 clone) of a supplier that is unwilling to accept the loss of revenue
(e.g.,add-on disk memories) or loss of control (e.g., database) is a flawed approach.
It will succeed only if the new technology is compelling and the competitor cannot
move prices rapidly (e.g., plug-compatible IBM mainframes).
Autodesk successfully attacked the mechanical computer-aided design (MCAD)
market by building a product that ran on a PC as opposed to the older minicomputer
(e.g., Computervision).It succeeded because the established companies neither saw a
threat nor were able to lower their margins, since they had fixed costs and fixed ways
of operating based on selling a few, expensive software packages.
Two major industries have formed through efforts to attack a large company’s
customerbase:plug-compatiblecomputers(pioneeredby Amdahl)and disks(pioneered
by IBM alumni).Both industriesoriginatedat a time when IBM had extraordinarilyh g h
profit margins on computers and peripherals. A1 Shugart, disk pioneer and IBM
alumnus, described the opportunity as follows: ”IBMs high profit was immoral. Any
self-respectingengineer would start a companyjust to bring lower-costproducts to the
mass market.”
However, by aggressive pricing and by increasing the complexity of the disk
subsystems, the established firms have decreased the significance of the plug-
 Marketing 235

compatibleperipheralbusiness.The 1980ssawfailuresby InformationStorageSystems,

Memorex, Storage Technology,Telex, etc. In place of the plug-compatible peripheral
industry, a substantially larger disk-componentindustry has formed, based on IBMs
Winchester technology, to serve the high-volume PC and workstation markets. Given
the cost difference(asmeasured in cost per byte)between largeand smalldisks, a strong
disk add-on industry could reemerge in the 1990s to attack the hgh-margin part of the
minicomputer and mainframe industries.

Cullinet. Cullinet was founded in the late 1970s by John Cullinane, an IBM
salesman who started the firm to sell a distributed-database product created by one
of his large customers. The company evolved to build products on a totally
opportunistic basis to fill the niche in IBMs product line. It succeeded for a while
selling its standards-oriented database before IBMs relational database became
popular. Becauseadatabasesystemconstitutesapredominantportionof a computer’s
operating system, IBM found it unacceptable to have such a key piece of its system
built by another vendor. In 1989, Cullinet became part of Computer Associates, a
large and successful company based on developing general-purpose software that
it derived from specific solutions it had encountered in consulting for IBM users.

Amdahl Corp. Amdahl Corp. was founded in the early 1970sby Gene Amdahl to
make high-performance IBM System/36Os. Amdahl was formerly the head of an
IBM laboratory that built a high-performance computer, but the laboratory was
closed because IBM felt that the demand for, and profitability of, large systems was
low and the development cost too high. During Amdahl’s start-up, the technology
took longer to develop than anticipated, requiring more funding. Fujitsu funded
Amdahl in return for 49 percent of the company and for technology transfer in the
form of training, CAD, gate arrays (derived from IBM-pioneered master slice),
packaging, software, and manufacturing rights. When Amdahl entered the market,
the cost of mainframe computing dropped by 40 percent and continued to decline
at a rate of 15 percent per year. Previously, the cost had remained nearly constant.

Andor-Amdahl could do it again in the 1990s. In 1987, Gene Amdahl started a

new company to build an IBM-compatible computer using complementary metal
oxide semiconductor (CMOS)gate arrays. Andor’s first deliveries are scheduled for
1991. The firm could be successful, unless IBM and the Japanese plug-compatibles
(Fujitsu and Hitachi) switch to CMOS4and sell lower-priced computers. A custom
CMOS computer (microprocessor)would be significantlyfaster and cost significantly
less than a gate-array version. When or whether such a chip could be built is
anyone’s guess.

4. Chapter 13describeswhy theECL-basedproductsaredoomed,except

atthevery highest performance.
236 Marketing and Sales

Building a ”JustAnother Product” of an Existing Type

Predicating a company on capturing a small part of a newly established market with

a product that is ”just another X (JAX)when a new technology becomes available
is a flawed approach. Virtually all new computer classesinitially attracted hundreds
of entrants (seeChapters 8 and 12).Only a few of those ventures were successful in
the beginning, and only two or three survived for as long as ten years.

Failing to Find an Adequate Market Niche

A company may attempt to carve out a suitable market niche with a product whose
cost is either too high (not enough buyers are available)or too low (sellingexpenses
cannot be covered). In either case, it will be unable to develop a business.
As indicated in the previous chapter, niches provide a protected space in which a
new venture can conduct its business, free of competition, until it becomes established.
A niche is often the only way a fledglingcompanycan develop a product that will return
high margins and, hence, be profitable enough to fuel growth. However, if the niche is
too tiny, the firm won’t be able to find buyers and will therefore have no market. If the
niche is too large, there will probably be many competitors, and prices will be too low
to obtain adequate margins. A strategy whose objective is to claim a niche from other
niche playersor fromnewly established,aggressivestart-ups is almost certaintobe fatal.

The elusive graphics supercomputer and the risks of nichemanship. Ardent and
Stellar attempted to define a new niche that they believed would be profitable and
unique. It was to be carved from two nearby niches: minisupercomputers and high-
performance 3-D workstations used for visualization. However, these niches were
owned by aggressive competitors (Silicon Graphics and Convex), which fought to
maintain their market positions. Trying to carve a niche from the Silicon Graphics
market position was essentially an attack on a ”walled city,” a flaw just discussed.
Trying to carve a niche from the Convex market position was essentially an attack
over a desert.The desert existed because the cost of the graphics supercomputer was
so low as to make it infeasible to sell a low-priced minisupercomputer. In addition
to the time-consuming problem of defining a new niche for a visualization
supercomputer, the selling costs were higher than anticipated, resulting in an
impracticable business plan.

Relying on a Single Customer to Distribute the Product

Virtually all the systems companies that have experienced sudden death have done
so because they depended on a single customer that would relabel and sell their
product and the customer then decided not to continue the relationship and
 Marketing 237

funding. For example, Cydrome teamed with Prime, ETA was part of CDC, and
Multiflow had an agreement with Digital.
The problem stems from the relabeler‘s changing its mind or not being fully
committed to the supplier. A start-up that is consideringdoing business with only one
customer should think again.Even with thebest relationship,the firmisstill at the mercy
of the reselling organization.
Rob Peglar, an engineer with ETA, commented on the CDC relationship:

Many people in the computer industry assume that most computer-company failures
must be a result of poor product, design, or manufacture of some kind. Not so.Computer
companies fail because of poor management and erroneous, ill-timed decision-r the
lack of coherent, timely decisions.

Having an Incomplete Product

A new venture may attempt to sell an incomplete, and therefore useless, product if
it mistakenly assumes that there exists a very broad market for a general-purpose
computing tool when the product is in fact differentiated only by having the ”right”
application software. Specific customers for a system have to be identified in the
beginning, and then the appropriate application software must be secured to
address the markets. The product must be complete!
A common oversight in building a new hardware system or platform, or a generic
softwaretool, is for a start-up to ignore the particular applications programs that must
be generated by either the user or independent software vendors until the product is
introduced.Asaresult, thecompany findsitself witha product that cannotimmediately
be used by the intended buyers. By the time the firm discovers the dilemma, it is in a
significant budget crunch as it scurries to persuade software vendors to ”port” their
applicationsto another platform.
Software vendors are generally very enthusiastic about porting the software
necessary to make a product complete, because they find it an effective method of
financing their companies.Larry Ellison, CEO of Oracle, described this as ’laundered
venture financing.”Oracle was able to charge various newly financed platform start-
ups as much as $1 million for porting its database. The optimum strategy for a software
vendor is to demand up-front financing and hope the platform venture goes under
before the port is done.

Failing to Identify Who Will Buy the Product and Why

The customer/application profile must be used to pinpoint the customers. The

buying rationale is then examined once the customers/applications have been
identified.
238 Marketing and Sales

Product developers are often tempted to simply develop products that, on the
surface, appear to be major leaps forward from an existing product. Both Analytica
(described in Chapter 11) and Javelin, its Boston-area counterpart, were founded to
develop a product that would extend and take market share from Lotus’s 1-2-3.Javelin
attacked 1-2-3head on and was repelled by the loyal user base that wasn’t interested in
switching to a new product, no matter how powerful. Although neither of the new
products was successful as a mainline replacement, when Analytica’s product was
repositioned asadatabaseand Javelm’sproductwas repositioned asa high-performance
analytical tool, both were able to find a niche at lower and hgher price levels,
respectively.

Assuming That Universities Are a Market unto Themselves

All too often, high-tech ventures focus on the university market segment. Univer-
sities apply computers in a broad range of academicdisciplines and really represent
only the leading edge and early adopters for the application of many products.
University users are demanding, critical, and provide user feedback. In addition to
locating beta tests at universities, as suggested previously, it is wise to sell the first
few products of a given application to universities and get their feedback and
imprimatur. Unfortunately, universities demand high discounts. Thus, unless a
start-up has a completely unique product (i.e.,a monopoly, such as Xerox had with
the first photocopier), it is hard to maintain adequate margins by predicating an
entire applications market on extensive university sales.

Having a Poorly Thought Out Cost Plan

Very high, fixed market-entry cost (e.g.,advertising, support) can make the sale and
distribution of a new product infeasible.Various products appear to suffer from this
flaw. The Ardent computer, for example, was limited because of its early pricing as
a workstation. It was simply priced too low to be sold in an established
minisupercomputer market at high enough volumes to cover the market-entry
costs.Similarly,the Analytica story, described in Chapter 11,involvespoor marketing
and the wrong price.

Preannouncing a Product and Having It Stolen

Calling on potential customers, or hyping a nonexistent product to verify the

product specifications, signals what a new venture is doing. This gives competitors
a chance to respond before the company can get off the ground. If the product is
 Marketing 239

being sold to a buyer that could develop such a product itself, the buyer is likely to
either be working on a similar product already or be prompted to begin working on
one. In some cases, the outright theft of trade secrets occurs.
The risk of giving away the storeby callingon customersand potentialcompetitors
during the seed and product development stages is very high. The Stardent story is a
wonderful example of the importanceof security in developingthe first product. Allen
Michels, founder of Convergent Technologies and Ardent, described Ardent and
Stellar, prior to their first product shipment, as "the battle of the big mouths." The
winners in the battle were competitors and users that obtained more competitive
products.

Visix Software. Visix was founded in 1987 to build a graphical user interface and
system manager for UNIX. Visix representatives called on several hardware plat-
form companies while the start-up was working on its first product, Looking Glass.
In two cases, established hardware firms began building competitive products
using ideas that had been discussed under nondisclosure agreements with Visix. In
one case, a recruiting company was employed to go after Visix employees to help in
implementing what the engineer in charge described as "a product we stole from
Visix." Visix did not file suit, choosing to concentrate on making money through the
sale of its products rather than through litigation.

Having the Start-up's Product Announced Prematurelyby Universities

The task of universities (facultyand graduate students)is the discovery,production,

and distribution of knowledge. Nondisclosure contracts signed by the university
community virtually guarantee instant, wide-scale disclosure of a product or
company. Conversely,if a start-up wants to gather intelligencecheaply, it should be
networked and simply ask the university community what potential competitors
are doing.

Hiring the Wrong People, Especially the Head of Marketing

Flawed hiring practices are common among high-tech ventures. As was the case
with the head of sales, locating the right head of marketing is unlikely. In my
experience, the probability of finding an appropriately qualified individual when
the company starts up is about one in four. Although the head of sales can finally be
tested with quantitative measures in the marketing development stage, it is very
difficult to measure the head of marketing. The full Bell-Mason Diagnostic provides
one such measure, but if a company is in trouble from a marketing standpoint, this
240 Marketing and Sales

will probably become evident by the end of the seed stage when it fails to satisfy the
marketing rules for that stage.
According to Ronna Alintuck, some common flaws among marketing candidates
hired for start-ups are that they:

Have MBA degrees and believe their degrees make them better marketeers

Believe they can precisely predict the outcome of a marketing program

Were not personally responsible for at least one marketing success, yet have
never failed with a marketing program

Are not passionate, emotional, and controversial

Are too easygoing or stop thinking about work the minute they leave work

Are afraid to say "I don't know"

Are process-oriented, committee people and make decisions based on consen-

sus

Are not both creative and technically adept [as evidenced by their ability] to
understand and enjoy the technology and product for which they are respon-
sible

MARKETING RULES

In the concept stage, the efficacy and uniqueness of the product or service is the
major determinant of market success.Thus, the product and technology dimensions
are emphasized more strongly.At one end of a product-demand-curve spectrum, no
market exists at the current price level. At the other end of the spectrum, the
company may predicate its business on capturing a small fraction of a very large
market with a marginally better, niche product or technology. Either strategy is
almost certain to fail. During the concept stage, the organization must focus on really
understanding whether there exists a market that is large enough or manageable
enough to enable the company to get the toehold it needs in order to develop.

Have the sets of customers (i.e., who) been identified for the product?

In order to begin working on a marketing plan, it is essential for a new venture to

start by clearlyunderstandingwhich customerswill buy the product, by profession and
by use, includingtheir organizationalaffiliations.This first step is carried out during the
concept stage.
 Marketing 241

Does the marketing plan at the concept stage contain a list of the customer/
application profiles ke., who/what) to be developed during the seed stage?

This rule, distinct from the rules for the product, focuses on whether the start-up
understands who the buyers of its product are and how they will use or apply the
product.This testing takes the form of a series of customer/application profiles, which
describe representativeindividualsin specificuse segments.These profiles include the
users, the operational environment, specific unmet needs, the ability to buy, etc. If the
organization does not have this type of detailed image of the user, together with an
understanding of the intended use, it lacks the information required to design an
effectiveproduct and reach customersthrough an effectivemarketingand sellingeffort.

Has the start-up identified a compelling buying rationale (i.e.,why) for each
of the customer groups to purchase the product?

In the case of a totally unique product, the company must construct a compelling
buying rationaleto attract new customers.Ideally, the utility of such a product (e.g.,the
first spreadsheet)will be self-evident. In the case of a more conventionalproduct, the
new product must add a feature or dimension that no competing product has. At all
stages, the firm must continue to be able to answer this question affirmatively.

Has a simple estimate of the market size been developed, supported by

articles and extrapolated market numbers as well as other public informa-
tion?

The market size must be quantified in terms of the aforementioned customers and
their applicationsin order that a business plan with numbers can be made. It is useful
to be able to size the market in various ways, includingstartingwith basic demographic
data. Most libraries can provide a plethora of "free" data that characterize the world-
wide information-productmarkets. And for nearly any product idea, regardless of its
merit, at least one or more market surveyscan be purchased at $1,000 per kilogram that
proclaim the existence of a viable billion-dollar market at some future time.

Has the start-up created a simple market map showing the paths the
company will use to reach each of the sets of users (i.e., how)?

This rule diagnoseswhether the firm knows how, or by what channels, to reach the
customers.Although salesis responsiblefor supplyingthe specificnumbers,marketing
has to identify the alternativesand recommend the best routes. The principal role of the
marketingmap, however,is to ensurethat the companyis aware of the need for "active"
distribution channels. A flaw in many marketing plans is to forget about all the other
242 Marketing and Sales

vendors in the distribution chain. In many cases, however, the start-up’s product will
not be ”complete” (i.e.,ready for use by the final buyer) without one or more products
that must be supplied by these vendors. Nearly all hardware and software products
depend on additional products in order to form a complete product and, hence, a
successful market.

At the end of the concept stage, does the start-up have a simple outline of a
market plan, which can be expanded during the seed stage?

This rule examines whether, at the end of the concept stage, the company knows
how to make a market plan so that, given a product, it can help salespeopleidentifyand
reach the customers. A finalized, detailed plan is not required at tlus point, only an
outline for such a plan.

By the end of the seed stage, has a product requirements specification been
written?

Definingthe product is a key activity of the seed stage. Marketingis responsible for
definingtheproductrequirementssothat engineeringcanmaketheproduct specification
for designing the product.

At the end of the seed stage, are the preliminary customer/application

profiles (with needs analyses), initial product concept, and projected unit
cost roughly in line with the hypothesis developed at the beginning of the
seed stage, and have any changes been factored into the business plan?

During the seed stage, the marketing person is finding users who understand the
product and may be duential in specifyingits details.This is an excellent time to form
and recruit a technical or customer advisory board (TAB or CAB), which will help
godfather/godmother the product into existenceby advising the company on critical
features and functions as well as how to build the product. The customer/application
profiles (CAPS)describe who is going to buy and what they require by way of product.
As the companybegins to build the CAPS,it must understand what informationit
needs from the data-gathering process, including a ranking of what it believes are the
critical features, functions, and benefits (i.e., why) in the buying decision.
A focusgroup is an effectivetechniquefor really hammering out product functions.
A small, select group of eight to ten potentialusers come together to give a product and
market critique.The idea is for the start-up to listen to direct, but not necessarilypolite,
feedback and to refrain from selling. The group must be moderated by someone who
really understands it all.
 Marketing 243

In determining system configurations, for example, one of the requirements for a

new system may be the ability to communicate with existing systems. At the lowest
levels,communication takes place according to certain industry and de facto standards
and protocolsand dialects,suchasIBMsSystemNehvork ArchitectureortheDepartment
of Defense's Transport Control Protocol and Internet Protocol (TCP/IP). Other com-
municationis via file and databaseformatsand standard languages.The start-upcomes
to fullyunderstandall thesedetailedrequirementsbybuildingthecustomer/application
profiles, reviewing customer inputs, and listening to a focus group. Each of these
activities will help the company understand how the product will be applied.

By the end of the seed stage,have the sets of customers and their applications
(i.e., who) been identified for use during the product development stage?

Once the organizationreaches the end of the seed stage, it must have a pretty clear
understanding of who is expected to buy its product, by profession and by use,
including their organizational affiliations.

By the end of the seed stage, has the concept stage market map been updated
and refined based on initial explorations and field research performed
during the seed stage?

By the end of the seed stage, a really complete market map is required in order for
the start-up to enumerate and understand all the ways in which the product can be
delivered to the market, although the specific route remains to be chosen. The map
should start with the SIC code/customer/application groups, then look at various
distribution and product-finishing channels, including VARs, independent software
vendors, retailers, dealers,distributors, OEMs, and brand relabelers.The final stages of
the map end up within the company as a supplier of a component or a system. Some
testing of the market map should have occurred in the process of understanding the
CAPSand determiningthe availabilityof other softwareto work in conjunctionwith the
firm's hardware or software product.

Does the person responsible for marketing have experience in successfully

marketing high-tech products in the start-up's market and product space?
Can he or she attract, lead, and manage a "grade-A" marketing staff?

"Grade-A marketing people are those who have been responsible for high-tech
market successes, workmg collaboratively with engineers, and can function with
minimal resources under severe schedule pressure and changing plans.
244 Marketing and Sales

Can the head of marketing bring to the company a vision of how the product
will be used to establish a unique and lasting market?

The responsible marketing person or the vice president or director of marketing

plays one of the key roles in a high-tech venture. Although the head of marketing is not
necessarily on board at the concept stage, having a leader is essential during the seed
stage and as the company starts up. Marketing must be a strong partner in the product-
definition and planning process.
A marketing person coming from a sales background can be highly creative but
may only deal in tactics, ignoring product and market planning and management. An
engineer may focus excessively on planning and management and not attend to tactics
(i.e., helping salespeople sell). An ideal background is an engineer who has spent
enough time selling successfullyto understand the requirements for marketing.

Is the preliminary market plan outline (i.e., what the company has to do in
order to deliver the product) in place, based on potential product position
and competitive analysis? Does it include a tactical plan for programs, with
costs and resources as a function of time?

As the company enters the product development stage, it must have a plan for a
market plan in order to establish goals for output, guide spending, and determine
resource requirements.

Does the market plan outline include the following components?

Preliminary corporate and product-positioning platform or statement with

competitive market environment
Simple product specification (features and functions), which has been
translated into potential benefits for users
Simple descriptive customer/application profiles of key market areas

Global targets of opportunity

Market map refined with a preliminary outline of requirements for selected

paths
Global tactics to reach the market, including advertising, PR, trade shows
and seminars, etc.
 Sales 24.5

The start-up’s market plan outline must contain substantive detail regarding the
key topics listed above.

SALES

Sales must produce orders so that manufacturing can ship the company’s products
for revenue. Since so much has been written about selling, I will provide only a brief
overview of this dimension. White (1977) offers a fine description of the sales
organization and the motivation of sales personnel, including enumerating all the
ways in which a sale can be closed.
During the concept and seed stages, only a model for the sales organizationexists.
If the company’sproduct is to be marketed within twelve months,the head of sales may
be hired by the end of the seed stage.
Often, sales is so closely related to marketing, particularly during the early stages
of a start-up, that it is hard to diagnose the two as separate entities. Once the product
begins to be sold, however,the sales organizationcan be measured quite easily in terms
of the booking of products to be delivered,with the companybeing rewarded according
to the amount actuallysold.Unfortunately,it takes at least six months to fully implement
any changes in the sales organization-e.g., a new head, regional managers, individual
account representatives,or a commission plan.

THE CONSTITUENT DIMENSIONS OF SALES

The sales function, like marketing and engineering, can be decomposed into its
constituent dimensions to form the sales balance sheet. The dimensions of the
balance sheet are:

Formal processes for running the sales organization: These processes include formal
trainingand periodicsalesmeetings;order processing, revenuerecognition, product
shpment,and revenuecollection;salesforecasting;customervisitsand presentations;
field seminars; field marketing program development;etc.

Sellirzg plan and model: These form the basis for controlling sales and sales produc-
tivity.

Presales-support and sales-support ou tpiits and quality levels.

Customer-support quality level.

246 Marketing and Sales

Head of sales.
Regional sales heads.

Sales resources: This dimension includes the field sales personnel, offices, and
infrastructure.

Operational control of the organization: This dimension includes MBOs and the
ability to meet the selling plan.

In the start-up’s later stages, rules test each of the preceding dimensions.
At the seed stage the only relevant factor for salesis a realistic sales plan outline and
model for selling.

The Selling Plan and Selling Model

Because sales costs determine whether the product is feasible at the price level, they
therefore directly determine whether the venture itself is really feasible. Thus, sales
is responsible for providing a realistic sales model for each of the customer groups
identified by the marketing organization. The following parameters must be de-
termined in order to make both a sales plan and the business plan for the company:

Time and cost to hire and train sales and sales-support personnel

Sales-cost profile, including the complete cost of making a sale versus time

Order-gestationtime, from first contact to final sale

Sales productivity (sales/year), including the learning curve of the company’s

salespeople

SALES FLAWS

As was the case with all the other functional areas-including engineering,
manufacturing, and marketing-the sales effort is subject to numerous flaws that
can limit a new venture or even cause it to fail. Several of the most common of these
flaws are described below. Sales management is simply an ”art form,” like other
areas of management, that demands understanding and experience.Nearly all new
ventures are plagued by a combination of marketing and selling start-up problems
that cause them to miss their revenue plan and require additional funding. When
this happens, the sales organization points to the marketing organization as the
cause of the problems, and marketing, in turn, accuses sales of being untrained and
incompetent. Both point to engineering for product deficiencies.
Having an Overly Optimistic Order-Gestation-TimeModel

New ventures are almost always too optimistic about how long it will take to get
orders. In the case of products entering emerging markets, all the selling in the
beginning is ”missionary selling,” which follows the time-honored model of first
selling to a research community, then to early adopters, and ultimately (it is hoped),
to a large market. In order to compensate for this tendency to underestimate the
order-gestation time, market calibration is included as a normal stage of a start-up’s
development.

Having Sales Costs That Are Too High to Support a Viable Business Plan

When the selling costs begin to be tallied in the market development stage, it may
become apparent that the start-up is in dire trouble because of high selling costs. The
trouble may stem from the company’s failure to understand where, on an economic
basis, the product will be sold at the price levels that are assumed in the business plan
and required by the marketplace. The nonexistent-niche market flaw is directly
related to the characteristicsof selling a product (price,sales-gestationtime and cost,
and support cost).

Lack of an Effective Sales Leader

The head of sales is the critical person for the marketing calibration stage. He or she
must hire and lead the sales team and assist in closing sales. The probability of
getting the right sales leader is less than fifty-fifty, in my experience. All sales heads
can sell themselves for a while, but ultimately, the numbers tell the story. Unfor-
tunately, the company will be operating at its highest expense rate by the time the
sales leader’s shortcomings manifest themselves.

Having an Overly Optimistic Hiring Plan

The sales plan may make rosy assumptions about the availability of job candidates
who are already skilled salespeople or who can be trained to sell the product in a
relatively short time. The sales plan may also neglect to provide for adequate sales-
support personnel. Although these individuals are called ”sales support,” they are
often the ones who actually do the selling when a complex product’s content is the
basis for the sale.
This common flaw comes from not understanding the support needs (costs)and
results in doubling the cost of sales. All of the systems companies with which I’ve been
associatedover the last decade consistentlyoverestimatedthe salespeople’sability to be
trained to understand and sell technical products. Invariably, successful salespeople
248 Marketing and Sales

either completely understand the product themselves (which only happens in a

minority of cases)or have someonewho might be termed a "technically knowledgeable
alter ego" accompany them on sales calls when the product is discussed in detail.
Depending on the energy and competitiveness of the sales representatives,between
one-half and two sales-support people are required for each salesperson.

Having an Incorrect Product-SupportCost Model

After the product's introduction, it may become apparent that the product is much
harder to use than was originally anticipated. Thus, a field organization is required
to support the product, including training customers and helping them apply the
product. Often, the difficulty in using the product is attributable to some type of
product flaw that results in a need for significant and inordinate "hand-holding."
Sometimes, users are simply unable to cope with the product's complexity within
a reasonable time. In either case, more time and costs are incurred before the product
can be sold in quantity.

SALES RULES

It is highly unlikely that a new venture can bring a head of sales aboard at either the
conceptor seed stage.However,having a model of the "sale," includingall the costsand
the time frame, is critical. The only way a realistic model of selling costs can be made is
by using a similar product as an example. Even the "worst-case model" will probably
turn out to be optimistic,however.In most instances,this occursbecause the start-uphas
prepared its model by comparing its product with a steady-state product from an
established company.

At the concept stage, does the start-up have an initial outline of channel-of-
distribution alternatives, their typical requirements (e.g., selling cycle[sl),
cost of sale, and a first model of the sale?

A selling model is required for each distribution channel, including sales-gestation

time and effort, geographical distribution of customers, etc. At this early stage of the
company's development, the sales organization is not a large component of the plan.
Thus, a requirement right from the start is a model of the sale that includes selling cost
and time. Without this model, the firm is likely to be flawed because its product will be
economically infeasible to sell and support.

By the end of the seed stage, has the start-up developed a sketch for a
preliminary sales plan-including distribution channels, organization, and
"model" cost-and verified it against similar products?
 Conclusion 249

The sales organization is usually formed after the product is well along, since it’s
usually inappropriate to hire salespeople at start-up time. However, a person who
understands the sales process in the specific market area is required in order to build a
credible sales model. In many cases, the head of marketing assumes this role in the
venture’s initial stages, especially if he or she has experience in selling.
Vitaldetailsthat are often overlookedin the sales-planningprocessincludethe need
and time for sales training,the requirementsfor a salesperson,and the need for technical
sales-support personnel.

By the end of the seed stage, has the company identified sales-management
candidates with the appropriate experience who will sign up to meet the
sales-cost and sales-productivitymodel contained in the business plan?

The start-upmust identify a potential sales leader who will check the efficacy of the
business plan. Although the person may not actually be hired at this time, identifying
likely candidates is critical.

CONCLUSION

Six questions determine the success of a product or service and, hence, of a company
that is started to produce that product or service:

1. What is the product, and is it complete and ready for use by the potential
customers?

2. Who will buy the product?

3. How will the product be used-i.e., for what application?

4. Why will customers buy the product, in terms of its features, functions, and
benefits?

5. Where will the product be sold-Le., through what distribution channels?

6. When will orders be received and filled-i.e., how long will the process take?

Marketing is responsible for answering all of the above questions. It shares the
responsibility for question 1with the engineeringorganizationand for questions5 and
6 with the sales organization.
The start-up can employ a variety of techniques to answer these questions and
evaluate the marketing organization. For example, a customer/application profile
addressesquestion 3. A market map is required to enumerateall the paths the company
250 Marketing and Sales

can use to distribute its product (question5). And the firm must have a sales model in
place by the end of the seed stage to guide the selling process (question 6).
Nine dimensionscharacterizethe marketingbalancesheet:the marketing processes;
the marketing plan; the marketing-support output (e.g., literature, public relations);
tactical sales support, including targeted customerlists; the head of marketing; the t o p
level marketing team; a customer and/or technical advisory board; the marketing
resources, tools, and people; and a control system, with MBOs and output measured
against the marketing plan.
Eight dimensionsare important for sales:processes; a selling plan and model; sales
support;customersupport;the head of sales;lus or herregionalsalesmanagers;the field
sales resources; and operational control. The need for quality pervades all these
dimensions. At the seed stage, the Sales Dimension is concerned only with a realistic
model for sales costs, productivity, and order-gestation time.
Chapter 10

THE BELL-MASON
DIAGNOSTIC

When you can measure what you are speaking about, and express it in numbers,
you know something about it: but when you cannot express it in numbers, your
knowledge is of a meager and unsatisfactory kind: it may be the beginning of
knowledge, but you have scarcely, in your thoughts, advanced to the stage of
science.
-William Thompson, Lord Kelvin (1824-1907)
Popular Lectures and Addresses, 1891-94

Everything should be as simple as possible, but no simpler.

-Einstein

The Bell-Mason Diagnosticand Prescriptive Method is a rule-based tool that is applied

manually to characterize and plot the status of a high-information-technologyventure
at each stage of its growth. The start-up is compared to the diagnostic’sdefinitionof an
”ideal” company by testing it against a set of rules, which are applied by answering a
seriesof questionsfor eachof twelveevaluationdimensions.The answersare tallied and
plotted on a relational graph, which is then compared to the ideal for that stage of
development.The graph highlightsthe firm’spotentialdeficienciesand pinpointsareas
that are in or out of balance.

251
252 The Bell-Mason Diagnostic

THE FOUR ELEMENTS OF THE BELL-MASON DIAGNOSTIC

The four major elements of the Bell-Mason Diagnostic include:

1. The five stages of company growth

2. The twelve dimensions that are measured to assess a start-up

3. The rules used to evaluate each dimension

4. A relational graph plotted against the ideal model for success

These elements are described briefly in the following subsections.

ELEMENT 1: THE FIVE STAGES OF COMPANY GROWTH

The range of computer-and communications-technology-basedcompaniesis largeand

will reach the trillion-dollar level before the year 2000. Hardware componentsstart-ups
produce and sell such items as integrated circuitsand disks.Softwarecomponentsstart-
ups serve all computing power levels and deal in dozens of software segments.A few
of the offerings in systemsprogramming include languages,operating systems,utility
programs, network management, and general software-engineering tools. Complete
computer systems manufacturers may create anything from voice-controlled,pocket-
size PCs to supercomputers.End-user applications software ventures bring us games
as well as programs for inventory control, word processing, and mechanical design.
Distribution, service, customization, training, and operations also constitute a major
segment of the industry.
But despite their variety, all healthy companies starting up in the information-
technologyfield must pass throughthe followingfour predictable,measurable,sequential
growth stages in a roughly identical fashion:

Stage I: Concept
Stage 11: Seed
Stage 111: Product development
Stage IV: Market development

These four stages correspond to key product, market, and corporate development
milestones and are intentionally distinct from a definition based on the infusion of
capital (i.e., the rounds of funding).
Assuming they have successfullymaneuvered through the preceding four stages,
companies then reach a fifth stage, known as steady state-a mature but still growing
 The Four Elements of the Bell-Mason Diagnostic 253

stage at which they are considered to be stable, solidly established, and sustainable
organizations.
This book has focused on the definition and analysis of the first two stages of
growth, concept and seed, when both the product and the market approach are
hypotheticalbut are undergoing detailed planning and development. Decisions made
during these stages are excellent predictors of the company’s performance in later
stages. In fact, the success of the entire venture is most often determined wholly at the
concept stage.

ELEMENT 2 THE TWELVE DIMENSIONS

THAT ARE MEASURED TO ASSESS A START-UP

The Bell-Mason Diagnostic enables the user to make a complete assessment of a

high-information-technology start-up by measuring twelve principal and relatively
independent dimensions. (Although the diagnostic is, as noted, geared toward hgh-
information-technology ventures, it might, if modified, be useful in evaluating other
types of companies,includingretailing.For example, ”location”could be substituted for
”technology” as one of the dimensions.)
The twelve dimensions are organized in four groups, each containing three
dimensions:

Technology/engineering (Chapters 5 and 6), manufacturing, (Chapter 7), and

product (Chapter 8)

Business plan (Chapter 3) and marketing and sales (Chapter 9)

CEO, top-level team, and board of directors (Chapter 2)

Cash, financeability,and operations/control (Chapter 4)

Thus, it should now be clear that each of the precedingchaptershas discussed either
one, two, or three of the dimensions.The dimensionsare designed to cover every aspect
of a start-up in a complete,independent,and nonoverlapping fashion, including input
(people, cash, financeability, and technology), output (product and service, and the
ability to produce and deliver products), balance sheets, the organization and people
who run the company, and finally, key processes.

ELEMENT 3: THE RULES USED TO EVALUATE EACH DIMENSION

Each of the twelve dimensions is evaluated at each of the four stages of a company’s
growth by comparing the start-up with an ideal for that stage. This comparison is
performedby havingkey participantsin theventureanswer a seriesof questions,which,
254 The Bell-Mason Diagnostic

in effect, constitute a checklist. The questions themselves are the rules that define the
"ideal." Thus, the companyis on track across all dimensionsif it answers "yes" to all the
questions.
Fully (or at least nearly) achieving the ideal values at one stage is a necessary
prerequisite for the firm to advance successfullyto the next stage. If a company fails to
satisfy the requirements of a rule (i.e., by answering "no" to any question at a given
stage),it will probably have to correct the situation at a subsequent stage. Thus, those
managing the start-up can choose to "pay now or pay later."

ELEMENT 4 A RELATIONAL GRAPH PLOTTED

AGAINST AN IDEAL MODEL FOR SUCCESS

Figure 10-1 shows each of the twelve dimensions as a spoke in a polar graph, with the
spokes separated by 30 degrees.Plotting the scores for the answers to the twelve sets of
questionsproduces the "value" for eachdimension.Thedimensionsgrow in value from
the center of the circleto its circumferenceas the companyprogressesthrough the stages
of growth.
The figure shows two of the four elements of the diagnostic: the stages of growth
and the dimensions that are measured. And as will be discussed below, this type of
graph can also be used to show the ideal model for success at each of the stages. This
enablesthe user to see at a glancehow a start-up's current status compareswith the ideal
values for all of the dimensions at a particular stage. The fourth element of the
diagnostic-the questions, or rules-does not appear on the graph but operates in the
background, permitting the evaluation of each dimension at a given stage.
Figure 10-2 shows how the ideal grows with each stage, as the company begins at
the concept stage with technology, a plan, a leader, and enough resources to get to the
seed stage and then progresses from there to the product development stage.The graph
reveals hot spots requiring attention by graphically portraying the organization's
strengths, weaknesses, and overall balance at each stage.

APPLYING THE DIAGNOSTIC

TheBell-MasonDiagnosticcan be employedin severalways, as described in the Preface.

The three most common uses are:

As a template, or reference, for planning a high-tech venture

As a tool for performing a diagnostic "outside review" or "self-assessment" of a

company
 The Four Elements of the Bell-Mason Diagnostic 255

Business plan

Stage IV

Technologylengineering

Operationslcontrol

FiUlXlMICdmUU&OU

Financeability Board of
directors
Cash

Figure 10-1. Relational Graph Used to Measure the Twelve Dimensions of a Start-up at
Each Stage of Its Growth.

Business pian

Cash

Figure 10-2. Relational Graph Showing the Status of an Ideal Start-up at the End of Each
of Its Four Stages of Growth.
256 The Bell-Mason Diagnistic

As a means of developing a prescription for change to achieve a more ideal

organization

Whatever the ultimate application, if the user becomes adept with the concepts
underlying the Bell-Mason Diagnostic,gains sufficientexperienceor knowledge of the
industry, and then applies common sense, he or she is likely to significantlystrengthen
the start-up’s position.
Although the diagnostic attempts to be resistant to the destructive effects of
ignorance and denial, which pervade many start-ups, readers should keep in mind that
the method is only as good as the people answering the questions and the people
evaluating the answers. For example, a company may have a business plan that meets
the diagnostic’sstandards with respect to content, but that content may nonetheless be
fatally flawed because the analytic work is poor or the staff has an insufficient
understanding of some key issue. Thus, it is possible for a firm to obey all the rules but
still fail because the quality level of the organization and/or its product is too low.

THE FIVE STAGES OF COMPANY GROWTH

The accompanying flowchart (Figure 10-3) illustrates the stages of growth for high-
information-technologyventures, together with the possible outcomes for each stage.
These are the same stages that were shown in a computer-program format in Chap-
ter 1.Note that three of the substages of market development also appear in the figure.
This section examines each of the five stagesin detailand closeswith a discussionof the
different ways in which a start-up may transition from stage to stage.

STAGE I: CONCEPT (0-? MONTHS)

The concept stage is the company’s starting point. It takes nothing to enter this stage
except a kitchen or dining-roomtable at which to sit and begin exploringand planning
the proposed venture.Participantsat this stage usually include one or two players who
want to develop an idea they have for converting some technology into a product.
The product might be targeted for a market that did not previously exist, as in the
case of a newly emerged market. Or it might simply be aimed at a niche of an existing
market, such as a performance- or cost-oriented segment of that market. If the product
represents a significant improvement in performance or price, the start-up may target
it as a replacement in an established, growing, main-line market.
Theconceptstagecanbe initiatedfromanyviewpoint-such asmarket,technology,
or product-but it requires the drive of a core group who have been infected with
entrepreneurial fever. Ideally, the founding team includes a CEO who is capable of
carrying the team through to stage V, steady state.
 -
-Build
+Build
a market model
a product model-bO--Build a product-
-Test
+Build
the market model&
the m a r k e t e B u i l d the company-

to product
C L L
Concept
$, directly fund start-up
0, project slip

'a
S. accelerate marketing

$? ...
n
$, accelerate growth with
mezzanine

to market Acquired)

I
to company
looks
feasible.
Why redesign recalibrate
not start a product business
company?
Kill License or Kill sell Die Acquisition Die Acquisition Die Merger'
find alternative prototype of company acquisition
way to exploit or find o r technology
technology alternative a n d o r product
or product way to exploit state...
technology walking dead

Figure 10-3. Flowchart of the Stages of Growth for a Start-up, Including the Criteria for Moving Among the Stages.
258 The Bell-Mason Diagnostic

The players remain in the concept stage for a few days to as long as a year. They are
"self-funded until they develop a skeletal plan and secure the funding to move either
to stage I1 (seed) or directly to stage I11 (product development) and begin staffing the
organization.

STAGE II: SEED (PLANNINGTHE COMPANY) (3461-12 MONTHS)

The purpose of the seed stage is threefold:

The entrepreneursmust ensurethat any criticaltechnologyis under controlin order

that stage 111(product development) can be scheduled.

They must create a cursory product definition so that the market can be assessed.

Theymust producea realisticbusinessplan, which ties costs and revenuestogether.

The seed stage lasts six months on the average. It can take over a year, however, to
prove technology/ product efficacy if the proposed company utilizes a particularly
difficult technology.
Although not all high-tech ventures go through a seed stage, it is strongly recom-
mended' that they do so in order to allow for the formation of a first-rate team and the
development of a detailed, high-quality plan for the company (as described in Chap-
ter 3). If the founders receive a large infusion of cash with which to start the firmmore
rapidly, they tend to skip the rigors of this critical planning and technology-solidifg
stage and instead redirect their attention to hiring people. Although the seed stage is
vital, it is also a difficult stage because potential employees want to know that an
organizationis properly funded before they agree to join and because an extra round of
financing means a further dilution of the founders' stake in the enterprise.
The technology and product feasibility are validated during the seed stage by
creating a breadboard of the product or a critical part of the product, together with a
product specification and a model of the corresponding target market(s1. A formal
business plan is prepared, as is a plan for stageIII (productdevelopment),with the latter
plan detailing resources, specifications,and product development schedules.Funding
is securedfor the entireproduct developmentstage,in accordancewith the advicegiven
in Chapter 4 and the answers to the key questions about financeabilitypresented there.

1 . January 5, 1990, I examined a well-written plan for $5 million that would have taken the proposed
company from stage 1(concept)directly to stage I11 (product development). I urged against it. The next
day, I found a critical technological breakthrough on which the entire product was predicated. Although
the necessary technological issues could have been examined by one or two people during a three-month
seed stage,attempting toassess thosesameissues withalargestaffattheproduct development stagewould
have been hopeless and would have led to compromising the product.
 The Five Stages of Company Growth 259

STAGE III: PRODUCT DEVELOPMENT(6-[23]-37 MONTHS)

The goals of the product development stage are to hire the staff, speclfy and plan the
product, and design and produce the actual working product. During this stage, the
product must be tested for several months under actual operating conditions by a
reasonable number of real users. (Theactual number of beta-test users depends on the
product’s cost and volume.)
The entire product development stage takes an average of just under two years,
with entry into the stage being marked by securing funds and exit from the stage being
marked by the existenceof a working and user-tested product. This stage consistsof the
following four substages, which correspond to the four main product development
phases (describedin Chapter 5, Table 5-4):

Substage IIIa: hiring and planning (0-131-6 months): The development team is hired
and then generates a detailed plan and product specification.

Substage IIIb: designingand building (4-1141-24 months): The product is designed and
built.

SubstageIIIc: alpha testing (1-131-5 months): The product is formally tested in-house,
under conditions as stringent as those of actual use.

Substage IIId: beta testing (1-131-5 months): Sinceit is highly unlikely that any product
will be flawless enough to be shipped without extensive testing and acceptanceby
the intended users, the product must be delivered to a number of actual users for
testing. (Producttesting by relatives and friends does not count.) Although initial
beta testingcanbefacilitatedbybringing the first usersintothecompanytoevaluate
the product on-site, the product must also be shown to work in the users’
environment.The product will then have to be modified as necessary in light of the
test results.

Following beta testing, the detailed plans for producing and marketingthe product
are created, and funding is secured, if necessary, for introducing the product into the
market.

STAGE I V MARKET DEVELOPMENT(2431-4 YEARS)

The market development stage is the culminationof all the work done in the preceding
stages.It is the period during which the planning performed in stages 1-111 is tested and
then tried out in the marketplace. Although the firm‘s ultimate fate often becomes
apparent during this stage, the seeds of successor failurehave already been sown in the
earlier stages, when the product was designed and the marketing plans were made.
260 The Bell-Mason Diagnostic

Entry into this stage is marked by the first customer shipment, and exit is usually
marked by company acquisition or IPO (initialpublic offering).Once the product has
proved itself with internal (alpha)and external (beta) testing in stage III, the start-up
must begin spending significantly more money to produce, market, and sell the
product. The rate of expenditure typically triples when the firm enters the market
development stage,provided that inventory costs canbe kept to a minimum.The three
substages of market development (discussedfurther below) include calibration of the
existing market model, expansion of the market to reach a break-even point, and
operation of the company at a profit for a minimum of six quarters. Just as product
developmentis the stageat which the technology-to-producttransition (i.e.,the product
plan) is tested, market development is the stage at which the product-to-market (use)
transition (i.e.,the market plan) is tested.
During the first three stages of a company’s existence, the size of its staff is limited,
which tends to minimize expenses. It is therefore relatively easy for the firm to appear
”incontrol”even though no output is being produced. In themarketdevelopmentstage,
however, the profit and loss statementhas lines that directly relate to producing revenue
at a planned level at some future time. These items include product cost and all the fixed
and variable sales and marketing expenses, such as advertising, salespeople, support,
installation, and service. The first sign of failure to meet the plan‘s “bottom line” often
shows up right away in the ”top line”-i.e., the revenue is not present. When this
happens, a number of the intervening expense lines must be cut instantly in order to
meet the bottom line. Otherwise,the organization gets sigruficantly”off plan,” with the
inevitable need for ”another round of financing and the resulting dilution of owner-
ship.Thebottom-linefailuresthat affectmost start-upsare actually a directconsequence
of failure to meet the top line-i.e., the sales plan. Most high-tech ventures suffer from
a top-line problem at some point.
The three substages of the market development stage, mentioned briefly above,
include:

Substage IVa: market calibration (3-161-9 months): This substage is entered with the
initial shipment of the product to customers and is the first time every line item of
the business plan is finally tested. During this product/market calibration, or
market-beta-testing, phase, which lasts an average of six months, the product is
introduced into the market and the product, market, pricing, and sales plans are
modified as needed until a refined plan for profitability is arrived at. The company
adjusts its fixed spending in engineering, marketing, manufacturing, and admin-
istration to meet the unit variable product and salescosts, so it can move toward the
”break-even” point. The major purpose of the market calibration phase is to
determinethe product’saverage selling price and its cost of sales, together with the
order-gestation time.
 The Five Stages of Company Growth 261

Substage IVb: market expansion (6-491-12 months): In this substage, which lasts an
average of nine months, the firm continues to calibrate itself and moves, under its
refined plan, to achieve its first break-even quarter, the exit criterion for this
substage.

Substage IVc: steady-state operation (18 months): During the final substage of market
development, the start-up demonstrates that it can run profitably by sustaining
steady-state profitable operations for six quarters.

Sustainingprofitability implies that a successfulcompany has been formed.At this

point, the firm has a number of options. It could remain in its current state while it
continues to build stature. Alternatively, it could ”cash out” in some form, with its
foundersand investorsachievingfinancialliquidityby sellinga portion of the organization
to someone.Virtuallyall start-ups aim toward the finalfundinground’sbeing the initial
public offering, which keeps the company independent while enabling founders,
funders, friends, and key employees who have invested in the firm to finally receive
value for their efforts.Although,ideally, the company reaches the steady-statestageby
going public and remaining independent, most hgh-tech ventures in fact end up being
acquiredby another firm.For example,in 1989,18personal-computer-relatedcompanies
went public at a valuation of $300 million, while 149 companies were acquired at a
valuation of $2.1 billion.
Although “cashing out” is a declaration of having entered the steady-statestage, a
healthy company may choose to remain private and profitable, thereby entering steady
state surreptitiously. In the case of a privately held firm, the main investor issue is the
ability to provide a return on the investment.

STAGE V STEADY (SUSTAINING) STATE

Although in substages IVb and IVc, the company sustained steady-stateoperations for
a period of approximately two years, it lived a relatively sheltered existence, beyond
public scrutiny.InstageV (steadystate),however, thegoalis todevelop theorganization
in such a way as to ensure its “immortality.” This stage requires continual strategic
maneuvering, whereby the firm attempts to retain and consolidate its niche in every
aspect of its operations, including technology, product, market, service, business plan,
finance, operational style, and culture.
There is a chilling alternativeto a healthy, dynamic,and fully mature steady s t a t e
namely, the companymay ”go public,” only to settleinto a stagnant steady state (known
perversely as ”the state of the walking dead). In this condition, the venture cannot
attract additional funding and is not viable for more than a few years, since it is unable
either to maneuver into permanent and sustaining product and market niches or to find
262 The Bell-Mason Diagnostic

an alternativeway of operating in the long term. Single-product firms are likely to end
at this point, unable to evolve a next product or create a unique and permanent way of
doing business.Even when such a company has a plan for permanency, the public may
perceive it as ”stuck,”with no way to financeitself, and thereforenot worth investment
or speculation.In this situation, the only recourse to death is some form of merger.
Hence, when1speak of a company’shaving successfullyarrived at the steady-state
stage, I am not referring to the creation of a stagnatingorganizationthat enduresmerely
through momentum but to the creation of a healthy and enduring organization that
operates in such a manner that indefinite growth and profitability may reasonably be
anticipated.

TRANSITIONING FROM STAGE TO STAGE

The transition from stage to stage is usually linked to the requirement that additional
funds be obtained to carry the start-up to the next stage of growth. Funds are also
required when the firm misses its product or market development plans and has to
remain in and loop within a particular stage. Thus, any company, within any given
stage, may choose, or be forced to choose, one of the following options, listed in order
of severity of consequences:

Move to the next stage: The firm progresses to the following stage in its ideal growth
pattern, but with some inevitable dilution of ownershipas stock in the company is
traded for funding to achieve the next stage of growth.

Loop within the current stage: The venture must receive more funds (i.e., obtain
anotherround of financing)and remainat thecurrent stageuntilitgetsbackonplan.
Increased funding usually means increased investment and therefore greater
(possibly complete) dilution of ownership for all the current investors.

License the technology/product to another company: The firm uses licensingas a means
of funding the current stage and thereby getting on the road to success.

Return to an earlier stage: The start-up backtracks without achieving the objectives
of the current stage. For example, a product recently introduced into the market
may be found to be fatally flawed and must then be redesigned from scratch.

Haveitsassetsacquired by,ormergewith,anothercompany:Thefirm turnsover itsassets

(including technology, capital, people, products, equipment, and buildings) to
another organization and ceases operating as an independent entity.

Cease operation: The firm sells any assets.

 The Twelve Dimensions That Are Measured to Assess a Start-up 263

THE TWELVE DIMENSIONS THAT ARE MEASURED

TO ASSESS A START-UP

Innumerable factors, large and small, indigenous and exogenous, influence the course
of a high-tech venture. These can be distilled and categorized into only twelve key
elements,or dimensions(shownin Table 10-l),which determine the firm’sultimate fate
in the marketplace. By using the Bell-Mason Diagnostic’s rules to evaluate the strength
of each of these dimensionsat each stage of growth, the start-up’shealth can be assessed
and its future outcomepredicted and managed. In effect, the organizationis compared
against an ideal. Of course, the very process of conducting the assessment (Le.,
identifying and carefully scrutinizing critical issues) is likely to have a significant
positive impact on the start-up’s outcome.
Because all twelve dimensionsare important, they are all given equal weightingon
the relational graph (shown earlier in Figure 10-1).Achieving superiority in only one
area, such as having the best people or producing the best design or even reaching the
market with the best overall product, is simply not enough in the competitive era of the
nineties.
Many catchy formulas have been offered for how to start a successful high-tech
business. One of the earliest venture capitalists, Arthur Rock, characterizes the
entrepreneur’s traits as follows: “a burning desire to start a company. . . . A person has
tobevery,very honest.. .recognizeproblems,foreseeproblems,recogruzeshortcomings,
and admit and learn from mistakes.” Rock reduces the whole issue to ”People,people,
people,” while others advocate a more balanced, but still simplistic, maxim: ”People,
product, plan.” Poduska, who believes that great people will rapidly adapt to any
situation, states his belief in people over product or plan like this: “ ‘ A people with a ’ B
plan beat ’ B people with an ‘ A plan.” In contrast, Don Valentine, the head of Sequoia
Capital, has no fears about recruiting or replacing key people in a start-up because he
looks for “markets first, products next, and then people.” Bob Keeley, a Stanford
professor who has studied start-ups,believesthat without a very good first product, the
company is likely to fail because it will run out of time.
Whenever stories of business success or failure are told, almost invariably, a
simplistic formula like one of the aforementioned will be cited as the moral of the tale.
The modern entrepreneur must avoid such maxims, no matter how clever they are or
how reliabletheir source or how true they may once have been, sincenone of them even
begins to capture the challenge of the contemporary high-tech venture. Such over-
simplificationsdeemphasize all sorts of critical factors, including the need for cash, the
ability to control the organization during a period of rapid growth, having the right
product before the start-up becomes just another company producing a commodity
product, and the complexitiesand challengesof competingwith a plethora of firms that
are being founded to produce what will become a high-tech commodity.
264 The Bell-Mason Diagnostic

Table 10-1. The Twelve Dimensions of the Bell-Mason Diagnostic.

~ ~ ~ ~ ~ _ ~ . _ _ _ _ _ __ _
_ ~
TechnologylProduct MnrketinglSales People FinancelControl
Technology/ Business plan CEO Cash
engineering and vision
Team Financeability
Product Marketing

Manufacturing/ Sales and Board Operational

product delivery product support of directors control

THE RULES USED TO EVALUATE (SCORE) EACH DIMENSION

Organizations are not subject to universal physical laws like those that govern much
technology.Instead,start-upshave to conform to the laws of moral and ethicalbehavior
and of governments.None of these ”contractual”laws determines whether a company
will be successful,although violating any of them will almost certainly spell its doom
at some future time.
”Laws of good practice” come from observation and result in “heuristics,” used
herein to define the ideal start-up. Each of the twelve dimensions is evaluated against
these laws of good practice at the firm’s point of transition from one stage of growth to
the next. The evaluation is performed by applying a series of rules to each dimension,
with the rules taking the form of a set of specificquestions.In the diagnostic,all the rules
are weighted equally to ”score” a dimension. In practice, however, the rules will be
given varying weights to reflect the difficulty and criticality of each issue (such as the
existence of a plan).
The relationship among the laws of good practice (i.e., the heuristics based on
observation),the rules or requirements of behavior that an ideal start-up should satisfy,
and a question to which the organization can answer “yes” or “no” is illustrated in the
following example, which shows the development of a diagnosticquestion that can be
asked in the course of evaluating a company’s technology:

Heuristic based on observation: Software-engineering experimentation has shown

that if a firm uses an inspection process in which one or more engineers examine,
or ”walk through,” another engineer’s programs, the resulting product will have
fewer errors. Although this law applies to “average” software engineers, a few
exceptionalprogrammers may produce correct codesby themselveswithout such
a formal review process.

Rule: The engineering organization must have a design-review process that

in’cludes code walk-throughs, inspections, or some other rigorous method of
verifying a design.
 The Rules Used to Evaluate (Score)Each Dimension 265

Diugnostic question (theprecedingrule, rephrased as a question):Does the engineer-

ing organization have a design-review process that includes walk-throughs, in-
spections, or some other rigorous method of verifymg designs before they are
integrated and become part of the final system?

Each "rule" is stated in the form of a question,phrased so that a simple "yes" or "no"
represents a "pass" or "fail" with respect to a particular issue when a dimension is
evaluated at the start-up's point of transition from one stage of its growth to the next. For
example, in order for the engineering organization to begin designing the product in
detail, the company must be able to answer the following question affirmatively: "Is
there agreement between engineering and marketing on the product (performance,
feature set, function, and cost) and schedule?"
Ideally, all rules must be adhered to (i.e., all questions must be answered in the
affirmative)before the start-upproceeds to the next stage.Failing to adhere to a rule (i.e.,
answering a question in the negative) at a given stage can have different implications,
however, depending on which rule is involved and why it cannot be satisfied:

If the rule is critical and the question cannotbe answered affirmatively,the venture
is likely to fail. (E.g., at stage 111: "Does the product work according to market
expectations?")

If the rule is critical and the next stage in the growth process hinges upon adhering
to the rule, the company is likely to remain in limbo until the question can be
answered affirmatively. (E.g., at stage IIIc: "Does the product work satisfactorily
during in-house testing so that it can be tested by real users?")

If the rule is so hard that virtually no start-up acheves the ideal, the company can
safely proceed to the next stage if it is doing at least as well as could be expected from
the average firm.

If the rule is irrelevant for some reason and can therefore be disregarded in the
scoring, the organization can safely proceed to the next stage. (E.g., at a software
company, manufacturing, though important, is low-tech and almost inconse-
quential.)

If the firm does not adhere to the letter of the rule but has found a better way of
adhering to the spirit of the rule, it can safely proceed to the next stage. ( E g , it hires
only "perfect" people.)

Each question should be answered with care. If time permits, the evaluation could
also measure the quality of the work, movingbeyond simple "yes" or "no" answers and
assigning grades. The transitional diagnostic questions are designed to elicit informa-
tion about each dimension at the level of detail required to effectivelybring the product
266 The Bell-Mason Diagnostic

to market and the firm to steady-state operations. These are sharply focused, hard
questions-precisely the sort of questions that a CEO or board should want the
organizationto address honestly. Such questions as "Does the company have a market
plan?" or "Is the product sound?" are too vague to bring the criticalconcerns facing the
start-up into clear focus so they can be properly addressed. In contrast, the diagnostic
focuses the issues sharply by using more detailed rules (i.e., by asking more detailed
questions),which permit a more meaningful assessmentof the critical issues. Here's an
exampleof an effective,sufficientlydetailed diagnosticquestion:"Have design reviews
for each critical milestone of the project been included in the scheduleand adhered to?"
The rules for each dimensionbecome more stringent with each stage in a start-up's
growth. Note how the product development rules (questions)evolve as the company
progresses through the following stages:

At stage 1 (concept): "Does the company have evidence of product concept pos-
sibilities, given the technology, that customers are likely to buy?"

At stage I1 (seed): "Does there exist a simpleproduct development specificationwith

features and functions that can be presented to potential users?"

At stage111(productdevelopment):"Are an appropriatenumber of beta-stage systems

(at least three for large systems and at least twenty for mass-user software)
operating in real user environments and demonstrating unique capabilities or
providing users with significant performance/price benefits?"

Notice that several of these questions have multiple parts. In such cases, each part
must be answered affirmatively according to the logic of the question in order to
determine the start-up's readiness to progress to the following stage.

THE OUTPUT
THE RELATIONAL GRAPH AND MODEL FOR SUCCESS

The relational graph, a type of polar coordinategraph shown earlier in the chapter (see
Figure 10-l), is also known as a Kiviat graph? Since the graph displays both the four
stages of growth and the twelve evaluation dimensions, it enables the user to quickly
assess a company by examining all twelve dimensionssimultaneouslyat a given stage.
When the relational graph is employedas a managementtool, it permits areas of concern
to be pinpointed so that problems can be corrected.

2. TheKiviat graph was first used to plot various dimensions of computer-systemsperformance. The polar
graph is commonly used in Japan, where it is taught in secondary school.
 The Basic Rules for Diagnosing a Company 267

The graph is formed by plotting in the area between the four concentric circles to
denote the state of a given dimensionat each of the four stagesof growth. At every stage,
the value for each dimension should lie in the range between the circles for the previous
and current stages. Since a company may not have completed all the requirements for
a particular stage, it is possible that the value for one or more dimensions may lie
somewhere within a previous stage. Such a discrepancysimply indicates that the firm
is underdeveloped in some dimensions for the stage in which it purports to be.
Once the twelve dimensions are plotted for a particular growth stage, they can be
compared against the idealrelationalgraph for that stage.Figure 10-2, presented earlier,
shows the ideal state for each of the four growth stages, as a requirement for each
dimension, and illustrates how the dimensions grow as the stages progress. The three
outlines plotted on the figure represent the state of evolution of each of the dimensions
that is required at the concept, seed and product development stages. Ideally, as the
company grows, each dimension evolves to meet the target standard for the current
stage.
In Figure 10-2, points plotted at the circumference of the circle for a given stage
represent the dimensions of greatest importance at that stage. For instance, at the
concept stage, the four dimensions that form the axes-the business plan, CEO, cash,
and technology-are the most critical. Other dimensions,such as sales and product, are
less important at this stage.All dimensionsshould be fully evolved and lie on the outer
circle by the time the company reaches steady (sustaining) state (i.e., becomes an
established firm)at the end of stage IV.
To cite another example, the emphasis during the product development stage is on
growth in the engineering organization and development of the manufacturing orga-
nization.Thesechangesare in preparation for the market developmentstage.Although,
during this stage, the marketing plan is also being developed and other activities are
taking place, they are not receiving the same attention as engineering (product devel-
opment),which is critical to the current stage. The product development stage is only
concluded when the product has been successfully beta-tested, such that it can be
introduced into the market for sale, which will occur at the beginning of the market
development stage.

THE BASIC RULES FOR DIAGNOSING A COMPANY

Each of the dimensions is, in effect, defined by the rules, or questions, that form the
diagnostic. That is, an ideal start-up will satisfy all the rules that define all the
dimensions. Figure 10-4 gives twenty-five rules (in the form of evaluation questions)
that will help the reader better understand each dimension.These rules can be used to
evaluate a company broadly and superficially, but quite easily, at any time.
268 The Bell-Mason Diagnostic

Technology/Engineering
Does the company have a fundamental,defensible, and measurably superior technology,
as indicated in its “technology balance sheet,” that enables the sustained conversion to
products by an engineering group of proven capability? Does the “technology balance
sheet” include the following dimensions?

Technology base

Standards

Design, quality, and other processes

Plan, with schedule and resources

Engineering specifications

Manufacturing specifications

Chief technical officer

Team and engineering culture

Architecture

Technical resources

Technology future

Operational management

Product
Does the product have well-defined and unique features, functions, and benefitsto support
the price and match the competitive market requirements?Can the company buildthe next
generation of follow-on products?

Manufacturing
Does the company have a well-defined organization and processes to produce products
at the cost, quality, and schedules required by its customers?
Does it manage its raw materials and finished goods inventories in an optimal fashion
according to just-in-time principles?
Does it introduce products into manufacturing rapidly, accompanied by clear product and
process specifications?

Business Plan and Vision

Does the company have a written five-year plan that is working and realistic and that
emphasizes (Le., spells out in particular detail) the plan’s first two years? Does the plan
provide an integrated overview of all aspects of the firm and specifically identify the
following: corporate vision and mission, lasting technological advantage in terms of a
“technology balance sheet,” product strategy, market segmentation and competitive
 The Basic Rules for Diagnosing a Company 269

market position, people and the reward structure, and the financial and financing
requirements?
Are resources and milestonesspelled out in the plan, and does the plan balance costs and
customersto give a realisticforecast of returns, as noted in the financial portion of the plan?
That is, quite simply, does the plan make sense?

Marketing
Does the company have a complete strategic and tactical market plan (both of which are
defined below), together with the leader and organization to implement it?
Does the strategic market plan cover the following topics?
What (the complete product)
Who (the buyers) and how they are reached via a "market map"
How (the manner in which the product will be applied, along with any missing
components needed to deliver a complete product to a buyer)
Why (the buyinghsing rationale, in terms of features, functions, and benefits)
Where the product will be sold-i.e., distribution channels
When (the time frame and cost model for selling and receiving the product)
Does the tactical market plan contain detailed information to support the marketing of
the product, including a definition of the programs, resource requirements, and
schedule?

Sales
Does the company have adriven sales group headed by a proven leader, and do the group
and its leader have the understandingof and experience with the product class, price, and
customers that will enable them to realize the selling cost and time model?

CEO
Does the CEO possess the level of intelligence, energy, ethics, and quality required to
establish the clock and culture for the proposed company?
Does the CEO recruit (help select and sell) great, critical hires?
Has the CEO demonstrated management,team-building, and leadership abilities involv-
ing product development, in a resource-constrained environment, and on a "do-it-from-
scratch" (i.e., start-up) basis, and is he or she likely to be able to manage the company
throughout all the stages of its growth?
Does the CEO attract capital, board members, key customers, and strategic corporate
partners?

Team
Is the top-level team composed of high-quality individuals with measurable experience
and expertise in the various areas? Are they capable of attracting grade-A personnel as
well as leading and managing their respective functions?
Is the team "do"-oriented rather than "management"-oriented-Le., can each of the
members "play" several positions on his or her team as opposed to just managing a team
of players?
Do the members function collectively as a team in an integrated fashion, as opposed to
operating as a collection of egocentric or warring individuals?
(continuedj
270 The Bell-Mason Diagnostic

(continued)

Board of Directors
Is the board composed of individuals whose experience and expertise enhance the
company's competence at its current and subsequent stages of growth?
Do board members act as reviewers,counselors, and company missionariesfor sales and
finance rather than behaving like corporate decision makers?
Is the board reviewingthe firm's strategic plans and direction as well as providing the CEO
with advice about current operations?

Cash
Does the company have enough cash to complete the current stage according to plan and
carry it into the following stage while it secures the next round of financing in concert with
its investors?
If the cash is below a three-month supply at the current rate of expenditures, can the
organization either obtain adequate cash from operations or seek extra cash through a
relatively predictable financing channel within that three-month period?

Financeability
Aremultipleinvestorswillingtocontributetothenextstageofthecompany'sgrowth, based
on the corporate, product, and market outlook for the firm, in the context of their feelings
about the economy, high technology, and the market sector?

Control
Is the company operating according to an overall plan, and are only a minimal number of
changes being made to that plan? (Have missed milestones, if any, been minor and
explainable?)
Does everyone in the organization operate according to a formal schedule and manage-
ment by objectives? Is everyone informed about the firm via effective staff meetingsduring
which review, direction setting, and problem identification/resolution take place? The
control dimension is verified by reviewingthe archivesof the team and of each department!
Are critical processes in place to govern spending and hiring so as to assure progress
against the plan in a controlled fashion that will produce high-quality output?

Figure 10-4. The Basic Rules for Diagnosing a Start-up.

CONCLUSION

High-tech start-ups are characterized by the growth of twelve dimensions throughout

four stages until the company reaches the fifth stage, steady-state. The preceding
chapter has described the 12 dimensions and the rules to evaluate each dimension for
a company in the concept and seed stages.
The Bell-Mason Diagnostic is designed to evaluate the dimensions and plot them
against an ideal at each stage using a twelve-dimensional relational graph. Diagnosisis
 Conclusion 271

carried out by answering questions that come kom heuristics or rules that define an
ideal company. The heuristics come from experience and understanding. The entire
diagnostic consists of over 600 rules, and the evaluation at a substage of product or
market development may embody 100 questions.
Twenty-fivequestions(rules),which will help you understand each dimension,are
asked in the diagnosticprocess. These rules can be used to evaluate a company broadly
and sufficiently,but quite easily, at any time.
Chapter 11

CASE STUDIES

This chapter provides a glimpse of how a number of actual high-information-technol-

ogy ventures have formed. The Bell-Mason Diagnostic is used, after the fact, to assess
the health of several of these companies at various points in their growth.
The first two examples come from software start-ups of the early 1980s:

Ovation: a now-bankrupt software firm that conducted an aggressive sales and

marketing effort but was unable to produce a product.

Analytica: a decision-support software company that had a good product but

conducted a poor marketing effort.

The next six examples involve firms that appear to be healthy and whose products
are representative of the types of products that a start-up might be founded to build:

Dragon Systems: a manufacturer of speech-recognitionsystems that used product

salesto fund the development of its technologyand its understandingof the market
as computers with sufficient processing power for speech recognition evolved.
Dragon’sstory providesan exampleof how a companycan startup ina self-funding
fashion and evolve slowly with and develop a very complex technology.

Civrus Logic: a $100 million,public ”semicomputercompany’” insiliconvalley that

builds complex chips for the PC industry.

1.A semicomputer company is a firm that supplies microprocessors or microprocessor peripherals in the
form of semiconductors.

272
 Ovation 273

Gutezuuy Design: an organization that began as a self-fundedcompany and is now

a division of Cadence, located in Lowell, Massachusetts. Gateway builds software
products to describe, simulate, and test complex digital systems.

Gensym: a small,privatelyheld,profitablefirm,locatedinCambridge,Massachusetts,
that builds G2, a real-time, expert system for process control.

MusPur: a SiliconValley company in the market developmentstage that is building

a massively parallel computing system for highly parallel applications.

Thinking Machines Corporation: a Cambridge,Massachusetts, firm that produced a

forerunner of the MasPar computer. Thinking Machines’ story serves to illustrate
how government funding, through research and product purchasing, can be used
to create, evolve, and understand the use of an innovative computer based on
massive parallelism.

The start-ups discussed in this chapter span hardware, software, and systems in
both the Boston and SiliconValley areas. I am on the board of directors of Cirrus Logic,
a user of Gateway’s Verilog product, an investor in Gensym and Thinking Machines,
and a friend of MasPar’s founders.

OVATION: THE CASE OF THE MISSING PRODUCT

Ovation was a company founded in 1982 to build ”the next generation of integrated
microcomputer software,” a development that many were saying was upon us at that
time.Ovation’sproduct was defined as a package that would seamlesslyintegrateword
processing,spreadsheet,database-management,and communicationsfunctions.It was
to have been marketed in volume to Fortune 1,000 companies at a price of $695, which
was comparable to the price of the most popular spreadsheet, Lotus’s 1-2-3. Healthy
funding was secured from choice venture capital firms.
Despite having gotten off to a strong start, Ovation stalled and ultimately failed in
stage I11 (product development).It declared Chapter 11 bankruptcy about two years
after its founding,having spent approximately$7 million without producing a product.
Since Ovation took a high-profile position in starting up, its failure attracted press
coverage. The question is, what went wrong?

PRODUCT DEVELOPMENT STAGE

The relationalgraph in Figure 11-1plots Ovation’sactual accomplishmentsagainst the

model of successfor stage 111(productdevelopment)at the time when the companywas
perceived to be at the end of that stage. The graph tells the whole story practically at a
274 Case Studies

Business plan

Technology/engineering

Operations/contro

Cash

Figure 11-1.Relational Graph for Ovation Showing a Lack of Product and an

Overemphasison Marketing and Sales at the End of the Product
Development Stage.

glance, clearly indicatingthat the dimensionswere out of balance with one another. The
product, as indicatedby its definition, had just reached the seed stage,but no technology
or development team was in place to build a product of inherent and protectable value.
Ovation’s prolonged inability to produce any tangible output eventually prompted
analysts to coin the word ”vaporware” for such products.

PLAN

The business plan was clearly not a dynamic control document. It was stuck at the seed
stage and reflected the realities of product schedule slips and rescaled forecasts.

MARKETING AND SALES

Insharpcontrastto theunderstaffingevidentin thetechnologyand product development

areas, however, the marketing and sales departments were completely staffed and
 Ovation 275

running at full tilt, producing all the sales collateral, demos, sales materials, and
communications programs. (In Figure 11-1,note that sales is substantially overstaffed
for the product development stage.) The group spent at a high rate, in hopes of
generating indeterminate future revenues on a product that didn’t exist and ultimately
never would. Themarketingandsales effortwasimpressive,asindicatedbythefactthat
Ovation had already won an award for having the most important and innovative new
product, and the approach should have proved effectiveif the actual product could
have been built.

PEOPLE

The CEO receives bad marks for his inability to get Ovation back on track once the
imbalance in its growth pattern became evident. The top-level group clearly was not
operating as a coordinated and unified team, with complementary functions making
steady progress. The board, in its capacity as primary reviewer and counselor to the
company, also failed to identify and rectify the problems before they turned into fatal
errors. This latter point is not surprising, since the board had virtually no outside
representationand was composed entirelyof the marketing and sales staff that the CEO
had drafted from his previous company, a Digital Equipment Corporation (DEC)
reseller.The head of engineeringwas not on the board and, by someaccounts,may have
been treated as an outsider rather than as a team member.

CASH, FINANCEABILITY, A N D CONTROL

Ovation’s substantial cash, more than what a normal start-up might have been able to
obtain, was spent aggressively at a time when the chances for future financing were
diminishing.The company’spoor track record for achieving any externally measurable
results (such as alpha testing, beta testing, bookings, or sales) negatively impacted
further funding. Control and operations were primitive in relation to where Ovation
was supposedto be in the productdevelopmentstage;hence, the firm was out of control,
and had been all along.

POSTMORTEM

Even with twenty-twenty hindsight, it is not certain that prompt intervention would
have saved Ovationor made it viable.However,carefuland constant monitoring, using
a method such as the Bell-Mason Diagnostic,would have served as an ”early warning
system,” guarding against the expensive eventuality of failure. In Ovation’s case, the
results of the diagnostic would have sounded an alarm almost at the very outset.
276 Case Studies

At the first signs of failure, Ovation had two viable options:

Cease operations and return any remaining funds to the investors because the
technology was inadequate to support the product. (However,I am unable to cite
a precedent for this course of action.)

Reduce the company to a minimal marketing effort until a product could be built,
which might have enabled Ovation to achieve an improved balance between its
technology (i.e., its ability to produce a product), its product definition, and its
future funding scenario.

Thus, Ovation is simply another case of Ken Olsen’swallpaper remover (described

in Chapter 5)-all concept and no product, which is too often a fatal flaw in start-ups.

ANALYTICA: FAILING WITH A GOOD PRODUCT

Analytica was formed by Eric Michelman and Adam Bosworth to build a decision-
support softwareproduct for the PC with an easy-to-usegraphicalinterface.Adam and
Eric had already worked together in 1982, when they explored a project-scheduling
program with Andrew Lehman, who ultimately built the program as Time Line. Since
Eric had been interestedin decision support from his student days at MIT, he and Adam
(who had designed eighteen successfulbanking applications)decided to develop such
a product.

CONCEPT AND SEED STAGES

In March 1983, the company was funded with $1.3million based on a plan for a revenue
growth of $60,000, $4 million, $15 million, and $34 million and an Apple I1 product
demonstration that Adam had built. The business plan was based on the wildly
successfulLotus model take nine months to build and introduce a product; spend lots
on marketing and sales; and charge a high price. Although everyone knew what
spreadsheets were, no one knew what a decision-supportprogram was, why one was
needed, or how to build one. The product, ultimately called Reflex, was initially
nicknamed ”What If and Why.”
Founding consisted of merging with a stalled company, Taurus, which was selling
a user-friendly interface shell for CP/M and DOS. At its inception, the plan of the
combined company (eventually named Analytical was to milk the existing Taurus
product and develop Reflex.
Taurus had a staff of about twenty, and its president became the CEO of the new
firm. The merged board consisted of four venture capitalists and the president; the
founders attended the board meetings. Some of the board members had operational
 Analytica 277

experiencebut no experiencewith softwareproducts.Taurus had to relocate to be with

the six-member Analytica team that was building the company’s future product. The
head of engineering,Brad Silverberg,an engineer on Lisa, came from Apple with Eric.
The CEO, two engineers,and three members of the support staff from Taurus remained
with the company after a year.
Figure 11-2showsthe relational graphsfor the Analytica start-upat the concept and
seed stages.Notice that at the concept stage, the diagnosticis able to pinpoint the flawed
business plan, given the newness of the product concept, the lack of specification,and
poor market understanding.The lack of a CEO also showsup right at the concept stage.
By the end of the seed stage, when the venture capitalists created the merged
company,the firm was already in serious trouble.Clearly,there was no vision and plan
for the company as a whole, since Analytica was nothing like what the founders had
originally proposed and the new president did not prepare a new plan for the combined
organization.Whereas marketing was behind in planning, sales was ahead, in view of
the fact that it was assembling a sales staff.The board at this point was OK, even though
its members lacked software experience and did not act conservatively with respect to
control.There was as yet no indication that the firm was out of control, except that the
merger with a failing company, by definition, meant that the new organization was
unable to meet its plans. The lack of a clear product specificationat the seed stage was
a sign that both the technology/engineering and the product had problems right from
the start.

PRODUCT DEVELOPMENT STAGE

With the company in full swing,everyonehad ideas for the product, includingthe fully
staffed marketing and sales groups; the new, strong engineeringteam; and the board.
The product specificationfor Reflex advanced the state of the art in the graphical user
interface;in databasedesign,includingimplicitdata typing; and in the ability to add and
subtract fields within the database. The product also had a variety of analytic powers,
with the ability to do charting.For example, it could make a chart by extrapolatingfrom
a historical database.
However, for a product with such a plethora of features, Reflex had a weak
underpinning because of the decision to run the program ”in core” instead of working
with a virtual memory environment.The fact that the program had to reside in memory
at all times significantlylimited the sizeof the applicationsand theunderlying database.
This was almost, by definition, a fatal flaw from the standpoint of potential corporate
users (particularlysinceAnalytica planned to market the productaccordingto the Lotus
model) and was therefore a misinterpretationof the market. After a year, Eric and Brad
were finally granted explicit control over the product specification.
Given the program’s size and complexity, the fact that it constantly grew in
functionality, its lack of definition, its complex but undefined graphical user interface,
and the naivet6 of everyone concerned with the schedule, it was inevitable that Reflex
 Business plan Business plan

Figure 11-2.Relational Graphs for Analytica Showing the Concept and Seed Stages.
 Analytica 279

would be late and could not be accurately scheduled. At least one of the developers
became so frustrated by the uncertainty that he left the company. Adam later estimated
that Reflex could have shipped six months earlier if charting had been omitted from the
first release.
In view of the product slips, the board regarded the team as grossly incompetent
and put incrediblepressure on its members.Eric made the followingobservations about
the schedule:

We really didn't know it was a two-year project. It was easier to deliver the bad schedule
news in six two-monthslips rather than a year. In retrospect, the development team did
an amazingjob by getting the product out in two years.

Two years is the standard gestation time2 for complex software products, especially if
they've never been done before.

Adam offered the following comments about a company's first software product
and its evolution:

The first product doesn't need to do everything;it only needs to do somethingthat a user
needs.It has to be incrediblyfocused,and thedevelopmentteam must understand it won't
solve all the problems of the world. For example, Paradox's first release was poor, its
secondreleasewas good,and the third one,great.Paradoxis a $50-million-a-yearproduct.
Like Analytica and Reflex, ANSA sold Paradox to Borland.

In April 1984, the CEO was replaced with a former divisional general manager of
Data General who had no experiencein developing or selling software products. At this
point, the company required its second round of financing. Despite the fact that the
product was behind schedule (it had not yet started internal [alpha] testing), the firm
continued to grow until it reached forty-four people. The head of sales, who came from
Lotus, was hired to introduce Reflex using the 1-2-3 model. Reflex's product position
was changed by Dan Rosenthal,a marketeer, who redefined the product as an analytic
database instead of a new class of decision-support tool. This repositioning made Reflex
comprehensible to potential users and gave them a basis for comparing it with other
products.
In 1985, the protracted third round of financing occurred, bringing the total
financing to about $7 million. Marketing, training, support, and sales personnel were
now being let go, because it was costing the company about $400,000 per month to stay

2. Ashton Tate's Dbase IV version 4.0 was shipped in October 1988 with bugs and was unusable. Version
4.1 was finally shipped in the summer of 1990. In the process, the company's president lost his job. Lotus
1-2-3 version 3.0 was almost two years late from its announced availability date, and Windows 3.0 from
Microsoft, which was introduced in the spring of 1990, was over a year late.
280 Case Studies

in business while the seven developers continued working on the product. In March
1985, Reflex shipped and attained a level of two hundred units per month, which was
far short of the plan. However, Analytica was clearly running out of money, and the
venture capitalists were fatigued. Adam approached Borland, and the firm was sold in
exchange for a few million dollars worth of Borland stock. The common stockholders
and first-round investorsreceived nothing. In October, the engineeringteam moved to
Borland and were given Borland stock options.
Adam left the company in 1989 for Microsoft and Brad did likewise in 1990, while
some of the team still remained at Borland. When Analytica closed, Eric became an
independent software developer, operating with independentbut highly experienced
software developers in a ”distributed garage-shop fashion” to develop Silverado, an
add-in database for Lotus 1-2-3 that was sold to Computer Associates, and Budget
Express,which was sold to Symantec.In 1990,he built and tested a work-group product
to sell to a software publisher.
Figure 11-3showsthe obviousshortcomingsof Analytica’sposition at the end of the
product development stage. When the product was first introduced, the engineering
group had a schedule fantasy factor (SFF) of over 2. The product wasn’t what was
promised, but it was stdl interesting and clearly salable. The business plan was at risk
because there had been no market calibrationfor an enormous PR machne, nor did the
companyhave the financingto proceed with a high-visibility,Lotuslikemarketingplan.
Sales expenses were always running ahead of plan. By now, there was clearly enough
blame to go around: two CEOs had failed in every respect, and the board had allowed
the organizationto operate for almost two years in an out-of-controlfashion.Although
Analyticahad cash when Reflex was introduced, it was clear that the venture capitalists
were tapped out and would close the firm down if the product failed to take off.

POSTMORTEM

Once Reflex was in Borland’s hands, the picture changed dramatically. Borland
immediately dropped Reflex’s price to $100 at a gross margin of 40 percent, and during
the first year, about 200,000 units were sold.
Adam and Eric believe that, in hindsight, the $1.3million first-round funding was
more than adequate to develop and introduce Reflex. They both give similar advice to
programmers who want to produce a product that can be developed by a small team:

Don’t use venture capital.Start small,build fromyour own equity,ship ASAP, and evolve
the product. Don’t believe your own b.s. Don‘t fight for elegance because you are paying
for it yourself.
 Analytica 281

Business plan

Technology/engineering

Operations/control

Cash
Figure 11-3. Relational Graph for Analytica Showing the Product Development Stage.

Furthermore, the company was constantly out of control. Since the board must
insist on control, and control is ultimately the responsibility of the CEO, both failed in
Analytica’s case. Across the board, spending must follow the development schedule
and not lead it (ashad happened in the caseof the founding)or be independentof it. And
it is pointless to hire a sales department until product testing begins.
A company must constantlyreviseitsplans wheneverit failsto meet any milestone.
In the product development stage, spending and control must be locked to the
availability of the product, not to an abstract plan.
Although Analytica’sbusiness plan was flawed,the fact that what emerged as the
product ultimately was able to be marketed successfully using a different approach
proved that the initial marketing effort was similarly flawed and that sales, which
should have performed a check on marketing, also failed. With better communication
among the entireteam,Analytica might havebeenable to understand and market Reflex
properly.
282 Case Studies

The bottom line: Analytica is a classic case of an out-of-controlcompany in which

expenses were running ahead of product availability. The development team built a
very good product, but it was a serious mistake to have tried marketing it according to
the Lotus formula. In view of the fact that Reflex was ultimately successfulafter being
correctlymarketed, the product really can't be faulted-only everyone associated with
marketing and selling the product and managing the firm.

DRAGON SYSTEMS, INC.:

SELF-FUNDING RESEARCH WITH PRODUCT DEVELOPMENT

JimandJanetBaker met at RockefellerUniversity and, as graduatestudents,had already

earned reputations in the speech-research community. In 1972, they switched to
Carnegie-MellonUniversity (CMU),which had more equipment to support research.
Threeyearslater,they received their I'h.D.'s fromCMU'sComputer ScienceDepartment,
where they worked in the speech-research laboratory.
After leaving CMU, the Bakers worked at IBM Research, in Yorktown, New York,
for three years. They then spent the next several years working in the Exxon Office
Systems group, doing speech research. When Exxon decided to switch all its effort to
development,they resigned in protest because they felt that research was essential for
future products.Left without an employer,they made a completelyspur-of-themoment
decision to start their own company.
In 1982,the Bakers founded Dragon Systems,Inc., which became a leader inspeech-
recognition-technologyproducts. They had saved enough money to fund a team of four
for a year and a half while the product was being developed. Since speech products
represent a slowly emerging market that is limited by the state of the technology, the
Bakers planned to operate at a low level of funding, extending their technology and
developing a product and market. They ruled out the "big bang" start-up approach as
inappropriate in the case of speech products, since the product development process
would be so tedious.
Dragon, while funding itself from product revenue, has consistently advanced the
state of the art. The basic idea of its product is to permit speech input to be substituted
for keyboard input in a completely transparent fashion, so that any application can be
operated via speech input.Dragon sellsits product directly to users and through original
equipment manufacturers (OEMs) and value-added resellers (VARs), which incorpo-
rate it into more specificapplications.For example,another start-up,ArticulateSystems,
uses Dragon's speech recognizer as a front-end control device for the Apple Macintosh.
Dragon operated with conservative goals and delivered its first product, capableof
recognizing a sixty-four-word vocabulary in real time on an 8088 (the PC), in 1984.In
 CirmsLogic 283

1990, the Dragon recognizer is capable of recognizing tlurty thousand words when
spoken in isolation, with only minimal speaker training, and sells for $9,000 as a PC
option.
Having bootstrapped its start-up without outside investment, the company is
employee-owned,wholly supported by customer revenues, and profitable.

CIRRUS LOGIC GOING FROM RESEARCH TO FIRST

PRODUCTS IN ELEVEN YEARS ON SHEER TENACITY

The history of Cirrus Logic encompasses several phases, beginning with a 1974
invention, wluch was further developed by a 1981 start-up company in Utah, which
then moved to Silicon Valley and became Cirrus Logic in 1984. The following subsec-
tions trace the development of this fun, which went public in 1989, and examine the
factors that contributed to its success.

RESEARCH STAGE THE INVENTION AND ITS DEVELOPMENT

In May 1974,SuhasPatil, an assistantprofessor at MIT, invented the conceptfor storage/

logic array (S/LA), a method for designing digital systems. S/LA represents the logic
of a design,alongwith the associatedphysical layout for complexVLS1(very-large-scale
integration) chips, in a structured fashion. Both the logic view of the design and the
physical structure are visible and part of the design. Use of S/LA has resulted in quick
design turnaround, with density and performance comparable to handcrafted VLSI
circuit implementations.Suhas described the development of S/LA as follows:

S/LA came from a recognition that my techniques for asynchronous logic design for
correctness by constructionhad to have a flat, 2-D implementationto match the topology of
VLSI to be of any value. . . . I was bothered by the gap between how a design was
conceptualizedand how it had to be physically realized.

The S/LA idea occurred as a ”flash recognition,” just like most inventions. By June, the
research team I headed had built a breadboard that worked the first time. We packaged
it in a 29-inch suitcase and toured Europe and the United States with it.

Next came the hard part. No one in the computer or semiconductorindustries cared. They
were quite content with the status quo. I tried to work with MIT’s Lincoln Laboratory,
which had a semiconductorlaboratory to do chips, but they had their own agenda.
284 Case Studies

In December 1975, I went to the University of Utah, where General Instruments (GI) had
set up a semiconductorlaboratory.The decision to leave MIT after ten years was hard, but
as a married assistant professor, I was dipping into savings I had made as a graduate
student and was headed for debt.Utah was a wonderful opportunityto test the invention.
NSF [theNational Science Foundation]let me carry my $250,000 research grant to Utah,
and Larry Hill, senior vice president of GI, supported the effort with a $330,000 grant.

At the University of Utah, Suhas continued to develop S/LAs and spent the
summer of 1977 at Sperry Research showing how the technique could be applied to
mainframe design. An eight-channel communications chip was designed in 1978 to
demonstrate the efficacy of the technique and its applicability to LSI (large-scale
integration), but again, traditional designers, including the semiconductor design
group at Ihgital, ignored S/LA.
InJune1980,a decisionwasmade to look for real problemstowhichSuhas's method
could be applied.Jerrold, a division of GI, had just farmed out a chip design for a TV
encoder to Motorola,and Suhas was able to carry out a design in parallel so his method
could be compared with traditional design.

PATIL SYSTEMS INCORPORATED:

A START-UPTO DEVELOP TECHNOLOGY

In January 1981, Patil Systems Incorporated (PSI) was formed. (In naming the firm,
Suhashad followed the advice of friends: "If you're serious about a company,put your
name on it.") Suhas described PSI'S strategy as "up-front, customer financing." PSI
would develop components (not designs for a client company),would not be a design-
automation software company, and would not have a semiconductor fabrication
facility.
PSI contracted with General Instruments to develop the software for S/LA and to
develop chps using the compiler, in exchange for exclusiverights to the compiler after
two years. By the winter of 1981, PSI's first chip was completed, beating Motorola in
time. This was the first demonstration that S/LA actually worked, since twenty
thousand parts were made and installed in a Jerrold product. Three other chips,
including a clone of a difficult Intel communications chip, were designed using the
software, but none was fabricated. By the end of 1982, with eight employees, the
company was profitable and retained earnings of $70,000.
Unfortunately, Hill retired from General Instruments, and the GI semiconductor
facility used for the work was dismantled.Suhashad lost his internaladvocateand was
cut loose from what had been a synergistic relationship. Suhas described Hill's
management principles as follows:
 Cirrus Logic 285

1. Assume something will always go wrong. Always have enough cash to cover
and withstand mistakes.

2. Controlling costs in a small company is done by controlling head count.

3. Always have a statement of work that everyone understands. This can be

changed as necessary, but always know where you’re going.

4. Every month, write a technical report to describe what you’ve done. [In 1990,
Suhas expanded this rule to include a weekly one- to two-page report on what
happened last week and what’s supposed to happen next week.]

In March 1983, PSI began its first private offering, selling9 percent of the company
for $380,000. Suhas‘s completed business plan did not attract much interest, perhaps
because Utah was simply off the beaten path. Suhas’s father, a retired executive from
India, was on the board and helped with the financing and control.
In June 1983,PSI got a letter of intent from Jerrold for three hundred thousand parts
for a new design at a price of $4 per part and a nonrecurring engineering charge of
$100,000. The design was started in August, with Suhas as the architect and two of his
students/employees carrying out the detailed design of the cells and the chip. By
October, the president of GI summoned Suhas and the heads of Jerrold and the
semiconductordivision to New York to ask why the company had to go to PSI for chips
when it already had designers. Fortunately, the team was able to demonstrate that the
design was nearly complete, and PSI was allowed to continue the contract, although it
had to turn over fabrication rights to the semiconductor division.The chip worked and
proved to be critical for future funding.
By the end of the year, Suhas concluded that it was pointless to continueoperations
in Utah without access to capital, foundries, trained people, and customers and
thereforedecided to move to SiliconValley.Four team members went with him, and one
entered academe but eventually rejoined the company.

CIRRUS LOGIC A SILICON VALLEY SEMICOMPUTER START-UP

In January 1984, Kamran Elahian, a Silicon Valley start-up veteran, was hired as a
consultant to help write a business plan and raise capital for PSI. By February, the team
moved to California,and in March,Nazemand Associateshad invested $200,000to seed
the California company. The firm continued to design and manufacture custom chips
for GI using several Silicon Valley foundries.
286 Case Studies

In May 1984, Suhas met Mike Hackworth, a senior vice president of Philips’
Signetics semiconductorcompany responsible for MOS (metal oxide semiconductor)
and linear products. Mike joined PSI in January 1985.Suhas’s advice on recruiting: ”If
you really want someone, go after them hard. Never take ’no’ for an answer. A ’no’
doesn’t necessarily mean ‘no’ forever.”
By September 1984,a plan to raise $5millionwas complete.The company’s revenue
plan (see Table 11-1) showed profitability in the fourth year. The plan called for
designingand supplyingcustom chipsfrom customerspecifications.A chip-fabrication
facility would be built in the second year. The firm would convert digital systems
designs that used older technologies to S/LA-based CMOS (complementary metal
oxide semiconductor)designs, with regional sales and satellite design centers. In the
third year, the company would begin selling its design tools.
The California firm, named Cirrus Logic, began operating in November 1984 with
$1.5 rmllion. Kamran served as the interim president. The first round of financing was
completed in May 1985,when $6 millionamved from venturecapitalcompanies.Suhas
was in charge of technology and served as chairman of the board.
For the first two years, Cirrus designed and sold custom chips. However, this did
not produce the desired revenue growth or product family, as shown in Table 11-1.
According to everyone in the semiconductor industry, the only way to ship large
volumes was to manufacture standard parts. This ”common knowledge” about high-
volume standard parts tended to impede the formation of companies to manufacture
and sell custom-designed parts, including gate arrays. In 1986, the decision was made
to concentrate on standard parts and to find “godfathers”for whom general-purpose
chips would be designed.The first of these was a mass storagecontroller for Seagate.By
1987, Cirrus had three standard parts and was on its way to becoming a semicomputer
company.

TECHNOLOGY, ENGINEERING,AND PRODUCT

Cirrus Logic and its technology evolved through the following classic stages:

1974 The basic idea for S/LA came with an “Aha!” during research at MIT.

1975 The idea was evolved.

1977 The idea was then tested in several stages of applied research at the University
of Utah and rejected by all the companies to which it was offered.

1981 Pat2 Systems, Inc., was formed with seed funding and contracts to develop
computer-aided design (CAD) software to support S/LA and to test the
software on a first production chip.
1983 The first private placement of stock in PSI took place.
 CirmsLogic 287

Table 11-1. Cirrus Logic’s Plan Versus Actual Revenue and Products.
1985 1986 1987 1988 1989 1990
1985 plan revenuea 1.00 6.30 23.0 52.0 95.0 -
1985 plan net profita -3.90 -5.30 -0.8 1.7 18.9 -
Actual revenuea 0.13 0.35 5.0 9.2 36.9 87.0

Actual net income” -0.76 -4.40 -6.4 -8.4 4.1 15.4

Products (custom) 1 3 - - - -
Products (standard) - - 3 13 28 -
”In millions of dollars.

1985 Cirrus Logic began product development as a custom semiconductorsupplier

using the technology developed by PSI.

1987 Cirrus evolved a standard semicomputer product line for PC disk control-
lers, communications, graphics, etc., and entered the market dmelopment and
calibration stage.

1988 Cirrus became profitable during the market expansion substage of market
development and expanded their product line.

1989 The firm went public (steady state 1, with revenues of $37 million.

Because technology is the basis for any new venture in the computer field,
maintaining the best technology balance sheet is the key to long-term success. The
technology balance sheet shows that Cirrus Logic has three principal strengths: S/LA
with automated tools to support quick turnaround of designs (a CAD process) and
trained engineersto use the tools; semiconductorprocess independence,wluch enables
the company to use a wide variety of fabrication facilities; and expertise in designing
microprocessor peripherals.With these strengthsin technology, Cirrus is able to build
its state-of-the-artproducts. Future options include moving to other hardware besides
the IBM PC (e.g., Intel 80x86) and to other peripheral areas, such as speech and image
processing.
Given Cirrus’sdependenceon complex design skills,it mightbe naturalfor the firm
to locate design facilities in other parts of the world to take advantageof the availability
of highly slulled talent, such as Gateway (discussedin the following section) has done
by locating a software-engineering laboratory in India. This approach would be
especially easy in Cirrus’s case, since a large fraction of Cirrus employees are foreign-
born engineers.
288 Case Studies

PLAN

From the beginning, both PSI and Cirrus Logichad a clear plan based on technologyand
products. PSI was founded to develop technology for designing chips and, in a second
phase, to test that technology by designing real chips. Cirrus’s first business plan was
to design complex custom parts for specific customers and to start by using Silicon
Valley fabricationfacilities.By the third year, it had become clear that a custombusiness
would not enable Cirrus to attain its planned revenues, nor was a foundry necessary.
Although,as shown in Table 11-1,Cirrus’sactualresultsturned out to be relativelyclose
to the plan, the companydeviated by getting out of the custom design business and into
the semicomputerpart business.Cirrus neitherbuilt a fabricationfacilitynor established
satellite design facilities.

MARKETING AND SALES

Mike Hackworth’s background at Signetics was oriented toward marketing. Cirrus’s

technological strength was thus complemented with Mike’sability to select the product
areas (e.g.,disk controllers)and to sell the right ”godfather”relationships (e.g.,Conner
Peripherals, Seagate)in order to achieve production volume.

PEOPLE

Throughout the development of the company, Suhas concentrated on building rela-

tionships and finding key people who could advise and help him and who would join
the firm. His advisers and mentors included a businessman in Weston, Massachusetts,
who served as his host family in the United States; various entrepreneurial MIT
professors (AmarBose, Ed Fredkin, and FrancisLee);Larry Hill of GI; Dean Brown,who
headed the University of Utah Innovation Center; Ted Bonn of Sperry Research; and
entrepreneur Vahan Gabusian. For example, Suhas, a most persuasive recruiter, hired
Mark Singer, a graduate fresh out of Berkeley Business School whom he met on an
airplane, to help with the business operation while PSI was in transition to California.
PSI had a good probability of succeeding in its CAD development because the
inventor led the original research team. The firm was profitable and might have
continued operating as a designer or supplier of custom chips in Utah, with the implicit
assumption that it might someday be acquired by GI. However, Suhas wanted more.
Suhasled the companyin its secondphase and, as mentionedearlier,recruited Mike
Hackworth.Mike, who had done it before, understood the semiconductorindustry and
had the confidence of the team.
 Gateway Design 289

Cirrus regarded recruiting as its number one concern. Although an advanced

degree is not a requirement for promotion in the technical area, the firm is staffed with
an unusually high number of Ph.D.'s who have made the transition from research to
product development.
The number of board members increased with the multiplerounds of financingand
in 1990 included Mike, Suhas, and six outside members with a broad range of
experience.

CASH, FINANCEABILITY, AND CONTROL

PSI operated in control as a contractingbusiness.It was very carefulabout spending and

taking financial risks. Suhas also offered the following advice with regard to salaries:
"Never be the highest-paidemployeein your organization,even though you may head
it. This sets an excellent example and helps keep salaries from running away." Hardly
any CEO follows this advice.
Cirrus Logic started out less in control than PSI had been, because it was operating
on a much larger scale than PSI and was subject to greater contract fluctuations.As a
result, Cirrus required five rounds of funding before the IPO (see Table 11-2).

GATEWAY DESIGN: TECHNOLOGY AND PRODUCT,

INTUITION, AND LEARNING FAST

Prabhu Goel, who founded Gateway Design,was born in India in 1949,the son of a civil
engineer.After receivinghis bachelor's degree from the Indian Institute of Technology
in Kanpur in 1970, Prabhu came to the United States to study, obtaining his Ph.D. in
electrical engineeringin 1974 from Carnegie-MellonUniversity. His dissertation dealt
with testing digital circuits. He then took a job with IBM, where he contributed to
testability. In 1977, he began working on electronic computer-aided design (ECAD)
programs, which were used within IBM. In 1981,he went to Wang Labs to head its CAD
development.
In 1982 Prabhu founded Gateway, planning to fund the company through his
consulting revenueswhile he developed and sold high-technologyproducts. During its
first four years, Gateway operated with no written plan, board, chief financial officer
(CFO),salesor marketingpersonnel,customer senicedepartment,or head of engineering.
From the first year, Gateway was profitable and controlled itself by never spending in
advanceofits shippablebacklog.Gateway'sstoryillustratestheimportanceofachieving
and maintaining a lead in product development and of having technologists in control
290 Case Studies

Table 11-2. Funding and Valuation of Cirrus Logic.

Round PricelShare No. of Sharesb $/Round" Raised" Valuation"
Round Total
3/83 0.001 1.ooo - 0.001 0.001 -
(common stock)
3-12 /83 3.000 0.127 1.27 0.380 0.380 3.8

Company is restarted, with first-round investors' stock taken as series A of PSI seed.
3/84 (A) 0.950 0.200 - 0.200 0.576 -
5/85 (B) 1300 4.150 - 7.500 8.100 -
3/86 (C) 2.500 2.000 - 5.000 13.100 -
11/86 (D) 3.750 1.200 - 4.400 17.500 -
4/87 (E) 5.410 1.960 - 11.ooo 28.500 -
6/89 (PO) 10.000 3.340 11.97 31.000 59.500 120.0
6/90 22.000 - - - - 316.0

"In millions of dollars.

millions.

at the top. Although the companybroke nearly all the conventional rules, it still scored
high in the Bell-Mason Diagnostic.

CONCEPT AND SEED STAGES

In 1982, Prabhu visited Cirrus (not to be confused with Cinus Logic, discussed in the
preceding section), a small, successful British CAD company that was selling a logic-
simulation program called HILO. In July of that year, he resigned from Wang and
started Gateway as a new CAD company.His unwritten business plan called for using
consulting and lecturingincometo fund his product development.Because of Prabhu's
strong reputation in logic design for testability,he could commandhigh consultingfees,
enabling him to spend most of his time developing the product.
The proposed product would be an automated test generationprogram for circuits
with scan paths. Having known and contacted many of the potential users for such a
program, Prabhu was confident that there existeda reasonablemarketfor it. Hebuilt the
productusing a terminalconnectedto a VAX. Figure 11-4showsthe company's position
at the concept stage (about March 1982)and at start-up (August 1982)-i.e., at the end
of what is normally the seed stage.
 Business plan Business plan

Technology/engmeenng CEO Technologylengineenng

~rahons/conml

w-
Finanmbdity Cash
Board of
dimtor8 Cash

Figure 11-4. Relational Graphs for Gateway Showing the Concept Stage (March 1982) and Seed Stage (August 1982).
292 Case Studies

PRODUCT DEVELOPMENT STAGE

In August 1982, Prabhu used a mail-order firm to help him incorporate Gateway in
Delaware, with an initial capitalization of $500. In December of that year, Prabhu
successfully ran the first customer benchmark program for Texas Instruments. The
second successfulbenchmarkwas run for Raytheonin February 1983.The alpha testing
of the product, TestScan, priced at $150,000 for a VAX, was complete. The final product
was then ported to IBM mainframes. In making the transition from VAX to IBM,
platformindependence (i.e.,having a program be independentof what platformit uses)
became a goal for all subsequent products. Thus, Gateway’s products would all be
rapidly portable to other computers. Charging for the porting of software to other
platforms became a source of revenue.
In July1983,after Gateway had experienceda year of profitability and had a product
in hand with two expected orders, a second engineerwas hired to maintain and extend
Testscan. While TestScan was being sold, the need for a second product with a much
larger market became clear. Prabhu thus began working on TestGrade, for testing
conventional, nonscan logic circuits.
A critical period ensued during which contract negotiations with the first two
customers took up hundreds of hours, with Prabhu usually being outnumbered by a
factor of ten. In negotiating a contract,he took special care about operational commit-
ments and liability while holding firm on price. Raytheon finally signed in December
1983.TestScan was then shipped,accepted,billed, and the revenue collectedduring the
first quarter of 1984.
Prabhu decided to build on his first product’s success by recruiting a world-class
simulator designer, Phil Moorby, one of the developers of HILO. Phil’s charter,
begnning in June 1984, was to build the ultimate mixed-mode simulator for specifylng
digital systems at the register transfer and gate levels. Phil offered the following
comments on why he joined Gateway:

Knowing Prabhu, I had a gut feeling that he would be successful.While at Wang, he grilled
us on HILO and really worked us hard, but he also didn’t expect utopia. He didn’t ask for
ridiculous enhancements that didn’t make business sense. He had a technical sense, a
business sense, and the ability to negotiate with people. He was unique and had
tremendous talent and made me feelconfident.The day1joined I felt successful.There was
always the sense that if you don’t succeed, then it‘s your own fault. We just went after
everythingwe were capable of doing. We never relaxed on an account.Prabhu‘s style was
such that you felt like you could never get anything by him so you always were trying to
do your best and cover all angles, because if you didn‘t, he would.

Figure 11-5 shows the status of the Gateway start-up at the end of the product
development stage.
 Gateway Design 293

Business plan

Technology/engineering

Operations/control

a
Financeability

Cash
Board of
directors

Figure 11-5. Relational Graph for Gateway Showing the Product Development Stage
(December1982).

MARKET DEVELOPMENT STAGE

With one product on the market, another product under development, and three
employees on the staff, Gateway moved to commercial office space and entered the
market development stage of its growth. Gateway's products were promoted solely by
word of mouth, and sales were made by the existing employees.For control purposes,
shppable bookmgs were used to manage the cash flow and eliminate the payment of
income tax. The diagnostic (Figure 11-61 shows Gateway to be a tiny company at this
point, but very well rounded and in control.
Plul's product, Verilog,was capable of describing digitalsystemsat multiple levels
of abstraction for simulation.It was ready for demonstrationby the first quarter of 1985.
A road show during which the product was demonstrated to thirty potential customers
in Silicon Valley and Los Angeles yielded an order from Sun Microsystems (the main
host for the simulator). In fiscal year 1985, Gateway had revenues of $1 million with
bookings of $1.1 million. A vice president of marketing was then hired. Before leaving
294 Case Studies

Business plan

Technology/engineering

Operationdcontrol

Cash
Figure 11-6. Relational Graph for Gateway Showing the Market Calibration Stage
(June 1984).

a year later, he was instrumental in hiring Craig Robbins as the vice president of sales.
Craig resided on the West Coast, where the bulk of Gateway’scustomerswere located.
In December 1985, Prabhu was contacted by three venture capital groups. In May
1986, an agreement was concluded whereby Greylock and Fidelity Ventures would
receive one-third ownership of Gateway for $2 million, with the financing being
completed in July. Henry McCance of Greylock and Gordon Kingsley of Fidelity
Ventures joined the board of directors. Clearly, Gateway was not in need of the cash.
Instead, like Microsoft, it sold part of the company to get expertise in hiring and
financing along with the financial community’s imprimatur on the firm so it could
advance rapidly within the high-tech business community.Prabhu believed a financial
partner was needed for future financing.
During the four years beginning in 1986, revenues went from $1.5 million to $12
million, with a backlog of orders worth $17 million (much of which had been shipped
but not billed). The first CFO, Dan Keshian, arrived in 1987.The usually high level of
bookings in the later years was used to manage the company’s cash flow, reduce its tax
liability to zero, and fund its growth during the following year.
 Gateway Design 295

In October 1989, Gateway and Cadence, the fastest-growing company in the

electronic CAD market, announced their merger. Cadence was challenging Mentor
Graphics,the leader in ECAD. Gateway’svalue was $80million in December,at the time
of the merger; in July 1990, the Gateway portion of the combined organization was
valued at $135 million. This gave the venture capital investors a return of more than
twenty times their investment after only four years. Although it might be argued that
Greylock reaped an exceptionally high rate of return on its initial investment and
ongoing advice, Prabhu stated: “Greylockjoining me gave me the insight to hire a CFO
and put the financialsin place early on so we would be prepared to manage the growth
we anticipated.”
By 1990, Prabhu was managing a substantial division of Cadence, with personnel
on both coasts of the United States and in India. Phil Moorby, who became the first
Gateway Fellow in 1988, started work on a completely new kind of design-automation
software.This approach allowed the technologistto continue working productively in
a way that was most beneficial to Gateway and most rewarding for h m , without his
having to become a manager.

TECHNOLOGY AND PRODUCT

The Gateway story shows how technology developed by two individual contributors
was capable of generating lasting product families. TestGrade, enhanced in 1985,
remains on the market in 1990.Verilog, the logic simulator,was introduced in 1985,and
a faster version, Verilog XL, was introduced in 1987.A simulator for the Department of
Defense-specified language VHDL, incorporating Moorby’s algorithm from XL, was
introduced at the Design Automation Conference (DAC)in June 1990.T h s algorithm
enabled Gateway to introduce software-only simulators that could compete with
specialized hardware accelerators.Bill Kaiser of Greylock commented:

From a product/feature comparison, we almost always won based on performance, so

there was a bit of an open field, which made it easier to do business.

Prabhu incorporatedlogic scan test methodology, a technologythat he had learned

at IBM. By having superb technology and technical leadership, Gateway was able to
attract the talent that was needed in order to keep the firm at the forefront.In 1988,when
Gateway had four evolving products, Gary Leive was hired as vice president of
engineering.
Quality was the cornerstoneof product development.Prabhu and Phil established
a culture in which products were designed to require minimal technical support,
because a product would not be released if it had bugs. A substantialamount of software
testing was automated through the use of test scripts. A customer service department
296 Case Studies

was added in the first quarter of 1986, to provide customer training and answer
questions.
In late 1986,Gateway established a development subsidiary in India. By 1990,the
twenty-eight engineers there were producing new, high-quality products at a small
fraction of the per-person cost of US. engineers.Ths subsidiary is scheduled to grow
to over a hundred engineers and, in the early 1990s, represents a major asset of the
company.

PLAN

A plan for developing a premium-priced, advanced line of products and marketing

them can be quite simple and need not be written down, provided the founder can self-
fund the venture. Self-fundingis an ideal approach to starting a company that others in
a comparable situation should use. The plan, similar to Autodesk‘s, was to write a
variety of programs and then select the winning product(s) for growth. Autodesk
developed and evolved a simple drawing program, AutoCAD, for a mass market of
architects and structural and mechanical engineers.Prabhu and Phil were the worlds
experts in a small but growing technical product area, and they understood their
customers’ needs. Gateway’s strategy, like Autodesk’s, was to use the emerging
powerful workstations in order to provide a design and simulation facllity for every
engineer. Gateway’sthird product, Verilog, became the cornerstonefor market growth,
and the company bet its future on it.

MARKETING AND SALES (AND SUPPORT)

Right from the start, Prabhu and Phil understood the market because they were the key
developers of the technology and products that defined the market. Prabhu, the chief
marketeer,had an uncanny ability to relate to users and their problems because he had
served as product developer and manager of CAD systems within IBM and Wang.
In the beginning, the responsibilityfor documentation,includingthe salescollateral
literature (documentationdescribing the product),rested with the product developers
because of their knowledge of the product, market, and customers. In late 1986, a
product-marketingperson was hired to help the engineersdescribe their products and
to prepare product information for the sales team.
Bill Kaiser of Greylock offered the following observation:

Gateway had an outstanding product that was adopted quickly by some blue-chip
customers.Early on, companieslike Motorola providedword-of-mouthreferenceaccounts
to help the sales process.
 Gateway Design 297

In the spring of 1987,Ronna Alintuckwas hired to handle public relations,corporate

marketing, and market positioning. Ronna was gven a free hand but only a small
budget. Her phlosophy, like Gateway’s,was based on doing only a few things well and
in a way that would communicate effectivelywith the company’sengineering market.
Prabhu acted as sponsor and coach for her nontraditional marketing.
Gateway decided to attend only two out of a possible eight trade shows, regarding
the remaining six as redundant. At the DAC (DesignAutomation Conference)and the
more prestigious academic conference, ICCAD (IntegratedCircuit CAD),it often gave
away promotional items.
Gateway’s brochures were initially limited to informative product-description
literature. Only in 1989did the firm produce a glossy, corporate-typebrochure. It was
filled with data about the company and its philosophy, people, and products and also
included puzzles designed to be intriguing to engineers.
In 1988, Gateway ran a series of space ads with testimonials and photographs of
Gene Amdahl; Forest Baskett of Silicon Graphics; Tom West of Data General; and me,
for Ardent. Gateway’sads were run in only one trade publication, EE Times, which was
read by the potentialbuyers. The ads used a quote from Einstein-”Imagination is more
important than knowledge”-as a recurring theme to position Gateway as a company
whose products inspire the creativity of their users. By 1990,most CAD companies had
switched to testimonial-style ads.
Gateway’sfirst users’ group meetingwas held to coincidewith the 1987DAC.Some
of the more important milestones that the company achieved in marketing included:

May 1988: Gateway was named the “hottest CAE [computer-aidedengineering]

growth company” by Electronics magazine.

February 1989: A CAE analyst described Verilog as ”the clear-cut winner in mixed-
level simulation.”

June 1989: Prabhu was named ”entrepreneur of the year” by Inc. magazine.

July 1989: Electronics magazine claimed that Gateway had become Mentor’s main
rival, despite the fact that Mentor was over twenty times larger than Gateway.

PEOPLE AND CULTURE

As noted earlier, Gateway operated during its formative years without a board, CFO,
sales or marketing personnel, a customer service department, or a head of engineering.
Responsibilityfor these functions initially rested with one person and, as the company
grew, was distributed among the developers. Thus, Gateway gave its employees
298 Case Studies

enormous responsibility. Occasionally, that trust was misplaced, and Prabhu had to
learn the difficult lesson that it is unwise to wait too long to make personnel changes
when someone gets in over his or her head. Prabhu summed up his feelings about
employees as follows:

Other than our focus on product and financial control, there was persistence,hard work,
and recruiting the right people. We let people run with their mission rather than trying to
control things too closely. We may not have always hired people on paper that had the
experience for the job, but as they exercised their positions,they flowered.I also think that
having a technologist at the top helped. In the beginning, when we were out there with
the customers, I was able to relate to their product needs in the marketplace.

Clearly, Prabhu set the product, technology, quality, and teamwork standards for
the company-i.e., he established its culture. The following story told by Prabhu
illustrates the important role that commitment and teamwork played in Gateway’s
success. It describes a n event that not only generated business for Gateway but
incidentally turned out to be an excellent team-building project as well.

On [the]Friday before Memorial Day weekend in 1986,Prabhu was on the West Coast and
learned that the company was third in a race for a Hughes benchmark of mixed-level
simulationthat the whole industry was watching. The leadingcontenderhad a capability,
stochasticanalysis, that Hughes felt they needed. Prabhu called a friend, who explained
stochastic analysis. Prabhu called back to the company and outlined a plan to add the
capability to Verilog. All eight engineers in the company canceled their weekend plans.

On the plane back, Prabhu worked on the benchmark and a demo to exploit the new
capability. The demo, which relied on having extended graphics monitoring of queues
(the company called these dancing bar charts),turned out to provide a dramatic demon-
stration of the product that also helped sell it. Gateway had listened to the Hughes
requirement (because they would have otherwise lost the sale), and by winning the
benchmark, Gateway got the business and a number of other contracts.

Henry McCance observed:

Gateway was such a joy because Prabhu operated openly and was a great listener and fast
learner. . . .We said, ”Why don’t you look at Hale and Dorr as counsel?”Within the month,
they were signed up. . . . The board suggested [creating]a technical advisory board, and
by the next board meeting, the company had a first-rate TAB.

Ronna describes the company like ths:

[Gateway’s]ability to market itself to itself was very important. At the first Annual
Gateway End of Fiscal Year Party in 1987, as he [Prabhu] was introduced as not being
affected by success in light of all the publicity,he shed his normally conservativebusiness
 Gensym 299

attire and came out in a sequin-covered suit to the Saturday Night Fever sound track. Next
year, he was dressed as Batman. The company wanted to reflect its goal of having
innovation and fun.

CASH, FINANCEABILITY, AND CONTROL (AND PROFITABILITY)

A key part of Gateway's culture involved being in control, not spending money before
it was earned, having adequate cash, and being profitable every year. In Prabhu's
words:

We never allowed ourselvesto believe that we could spend more than we had. From the
beginning, there just wasn't a huge cash reserve, and I felt the time I'd spend getting the
extra money would take longer than it would to get the product out the door. As we grew,
we also had the independence to focus on the fundamentals but not constrain ourselves
to a quarter-to-quarter focus.

Henry McCance described several elements of the Gateway culture that were
instrumentalin the firm's success:

Prabhu went to auctions and bought supplies from companieswe had funded that were
going under. . . . Prabhu did it differently. Most software companiesbook products they
can't shipand then start spending against thesoft bookings. Prabhu only booked products
he could ship and then shipped products against this backlog. He just forgot to bill until
the next fiscal year and was able to save on taxes.

Similarly,when Gateway went to its first trade show in Las Vegas, it hired a demo
van (whichdoubled as a hotel room and moving billboard)because the company might
not havebeen able to demonstrate its productsin a hotel suite.In this regard, it is unclear
how much Gateway's culture will change now that Gateway has become part of
Cadence and Prabhu has stopped staying in inexpensive motels.

GENSYM: OUT OF THE ASHES

In September 1983, Bob Moore came to Lisp Machines, Inc. (LMI),to head a division
developing expert systems for process control. Bob had a Ph.D. from MIT in electrical
engineeringand had been in the process-controlindustry for more than a decade, with
experience in managing marketing, sales, and engineering. Lowell Hawkinson, an
alumnus of MIT's Artificial Intelligence (AI) Laboratory, led the development effort.
Over the next three years, LMI spent about $3 million to develop Picon, a rule-based
process-controlprogram, the first product of its type. Picon was demonstrated in 1984
and in 1985 was put on-line at Exxon and Texaco. The division, with about ten
300 Case Studies

professionals, had roughly a dozen major customers, and its sales exceeded half a
million dollars.
At the AAAI (AmericanAssociation of ArtificialIntelligence)Conferencein August
1986,thepiconteamcame to therealizationthatLMI hadnofuture.The team’smembers
reasoned that unless they formed themselves into a separatecompany,it was likely that
little would ever come of their product and their work.
Two LMI vice presidents had just been dismissed, and the firm was headed for a
painful restructuring. However, strong forces on the board supported Picon. LMI was
in the process of getting yet another round of financing. Its president, Ward McKenzie,
had come from Digital over a year earlier as part of an aggressive expansion of LMI
based on building its second-generationLISP machine, whch it and Symbolics had
originally licensed from MIT. Ward raised $25 million and built a large, full-fledged
computer company to manufacture, sell, and service products for the minuscule and
slowly emerging AI market. LMI had wanted to compete with Symbolics,which was
successfullybuilding and selling essentially the same product. Symbolics was subse-
quently forced out of the LISP machine business as general-purpose workstations
became faster.

CONCEPT AND SEED STAGES

Ed Fredkin-an entrepreneur,inventor,BostonUniversity physics professor,computer

pioneer, and venture capitalist-agreed in August 1986 to cofound and seed-fund the
Picon development team. The concept stage was completed in two weeks, with a plan
created by the six key team members. Their intention was to license Picon from LMI or
build a next-generation version of a Piconlike product from scratch if LMI refused to
license Picon. They stopped communication with LMI, and the team resigned (taking
no proprietary material with them).Gensym was incorporated two days later, with the
aid of $300,000 from Fredkin. The team began looking for office space and computers.
Figure 11-7shows the status of the Gensym start-up at the concept and seed stages.

PRODUCT DEVELOPMENTSTAGE

The product development stage began in late September 1986. Host computers were
provided by Symbolics and Texas Instruments (TI). The team started developing a
product that would have substantiallymore capabilitythan Picon. Gensym’spresident,
Bob Moore (using the Finis Conner model of product development: “Sell, design,
build), began talking to customers to survey their needs.
Gensym’sfive-year business plan, written in the fall of 1986and revised in January
1987,calledfor$93millioninsalesin1991.Theplanshowed$l millioninrevenuein1987
 Gensym 301

Business plan

Technologylengineenng

Operationdcontrol

Cash

Business plan

Technologylengineering

Operationdcontrol

Cash

Figure 11-7. Relational Graphs for Gensym Showing the Concept Stage (Labor Day 1986)
and Seed Stage (Late September 1986).
302 Case Studies

and an 11percent after-tax profit based on sales of $6 million in 1988.In April 1987,the
firm’s product, G2, entered beta testing. The basic funding plan was to keep a low
profile, minimize funding from the venture capital community, and make a few
strategic alliances. In March 1987, Gensym had contact with about fifty venture
capitalists. Bob Moore believed that the following factors accounted for their lack of
interest:

AI and expert-systems software were out of favor.

The venture capitalistsdidn’t believe the overly optimisticbusiness plan for selling
process-control systems.

The company’svaluation was considered excessive, since it was unwilling to give

up controlling interest.

Gensym’s estimate that it would need only $2 million was thought to be naive.

The development schedule was regarded as overly optimistic.

The company would not send out business plans except on a strict nondisclosure
basis.

In May 1987,$1million of stockwas sold to private investorsand Palmer Associates,

providing capitalization of $1.6 million. The founders retained 70 percent of the
company for their investment of $50,000. Gensym demonstrated its first product in
August. Beginning in September, it sold and installed the early version of the product
at several sites, including Du Pont, in what Gensym’s chief scientist described as a
combined alpha/beta test. G2 was also used as part of a more complex product being
developed by Reliable Water, a company founded by Fredkm to build desalinization
plants. In September 1987, Gigamos, the firm that had bought rights to the LMI
technology after LMI’s bankruptcy, sued Gensym for damages; in June 1988, the case
was settled out of court for an undisclosed amount, with the parties agreeing that both
were free to pursue their own interests.During the period when the product was being
finished, alpha/beta-stage sales were about $250,000.
TherelationalgraphinFigure11-8showsGensym’sstatusattheendof the product
development stage.

MARKET DEVELOPMENTSTAGE

In April 1988, just nineteen months after Gensym started up, G2 version 1.0 was
introduced, marking the beginning of the market development stage. Another $1.1
 Gensym 303

Business plan

Technology/engineenng

Operationdcontrol

Cash
Figure 11-8. Relational Graph for Gensym Showing the Product Development Stage
(April 1988).

million of financing was raised at $1.60 per share, to give Gensyrn a valuation of $6.4
million.The company had its first profitable quarter ending September 1988.By the end
of 1989,Gensym had operated profitably for six consecutivequarters.Its after-taxprofit
was very nearly 15 percent on revenues of approximately $5 million, and Gensym
entered the steady-state stage. The relational graph in Figure 11-9 shows Gensym’s
status at t h s point.
Although the market calibration and market expansion stages, which are marked
by the first profitable quarter, usually take a total of between one and two years, these
stages actually took six months in Gensym’s case. The fact that the product was first
delivered in September 1987 accounts for the shortened time frame. In effect, the
company overlapped the last stage of product development with the first stage of
market calibration.Market developmentreally started when GensymdeliveredG2 beta
units to the first customers, even though the final product wasn’t announced for wide-
scale sale until seven months later, in April. The second major version of the product,
1.11,shipped in March 1989, and version 2.0 shipped in August 1990.
304 Case Studies

Business plan

Technology/engineering

directors
Cash
Figure 11-9. Relational Graph for Gensym Showing the Market Development Stage
(January1990).

TECHNOLOGY, ENGINEERING,AND PRODUCT

Gensym’s technology was developed over a ten-year period, with ideas coming from
Moore’s process-control experience as well as the experience of Lowell and Michael
Levine (the chief scientist) at MIT’s Artificial IntelligenceLaboratory. Clearly, the bulk
of the firm’s technological understanding came from building Picon at LMI.
Gensym’s engineering style was characteristic of the MIT AI Lab‘s tradition of
attractingbright people, giving them lots of computingresources,and generally leaving
them alone. Several companies emerged from this environment, including Gold Hill,
LMI, Symbolics,and Thinking Machines Corporation.
The January 1987 plan called for a beta release in April of that year (three months
later),but this did not occur until August (sevenmonths later),giving a schedulefantasy
factor of 2.3. The chief scientist described the development project as mildly chaotic,
since the group did not work according to any classic software-development-process
model, nor did it start with a rigorous schedule or specification. A cursory product
 Gensym 305

description came into existencein the spring of 1988. According to the usual definition
and metrics of software control, the software-engineering process was out of control
from the beginning until the completion of product development. The development
project simply proceeded on the faith that the individuals involved could ”do it again.”
After all, given G2’s relationship to Picon, it was the team’s second system.
In 1990, Gensym’s ability to complete software engineering according to its
schedule has improved considerably.Both the SFF and the quality level of its releases,
measured in faults per one thousand lines of LISP code, have also improved.
Gensym epitomizesthe softwareproduct company whose engineerswere trained
to be creative and then rewarded for their creativity. As Gensym grows, its challenge
will be to keep ahead of competitors who build knockoff products by the software-
engineeringbook.

PLAN

Gensym’s initial plans are shown in Table 11-3.

Bob Moore’s 1983business plan prior to coming to LMI was to produce revenues
of $25million by the end of the fifth year on an investmentof $2 million.This plan turned
out to be more accurate and realistic than the January 1987 and December 1987 plans.
All three of these plans were highly optimistic and tended to focus on what was
possible rather than what was likely to happen. For example, the entire distributed
process-control market was projected by one market survey company to grow from
$200 million in 1980to $3billion in 1990!However,all the plans were difficultto validate
in terms of the time it would take for the conservativeprocess-controlmarket to develop.
Quite probably, each plan was oriented toward the ”greedy” investor who wanted to
see a $100 million business in five years. None of the plans was calibrated or modeled
as to likelihood or according to market development and order-gestation times.

MARKETING AND SALES

From the start, Gensym tried to interest various hardware platform and process-
control companies in becoming strategic partners in order to help with funding.
Symbolicsand TI saw this need and responded with hardware.A number of early users
helped with the development, including DLIPont and Reliable Water.
The president of Gensym also headed marketingand sales,where his experiencein
all phases of the business, including development,proved especiallyvaluable.Because
its president wore several corporate “hats,” Gensym knew the critical applicationsand
customersintuitivelyand quantitativelyand was also able to maintain a direct link with
customers at the highest level.
306 Case Studies

Table 11-3. Gensym‘s First Plans Versus Actual Revenue.

1987 1988 1989 1990 1991
1/87 plan revenuea 0.98 6.10 21.10 54.2 92.8

1/87 plan profit“ -0.60 0.60 2.80 10.2 19.7

12/87 plan revenue“ 0.60 2.30 4.80 14.2 42.9

12/87 plan profit” -1O
.O -0.47 -0.50 2.2 7.8
Actual revenue “0.52 1.87 4.50 8.0 -
Actual profit“ -0.90 -0.50b 0.70 N.A. -
G2’s product-release cycle beta 1.0 1.11 2.0 -
“In millions of dollars.
”Includesthird-quarter profit of $80,000 and fourth-quarter profit of $80,000.

All of Gensym’s founders had worked together on Picon as a team at LMI and were
technically oriented toward the product and the market area. Lowell, the chairman of
the board and CEO, and Bob run the companyas a team and focuson the inside (product
development and operations) as well as the outside (marketing and sales).
After the first round of investment, the board consisted of the three founders
(Lowell,Bob, and Ed), together with one of the venture capitalists and Ted Johnson,
former vice president of sales at DEC. Ed claimed that the board played a critical role by
insistingon early profitability,which Lowellfelt may have been achieved at the expense
of a potentially higher growth rate.

CASH, FINANCEABILITY, AND CONTROL

By one measure of control, Gensym didn’t make its top line. Faced with this problem,
a typicalstart-upwould begin to decline,enteringa stage of ”hopingfor a better top line”
as it ran out of money. The keys to Gensym’s success were financial control, not
spendingahead of its reduced revenue, and being able to meet the bottom line. The fact
that the foundersown a largefraction of the companyperhaps helped with control,even
though they had only invested a relatively small amount of cash. Bob Moore contrasts
the plan with the actual outcome:

As far as the early business plan is concerned, we have actually followed the plan, except
with the “time axis” stretched out. . . . [Tlhe investment to start Gensym was projected at
 MasPar 307

$2 million, and this was almost exactly what it took. Basically, everything happened
slower, including product development, hiring, sales,etc. But we kept in balance and built
the company ”according to plan” at a slower pace.

Gensymbecame profitable inits eighthquarter.Financingtooka minimum amount

of time away from running the firm.
With a healthybalance sheet,Gensym has lots of options if it needs cash for growth,
provided it remains profitable. Going public would force the team to spend more time
dealing with the externalities of finance, and this would hardly be worth the risk until
the company is substantially larger. A merger might be possible, but this would
jeopardize Gensym’s culture and growth potential. Clearly, the firm is in an ideal
position until it becomes visible enough to attract significant competitors.

EPILOGUE

In July 1990,Gensym bought Picon for slightly over $50,000 at an auction conductedby
a judge selling the assets of Gigamos.

MASPAR
A MASSIVELY PARALLEL, COMPUTER SYSTEMS COMPANY

In the late 1960s,Westinghouse built the first computer with a singleinstructionstream

controllingmultiple data streams(referredto as an SIMD-typecomputer).Other SIMDs
were constructed in the 1970s, including ILLIAC IV,built by the University of Illinois
with Burroughs; the MPP, built by Goodyear; and the DAP, built by ICL. In 1990, the
basic DAP architecture is still being sold. In 1985, Thinking Machines Corporation
(TMC)introduced a massivelyparallelSIMDcomputercalled the ConnectionMachine.
Based on a 1981 Ph.D. dissertation by TMC founder Danny Hillis, the computer has
sixteen thousand to sixty-fourthousand processing elements.In 1982,Tom Blank wrote
a Stanford Ph.D. dissertation on an SIMD computer to be used for simulating digital
systems.During a 1987semiconductorconference,a team fromJeff Kalbs group at DEC
reported on the design of VLSI chips for an SIMD computer. Each chip had thirty-two
processing elements.

CONCEPT AND SEED STAGES

In July 1987, Kalb-a technologist and manager with fifteen years’ experience at
National Semiconductors, Data General, and DEC-moved to California to explore
new opportunities, including the possibility of exploiting DEC‘s SIMD architecture
with its blessing.In September,at the offices of the venture capital firmMerril1,Pickard,
308 Case Studies

Anderson, and Eyre (MPA&E),he began writing the MasPar business plan to develop
a massively parallel computer based on DEC’s VLSI chips and architecture. In October,
MasPar was founded, and the second employee, Jim Peachy, a Hewlett-Packard
manufacturing alumnus and former chief operating officer (COO)of Pyramid, joined.
In November,negotiationswith DECbroke down, and Kalb sought out Tom Blank,who
had become a Stanford professor, to help write the business plan based on an alternative
architecture.
On January 6, 1988, the business plan was complete, and Blank signed on as
architect and head of computer applications. MasPar began the search for start-up
funds. The plan called for raising $24 million, in three installments, for the firm to go
public. MasPar’s product, priced at less than half a million dollars, would deliver
computing power equal to roughly half that of a $22 million Cray YMPsupercomputer.
Specifically,the MasPar product was a scalable, massivelyparallel computer that could
deliver a peak of 1.2 gigaflops or 30 gigaops on 32-bit integers using up to sixteen
thousand processors in steps of one thousand processors. The price range was $60,000
to $350,000 for 128megabytesof memory, excluding the front-endworkstation.The first
shipments touniversitiesandlaboratories werescheduled for the fourthquarter of 1988.
For the first four years,beginningin 1989,MasPar projected revenuesof $1.8million,$11
million, $26 million, and $93 million.
The market sizes were derived from Dataquest’ssegmentationinto price bands and
application. MasPar targeted the two largest Dataquest segments, design automation
and scientific(avery general market).The ”whybuy” marketingquestionwasanswered
as follows:one to two orders of magnitude in performance and/or price/perfonnance,
software libraries to handle the parallel operations, a VAR and OEM strategy for
applications, high reliability through engineering, and a full systems capability with a
well-funded field sales and support organization.
In February 1988, John Nikolls and Peter Christy arrived to head hardware and
software development and to complete the core development team.
The relational graph in Figure 11-10 shows the status of the MasPar start-up upon
completion of the concept and seed stages.

PRODUCT DEVELOPMENTSTAGE

By March 1988,the first venture financing was complete, with $6.5 million in the bank
at $0.67per share.The company valued itself at $10.8million.Jim Anderson of MPA&E
and Floyd Kvammeof Kleiner, Perkins,Caufield,and Byersjoined Jeff to form the board
of directors.Andersonand Kvammealso helped with the early marketingand planning
work.
Since MasPar did not conduct a seed stage during which the product and devel-
opment project could be planned in great detail, all the detailed planning was pushed
 MasPar 309

Business plan

Technology/engineering

Operations/conaol

directors
Cash

Business plan

Technology/engineering

Operations/control

Cash

Figure 11-10. Relational Graphs for MasPar Showing the Concept Stage (October 1987)
and Seed Stage (March 1988).
310 Case Studies

into the first part of the product development stage, during which the product
specifications and schedule were being prepared and hiring was talung place. By
September 1988, the product was specified,a seventeen-memberengineeringstaff was
on board, and a detailed development schedule had been prepared. At this point,
detailed design was started, and by November, a simulation of the architecture was
completed. The custom chips operated in April 1989, and in July, the logic was
completely simulated using a hardware simulator.The product-design stage ended in
October, when the fabricated product prototype entered the debugging stage (IIIc) to
begin alpha testing. On October 18, during the northern California earthquake, a logic
analyzer fell on the system, disrupting the schedule, Another six-week slip occurred
because certain components failed to meet their specifications.
In June 1989, MasPar acquired its second round of financing, raising $11 million at
$1.15 per share, giving it a valuation of $30.8million. Daniel Tompkins, a former CFO
and a partner in DSC Ventures, was added to the board.
In December 1989, the product operated. On January 9, 1990, alpha testing was
completed, and MasPar announced its MP-1 family of products. Five computers were
operational. With the announcement, the first computer was also delivered to a
customer to begin beta testing. Two basic models were introduced, with from one
thousand to four thousand and from one thousand to sixteen thousand processing
elements, delivering a peak of 1.3gigaflops or 26 gigaops and priced from $170,000 to
$810,00Ofor systemswith256megabytesofmemory and the front-end workstation.The
targeted markets included computational fluid dynamics, computational chemistry,
image and signal processing, and VLSI design.
The relational graph in Figure 11-11 shows the status of the MasPar start-up at the
end of the product development stage.

MARKET DEVELOPMENT STAGE

Although beta testing was not completed until May 1990,with the slupmentof six units,
production-unitshipmenthadbeguninMarchofthat year.StageIVa(marketcalibration)
began with the product announcement.By May, the order-gestationtime was six to nine
months, versus a planed three to six months. Critical applications programs in each
targeted market area began being ported to start a vertical marketing approach.
Distribution relationships were established to provide multiple market segments.
In late June 1990, the third round of financing was completed,based on a private
placement memorandum issued in late March. The five-year plan, beginning in 1990,
projectsrevenuesof $3.1million, $19million,$52rmllion,$106million,and $177million,
with profitability in 1392.Another $15.5million (for a total of $32.6 rmllion) was raised
at $1.85per share,bringing the valuationto $66.5d i o n . Althoughonly $10 million had
been sought, the oversubscriptionwas high enough to increase the financing. As Jeff
 MasPar 311

Business plan

Technologylengineering

Operations/control

Cash
Figure 11-11. Relational Graph for MasPar Showing the Product Development Stage
(January 1990).

Kalb remarked, given the enormous investment in raising the capital, MasPar followed
Kleiner‘s law: ‘When the hors d’oeuvres are passed, take two.”
After the introduction, MasPar was shipping at a rate of almost one machine per
week at an average sellingprice of over $200,000. Demand in Europe was the highest for
large machines.Smallmachines were shpped in the United States,Japan,and Australia
to early adopters and to seed OEMs for product development.
Dave Cane,a founderof MassComp,liststhree necessaryconditionsfor a successful
start-up:“[having]a very good product, keeping the customershappy, and not running
out of money. In order to do this, the team must find the work relatively easy to do
because everything else about starting up is so hard.” The MasPar story seems to
embody these three principles.

TECHNOLOGY, PRODUCT, AND MANUFACTURING

Product development actually took twenty-four months, in contrast to the originally

planned nineteen months, which translates into a schedulefantasy factor of 1.26 (i.e.,24
312 Case Studies

divided by 19). The development cost of roughly $10 mdlion was about 1.5 times the
originalplan. However,in absoluteterms, the developmenttime was short, considering
the complexity of the MasPar product, including the development of two VLSI chips.
Although the product was one of the first of its kind and involved brealungnew ground,
the entire project team did have experiencein developing successfulproducts, and the
design process,includingthe design of the two chipsthat formed the product’score,was
within the team’s capabilities.
In 1990, any team that is building its first custom chip of four hundred thousand
transistors can expect the process to take at least two years before a final product with
the chip can be shipped. Dave Nelson-a founder of Apollo and CEO of Fluent
Machines, a companybuilding a workstationincorporatingdigitalvideo-sums up the
conventionalwisdom regarding custom chip design like this: ”The way to make a small
fortune is to start with a large one and do custom VLSI design.”
Kalbran the project,coordinatingarchitecture,applications,hardware,and software.
He was able to handle all these responsibilities because some of his other functions
including managing manufacturing were delegated to Peachy, who acted as a chief
operating officer.
The first MasPar products performed according to plan and delivered their rated
performance and performance/price characteristics on highly parallel applications.
Provided that a particular applicationrequires parallelism, the MasPar computer often
outperforms the Cray YMP supercomputer.
The situation in massively parallel computing is highly competitive. MasPar is
providing a lower-costalternativeto theConnectionMachineusingtheSIMDapproach.
Alliant, Intel, and NCUBE use the multicomputer approach by interconnecting a large
number of ordinary microprocessors. Both approaches are programmed in a similar
fashon.

PLAN

The plans used to secure funding were very crisp, clear, and basically correct. The
originalplan for startingthe companyconsisted of seventy-sixpages,whosebreakdown
is shown in Table 11-4.
Had MasPar gone through a seed stage for planning,its plan would have been more
accurate with respect to the project schedule and resources, but it‘s unllkely that all of
the team would have signed on to write a seed stage business plan.
The private placement memorandumconsistedof forty-twopages, not countingan
additional twenty-five pages of financial and stockholder information and product
literature.
 MasPar 313

Table 11-4. Elements of MasPar’s Original Plan.

Element No. of Pages

-~

Summary and short form 8

Parallel-processingtutorial 8
Product and implementation 13

Potential applications 3

Business strategy, including market size, 20

development schedule, and manufacturing
Competition 7

Risks and future product directions 3

Financial plan with detailed product cost 12

People 2

MARKETING AND SALES

When MasPar first started, the board served as marketing advisers. In October 1988, a
vice president of marketing was hired from Convex and became responsible for
industry applications marketing. In April 1989, the head of sales joined from Silicon
Graphics,where he had headed North Americanfield operations.The strategyof selling
computational-intense, parallelizable applications programs for customers buying
minicomputer-pricedproducts is based on a highly segmented and niche market.
Although MasPar achieved its product-design goal, the company’s success will
depend first of all on how well the marketingand salesorganizationsare ableto segment
the applications (e.g., logic simulation, computational fluid dynamics) that can fit the
architecture.This challengeis identical to the one that Ardent faced when it introduced
the graphics supercomputer.Next, MasPar must find the right customers, including
developing the channels of distribution with VARs, OEMs, system integrators, and
independentsoftwarevendorsthat developapplicationsto exploitthe SlMDarchitecture.
No small part of the problem will be to educate potential users about the nontrivial
nature of parallel programming. Alternatively, given the computer’s input/output
structure, it can be configured for a variety of tasks, including databases, communica-
tions switchmg, and display.
314 Case Studies

Kalb sees Blank’s role as head of both architecture and applications as critical
”Unless the architect understands how the computer does on real problems, we have
no way to exploit and eventually improve the architecture.”
As noted above, MasPar’s future will depend solely on how effectively the
marketing organization first segments the market and then develops the critical
segments.Given the finicky and difficultnature of parallelism,MasPar’s successwill be
achieved on an application-by-applicationbasis. The Stardent product and market-
positioning flaw (discussedin Chapter 8) should serve as a useful guide.

PEOPLE

MasPar started as a product development-centeredteam and was led by a technologist

and manager. The top team members all had the right experience and backgrounds.
Having a COO was critical because Kalb played such a key role in engineering the
product. In later stages, a vice president of engineeringmay be required to take some of
the load off of Kalb. Overall, it is hard to believe that the people dimension could have
been improved. However, until the product is successful, the marketing and sales
strategy and personnel remain untested.

CASH, FINANCEABILITY, AND CONTROL

MasPar performed in an almosttextbookfashion with regard to the cash,financeability,

and control dimensions,despite the product-schedule and cost overruns. The fact that
the original plan overestimated general and administrative costs helped to offset the
increase in research and development costs. Hiring the original team took longer than
expected, thereby decreasing spending and lengthening the development schedule.
However, by takmg more time and operating with fewer people, MasPar probably
developed a better product and plan.
MasPar used management by objectives, staff meetings, and weekly reviews to
manage and track progress.

THINKING MACHINES CORPORATION:

THE FIRST MASSIVELY PARALLEL COMPUTER COMPANY-
GOVERNMENT AND VENTURE FINANCING

Massively parallelcomputing is no exceptionto Silver’sstart-uprule: ”Anythingworth

doing is worth duplicating.”In 1990,AMT (derivedfromICL‘s DAP),MasPar,Thinking
Machines Corporation (TMC),and Wavetracer offer SIMD computers, segmented by
 Conclusion 315

price and application.About twenty computer companies offer multicomputersbased

on interconnected microprocessors. A11 are aimed at providing large amounts of
computation for highly parallel applications, generally in the technical computing
market.
In the spring of 1983,TMC startedup in Cambridge,Massachusetts,to build a large,
massively parallel SIMD computer. The company was initially financed by a number
of private investors, including William Paley and Frank Stanton of CBS. By 1990, over
$40 million was raised to finance the firm, which has a valuation of nearly $200 million.
TMC’s first Connection Machine was delivered in 1985.Unlike nearly all its SIMD
competitors,TMC sellshigh-end computersin the $1million to $10million range, which
it defines as supercomputers. As is the case with Cray and IBM products, the cost of
TMC‘s machines includes support personnel who help with customer applications.As
a by-product, these support personnel become experts in applications and help sell
additional computers. In 1989, TMC‘s first profitable year, the company’s product
revenues reached almost $40 million.
TMC is backed by the Defense Advanced Research Projects Agency (DARPA)
StrategicComputingInitiative,whose goal is to developa teraflop computerby the mid-
1990s.A substantialamount of the product development funding comes from DARPA.
For example, in 1989, TMC was awarded a three-year, $12 million contract aimed at
building a 1-teraflop computer. The SIMD and the multicomputer, unlike traditional
supercomputers, can be built in a scalable fashion to provide the most peak computing
power.Governmentagenciesand universitieseitherbuy or are givenDAIWA-purchased
machines in order to encourage, understand, and develop the use of such machines. By
1990, TMC had sold about fifty computers for large database retrieval, petroleum
exploration, and computational fluid dynamics.

CONCLUSION

This chapter has presented the stories of eight high-tech ventures to give readers some
insight into how it should-and shouldn’t-be done.The first two examplesillustrated
two classic failure modes for a start-up: no product but excellent marketing (Ovation),
and good product but wrong approach to the market (Analytica).The last six examples
illustrated how to create healthy component, system, and software firms. The Bell-
Mason Diagnostic (Chapter 10) was used as an overall framework for discussing and
evaluating the companies.
Chapter 12

TECHNICAL
WORKSTATIONS:
HEURISTICS FROM HISTORY

The first decade of scientific and engineering workstation development provides

useful lessons for information-technology start-ups. Many of the observations
presented in this chapter have been used as heuristics to develop the rules that form
the ”ideal” model for the Bell-Mason Diagnostic.
In the mainframe and time-shared systems of the 1970s, Teletypes and simple
alphanumeric cathode-ray-tube (CRT) terminals provided the primary user inter-
face. High-priced, high-performance terminals were employed for scientific and
engineering applications that required the display of graphical and image infor-
mation.
In the 1980s, distributed workstations interconnected via a local area network
(LAN)were developed. These featured high-performance graphics combined with
computation and were dedicated to a single user. This powerful tool enabled new
applications in computer-aided analysis, design, engineering, and visualization
that had previously been economically infeasible or impossible.
In 1981 and 1982, Apollo Computer and Sun Microsystems came out with the
first low-cost and practical workstations, followed by Hewlett-Packard (HP).These
machines were based on the Motorola 68000 and its successor microprocessors.
During the first five years after the 68000 was introduced, over twenty-five new
companies were started. Products were introduced so fast and furiously that they
were nicknamed JAWS,for ”just another workstation.” Soon after Apollo and Sun,

316
 Background 317

Silicon Graphics Incorporated (SGI) was founded to deliver high-performance

terminals and workstations capable of displaying 3-D objects dynamically.
By 1988, faster, second-generation workstations were introduced based on the
reduced instructionset computer (RISC)architecturesof Apollo, HP, IBM, Intergraph,
MIPS, and Sun. Ardent and Stellar introduced graphics supercomputers that
provided both very-high-performancegraphics and computation capability using
vector processing and multiprocessing. These entrants became the forerunners of
third-generation workstations capable of providing personal supercomputing.
By 1990,HP/Apollo, IBM, SGI, and Stardent had all introduced workstations of
such power that they became known as superworkstations. With the cost of 3-D
graphics dropping to affordable levels, permitting use by many professionals in
technical fields, these workstations are beginning to supplant traditional
supercomputing.
Nearly all the traditional mainframe and minicomputer companies regard the
workstation as a critical product. However, despite the great numbers of firms in the
business, a scant six of them provide 95 percent (byvolume) of the workstations sold.
Of these, two are 1980s workstation start-ups that have remained healthy and
independent:Sun and SGI.The other four are established companiesthat introduced
workstation products to serve their customer base, which was being eroded by the
new form of computing: Digital Equipment Corporation (DEC), HP (including its
Apollo division),IBM, and Intergraph. These big firms intend to become even bigger
players in the workstation field. IBM, for instance, has stated (in its R6000 product
announcement) that it intends to capture 30 percent of the workstation market by
1993.
The rampaging progress occurring on the workstation front does not mean that
the lowly PC is a thing of the past, however. Far from it! In 1990, PCs with VGA
graphics have about half the resolution of workstations and only display 2-D
graphics, but by using Microsoft Windows software on a PC, much of the capability
of workstations (including 3-D interaction) can be provided at a low cost due to the
high production volume of PCs. By the mid-l990s, it is possible that the workstation
market will decline into a niche, to be replaced by a range of high-performancePCs.
Alternatively, the workstation could limit the PC’s upward growth. Given the
differencein the softwarethat runs on the two environments(UNIXversus Microsoft’s
DOS and OS/2 evolving toward UNIX), the two are unlikely to merge.

BACKGROUND

The notion of a personal workstation is a matter not just of market application need
but also of technology, sociology, and economics.Owning one’s own computer has
been the goal of designers and users since computers were first introduced. The
318 Technical Workstations

earliest laboratory computers were used interactively and personally, even though
they often cost millions of dollars. In the early 1950s,MIT's building-sizeWhirlwind
computer (with CRT) was used to experiment with the first air-traffic-control
system, do computer-aided design, and visualize scientific and engineering simu-
lations-exactly what computers are still likely to be doing in the twenty-first
century.Readersinterested in an in-depthdiscussionof theemergenceof workstations
can refer to Goldberg's (1988)History of Personal Workstations.
In 1957, MIT and MIT's Lincoln Laboratory provided both the technology and
the engineer training that led to the formation of Digital Equipment Corporation.
DEC built the first small, interactive computers, which were the basis of the
minicomputer industry. Its first computer, the PDP-1 (with CRT), was introduced
in 1960at a price of $120,000. This might have set DEC on the path toward building
workstations, but it went on in the early 1960sto build large time-shared computers
because of the economics of sharing, while forgetting the paradigm of one person/
one computer upon which the company was founded.' Similarly, it did not take the
opportunistic path of building workstations based on Motorola's 68000 but instead
waited until it had a VAX on a chip. By then, however, Sun was too firmly
established.
In the late 1960s, the key ideas for graphics were being researched at the
University of Utah by Dave Evans, Ivan Sutherland, and their graduate students
under the sponsorship of DARPA (Defense Advanced Research Projects Agency).
Evans and Sutherland company was created to commercialize this technology. The
most important output of Utah and Evans and Sutherland was the generation of
alumni who created workstations and graphics, including Jim Blinn (JetPropulsion
Laboratory);Ed Catmul (Pixar);Jim Clark (SGI); Charles Csuri, Phong Bui-Tuong,
and H. Gouraud (who developed shading algorithms); Alan Kay (Apple); Martin
Newell; and John Warnock (Adobe).
As is the case with virtually all technological shifts, the suppliers of graphics
terminals connected to mainframes and minicomputers were unable to make the
transition to become workstation suppliers. Such a transition would have required
terminal companies to fund and become expert in computer systems and their
application.

1. I accept much of the blame for this oversight, since I headed research and development at Digital from
1973to 1983. When a computer is shared, the operating system's goal is to prevent a user program from
gaining access to a terminal. With workstations, the major goal is to provide the best possible
communication between the display and the program.
 Xerox 319

XEROX
Xerox's Palo Alto Research Center (PARC) built the first distributed LAN-based
computing environment. Human-interface aids such as the "mouse" and "icons"
were derived from EngeIbardt's Augmentation Research Center at the Stanford
Research Institute during the period 1964-1975. Although the PARC researchers,
trained in computer science, were directed to provide a computing environment for
the office,their main concernwas providing a computingenvironmentfor themselves.
Workstations and time-shared systems exemplify instances where the designers
themselves were the first users or target consumers of the system.2They wanted a
computing environment that they could own individually and that would therefore
not degrade with more users.
The PARC workstation, based on Data General's Nova architecture, was called
the Alto. Altos were interconnectedvia the first local area network, a 3-megabit-per-
second ancestor of today's Ethernet. By October 1976, the first Altos operated in a
network environment with applications, and the first user's manual appeared. In
1980, Xerox introduced the STAR workstation based on the Alto. Lampson (Table
12-1) has traced the ideas that influenced successor commercial computers. An
accountof PARC's influence oncomputing is also given inFurnbling theFuture (Smith
and Alexander, 1988).
The following are some observations based on the Xerox PARC story:

Anytime a system can be built in which the designers are also the users, the
system evolves much more rapidly, with higher quality, and to a higher state
than with other types of products that rely on specification^.^

The disadvantage of designing a product for oneself is that such products,

though powerful, may not be easy for novices to use.

Achieving successin the laboratory is no guarantee of successin the marketplace.

9 Unless a large organization is already in the business in which new-product

2. Hewlett-Packard successfullyincorporated this idea into its culture from the beginning, as "the next-
bench syndrome," whereby HP engineers built instruments for use by their fellow HP engineers.

3. UNIX is an excellent example of such a system designed by computer scientists for users with a
computer-science background. Because it was oriented toward technical users, UNIX provided an
opportunity for other start-ups to develop a superior graphics user interface for humans.
320 Technical Workstations

Table 12-1. Descendants of Xerox PARC‘s Alto.

Engineering workstations PERQ, Apollo, Sun, Tektronix, DEC

AI [artificial intelligence] workstations Xerox llxx, MIT Lisp machine,

Symbolics and Lisp Machines, Inc.

Personal computers Xerox 8065, Apple Lisa, Macintosh,

Metaphor

Office workstations Xerox 8010, Convergent, Apple

Macintosh, Grid

Graphics terminals BBN Bitgraph, Bell Labs Blit

Local area net[workls Ethernet/IEEE 802.3

Network protocols DARPA TCP/IP, Xerox Network

Services

Laser printers Xerox 9700, etc.; Imagen; Apple

Laserwriter

Printing protocols Xerox Interpress, Adobe Postscript

Servers 3Com file server, Xerox 8044, Apple

Laserwriter

User interfaces Xerox 8010, Apple Macintosh,

Microsoft Windows, ParcPlace
Editors MacWrite, Microsoft Word

Illustrators Xerox 8010, Apple MacPaint and

MacDraw, Aurora Systems

Source: B. Lampson, ”Personal Distributed Computing: The Alto and Ethernet Software,” in A
History of Personal Workstations, ed. Adele Goldberg, 0 1988, by ACM Press. Table 2 , page 331.
Reprinted with permission of Addison-Wesley Publishing, Inc., Reading, Massachusetts.

research is carried out, the company will find it quite difficult to transfer n e w
technology a n d products into either a new division or the mainstream of the
firm.

It is easy to transfer research results to start-ups if a n entrepreneurial environ-

ment, such a s Silicon Valley, exists.

The major beneficiary of general corporate research is most likely to be those

w h o leave a n d start companies.
 Three Rivers’ PERQ 321

THREE RIVERS’ PERQ: THE PIONEER GETS THE ARROWS

The first start-up to use the Alto workstation idea was the Three Rivers Computer
Company, located in Pittsburgh. In 1978, Brian Rosen, a former PARC researcher,
rejoined the firm (where he had worked while a CMU student) to design and build
the PERQ computer. The location seemed appropriate because computer scientists
at Carnegie-Mellon University (CMU) believed in and proselytized PARC‘s dis-
tributed model of computing. Many of PARC’s researchers were CMU alumni, and
the computer-science community was tired of sharing large, overloaded, time-
shared DECSystem-10s. Furthermore, several CMU Science Department members
invested in the company, bought PERQ computers for the department, and did a
substantial amount of software development for the PERQ.
CMU had outlined the idea for the computing environment in a 1979report for
a large research project, stating that ”the era of timesharing has ended.” Time-
sharing would be replaced by powerful ” 3 M (1million instructions per second, 1
megabyte of primary memory, and a 1-million-pixel screen) distributed personal
workstation computers, like the models seen at PARC. The first PERQ was delivered
in 1980.
By 1982, Three Rivers was delivering tens of computers per month and had an
original equipment manufacturer (OEM) relationship to supply computers to ICL
for the British and European markets. Although Carnegie-Mellon was developing
a substantial amount of software for the PERQ, Three Rivers was unable to
incorporate it into the product and hence had a technology-transfer problem. By
1983, workstations from other companies were all being fabricated easily and
cheaply using the 68000 microprocessor.The PERQ was based on designing a more
expensive computer using microprogramming, a premicroprocessor method of
computer design. The firm ceased operations in 1985 after delivering about 350
workstations.
The following are some observations based on the PERQ story:

Being first is no guarantee of success.

In fact, being first is often a detriment if the wrong basic technology is used.
Potential competitorsbenefit from seeing how to make the product a better way.

Hiring people with experience in the computer industry to build a Pittsburgh-

based computer systems company proved very difficult, even though CMU
trains key people in the industry. Several software-only start-ups, such as
Tartan Labs, were unsuccessful because their founders lacked good local role
models.
322 Technical Workstations

In Pittsburgh, the infrastructure (e.g.,manufacturing, marketing, legal, and PR

firms) to support computer manufacturing and companies in the computer
industry is quite small. If an area is to support a technology, it must build up a
complete infrastructure that understands the industry and how it operates.
Areas such as Austin and Boulder have built industries to support high-
information-technologystart-ups,whereas Pittsburgh’s industrial infrastructure
supports heavy industry, such as steel and petroleum.

Transferring software from a university to a company/product is difficult at

best, even though the survivability of the company and the efficacy of the
product might depend on the transfer.

The best way to transfer software is to transfer the people associated with its
development (e.g., Bill Joy brought the Berkeley version of UNIX, BSD 4.2, to
Sun).

A company must be able to translate its own vision of computing, which it

usually has, into a vision that appeals to users and other third parties who might
apply the product to solve actual problems.

PERQ was evidently unable to do this, since it was ineffectiveat securing either
OEMs or end users.

Usually, there exists a single technology path on which to base a product, and
virtually every successful company follows that path. Those that do not follow
the right path perish.

In the case of workstations, an open standard based on the microprocessor and

UNIX was the correct path. PERQ was not based on the right technology and had
to develop its own system. When the market finallybegan to open up, doing all
that development proved too expensive and time-consuming.

A general-purpose computer built in a mass-production fashion almost always

wins against special architectures.

For example, two specialized LISP-language workstation companies formed

during the early 1980s: Symbolics and Lisp Machines, Inc. LMI is defunct, and
Symbolics is reduced to selling software, because traditional microprocessors
provided greater performance at negligible cost.

At DEC, I institutionalized the principle governing what to make versus what

to buy: “Make what we sell, not what we buy.”
 Apollo 323

If a part or product can’t be sold competitively at the required level of integra-

tion, buy it. There may be other considerations,but this one is key, obvious, and
often overlooked. The inability to make good make/buy decisions can become
a fatal flaw for both new and well-established companies.

APOLLO

The concept for Apollo originated with several researchers and engineers at Prime
computer, following the outline of the CMU Spice Project that was using PERQ
computers. The idea was to replace the central minicomputer and its expensive
graphics terminals with distributed personal, computing graphics workstations.
Prime, however, rejectedthis notion becauseit believed itself to be in the minicomputer
business. This attitude is reminiscent of Levitt’s (1986)classic observation that ”The
railroads didn’t understand they were in the transportation business.”
In January 1980, Bill Poduska, one of Prime’s founders, Mike Greata, of engi-
neering, and Dave Nelson, head of research, left Prime to secure funding for a
workstation company. The Motorola 68000, the first microprocessor to have the
necessary characteristics for a workstation (32-bit address space), had just been
announced. The Prime engineers, as MIT Multics alumni, had experience in time-
sharing and virtual memory. They also had experiencein distributed LAN systems,
having introduced a proprietary LAN ring to interconnect Prime minicomputers.
Apollo’s concept stage lasted about a month. The firm was incorporated in mid-
February 1980,and about ten people began developing the product. The average age
of the founders was over forty, and all had start-up experience. The two-month seed
stage was self-fundedby the three founders while they planned the company, wrote
their business plan, and secured capital. During the seed stage, the team continued
product design and project planning in various living rooms and kitchens. The team
was melded together from the two cultures of research and development. On April
22, funding totaling $1.6 million was secured based on a six-page, handwritten plan
(page 391, and the firm moved into its first building.
In March 1981, after ten months in development, the first two-node networked
system was shipped to Harvard University. Apollo shipped $3.5 million (versus a
planned $5.5 million), $18 million, and $80 million and went public with an initial
public offering (IPO)in March 1983.One of its customers, Mentor Graphics, started
up at the same time, using the Apollo environment to create an electronic computer-
aided design (ECAD)system for logic designers. About half of Apollo’s sales were
to Mentor during much of its independent life.
324 Technical Workstations

By the end of 1983, with an annualized run rate of about $100 million, Apollo’s
market value was $700 million. Poduska claimed that sixty of the thirteen hundred
people in the firm became millionaires. In early 1989, before HP acquired the
company, Apollo had annual sales of $700 million, and its valuation was about $300
million.
While Poduska was the chairman and CEO, Charlie Spector joined from Digital
as president. Following Spector’s departure, Tom Vanderslice was brought in as
president and CEO in 1985.Vanderslice had managed a division at General Electric
unrelated to computers or communications and had most recently been at GTE
working in telecommunications.
After a year of studying how to build higher-power workstations, Poduska left
Apollo to form Stellar Computer (incorporated in January 1986),with the purpose
of building a graphics supercomputer to attack 3-D problems that Poduska believed
could be solved neither by Apollo’s existing 68000-based product line nor by the
new DNlOOOO architecture then under development.
In 1989, Vanderslice sold Apollo to HP after it began to decline and become
unprofitable. Nelson left to start Fluent Machines, a company that would build a
multimedia workstation.
The Apollo experience gives rise to the following observations:

Understanding and mastering the technologies is what enables a start-up to

remain healthy and meet its product development schedule.

Nearly all the technology required to build the Apollo workstation environment
was fully understood by the company’s founders. Although they had not done
it before, they did have experience and spent a significant amount of time
building the basis for Apollo’s technology (i.e.,determining what to do and how
to do it).

When blending people (in this case, engineers)from several cultures, it is critical
to have an integration period during which people come to agree on goals,
constraints, and values.

During this initial period, various members are likely to leave. But as a project
is started, the entire team must agree on its direction!In fact, venture capitalists
commonly insist that a new venture’s founders have worked together before.
The seed stage can accomplish this goal. Many successful start-ups require each
new member of the team to spend up to six months on probation before finally
being accepted into the company.

An initial relationship with an OEM customer can be a serendipitous event to

fuel a newly launched company.
 Sun Microsystems 325

Such an OEM is likely to be a start-up itself, since a large firm will probably not
want to risk buying from a start-up until the new venture is well established.

Although conventional wisdom holds that entrepreneurs should ”never put

risks in serial by basing a new start-up on another start-up,” this observation is
flawed.

When entrepreneurs want to found an innovative company, their search for

new technology is likely to lead them to a start-up. Part of the reason Mentor
succeeded was that it concentrated on software and left the hardware to Apollo.

From the earliest experiences in computing, when General Electric formed

several unsuccessful computer divisions, it became obvious that hiring a
general manager from outside the computer industry to head a computer
division is risky.

Apollo declined under the leadership of Tom Vanderslice. SteveJobshired John

Sculley, an event that ultimately caused Jobs to leave Apple. Whether Apple has
been and will be better or worse off without a visionary as its leader won’t be
clear until 1995, a decade after Jobs’ departure.

Beware of being overconfident when a new technology appears.

Having mastered past technology is no guarantee that a company will master

future technology. Stellar was founded on a product architecture and custom
complementary metal oxide semiconductor (CMOS) chips; the project took
twice as long as predicted, both because mastery of new technology was
required and because the company built the processor portion, which it could
have bought.

SUN MICROSYSTEMS

PARC also inspired Professor Forest Baskett and Andy Bechtolsheim, his graduate
student at Stanford, to build the Stanford University Networked (Sun)workstation,
which became the basis for Sun Microsystems. Vinod Khosla was the entrepreneur
who stimulated the founding of Sun. Vinod was employed by Daisy and was
searching for technology to make an electronic CAD workstation. Vinod became
convinced that a general-purpose design was required and would be used by many
other engineers. In May 1980, he visited Stanford and saw Andy Bechtolsheim and
the Sun workstation. Andy offered to serve as a consultant on the design for $10,000.
However, Vinod convinced Andy to abandon his quest for a Ph.D. and help him start
326 Technical Workstations

Sun. Vinod has succinctly expressed a most critical principle of technology transfer
as follows: ”Given the choice of the goose and the golden egg, I’ll take the goose.”
By 1981, about twenty Sun workstations had been under test at Stanford for
roughly a year. In February 1982, Vinod Khosla, Andy Bechtolsheim, and Scott
McNealy founded Sun Microsystems. In June, Bill Joy, who was responsible for the
enhancements to the Berkeley version of the UNIX operating system, joined Sun to
head software. Scott had been Vinod‘s Stanford business-school classmate and was
responsible for manufacturing at the first UNIX computer start-up, Onyx. The
team’s average age was twenty-seven. Several other companies-including Codata
(defunct), Forward Technology (defunct), Cimlinc (formerly CADLINK), and Sili-
con Graphics-licensed the Stanford design, but none succeeded in building and
marketing the original design.
The Sun Microsystems plan was six pages long, and it succeeded in getting the
team $300,000, plus a $200,000 line of credit with an option on the next round of
financing. The seed stage plan was to (1)fund the preparation of a proper business
plan, (2)get a marketing person, and (3)make a prototype. A proper business plan
was never written, but an operating plan was put in place. The group proceeded to
start building the Sun workstation and selling it sans software, as a Tektronix
terminal emulator, followed by a workstation running Unisoft’sUNIX. The company
turned a profit in September 1982,after a few months of operation, and has remained
profitable almost continually since then.
In early 1983,Owen Brown, a former DEC regional sales manager, was hired as
Sun’s president to solve the marketing and sales problems and stayed for a year.
Vinod remained CEO until 1984,when Scott McNealy took over that position. In the
summer of 1983, Bernard Lacroute was hired as chief operating officer. Bernie had
been at DEC, where he was then responsible for all communications products,
including the development of Ethernet.
Sun’s revenues for the first six years (1983-1988) were $8 million, $39 million,
$115 million, $210 million, $538 million, and $1.052 billion. Sun went public in
February 1986. By January 1989, Sun’s revenue rate was running at roughly $2
billion, reflecting more than a doubling of size every year. The shareholder value of
the company was roughly $1.3 billion, representing a 300 to 1 return on the
investment.
Both Apollo and Sun emphasized relationships with third-party suppliers that
would take their product into end-user environments, either by reselling with
software or by joint marketing. Sun tended to emphasize reselling, and hence, its
growth rates became significantly higher than Apollo’s, even though Apollo had
initial success with Mentor Graphics as an electronic CAD supplier.
 Sun Microsystems 327

The following are a number of observations based on the Sun experience:

A product prototype that has been funded by a public research institution,

together with the people who embody the technology, is an ideal start-up
source.

DigitalEquipment,SiliconGraphics,Sun,and Evans and Sutherland all followed

this pattern of technology transfer.

Advanced development at another company can be a source of start-up tech-

nology, albeit this source is somewhat subject to possible litigation.

Apollo came from Prime, and Stellar (Stardent) came from Apollo. Amdahl
Computer came from an IBM laboratory, and Trilogy came from Amdahl.

As a source of start-up technology, corporate research laboratories are even

better than advanced development, albeit this source, too, is somewhat subject
to possible litigation.

The closer a start-up is to research, the higher the company’s likelihood of

producing more lasting technology. Many start-ups came from Xerox’s PARC,
including 3Com and Adobe.

Although many recommend that start-ups build niche products for niche
markets, as a rule, building nongeneral products that only solve a particular
user problem is an expensive approach that often proves fatal to a firm that
resorts to it.

Of the three original and largest CAD companies (Daisy, Mentor, and Valid),
Mentor has been the most successful, because it used Apollo workstations and
concentrated on solving user problems, not on building just another worksta-
tion. By solving an overly specific problem while trying to maximize its return,
a firm is likely to miss the larger market.

A company that makes a product for a specificcustomer is doing so at very high

risk, since the customer and its need may change and thereby invalidate the
product concept.

Wilson, a former head of General Motors, once observed: “What‘sgood for GM

is good for the country.” In the high-technology industry, however, “What‘s
good for Company X is, at most, only good for Company X.”
328 Technical Workstations

Technologytransfer is very unlikely to occur without the transfer of key people.

As mentioned above, Khosla has succinctly expressed this idea by stating his
preference for taking the goose rather than just the golden egg.

In cases where a product is available to a start-up right from the outset,

becoming immediately profitable should be the goal!

Start-ups that are fortunate enough to operate on this basis appear to have an
unfair advantage. The basic circuit-design technology on which DEC was
founded, together with the $70,000 initial funding that it received in 1957, were
enough to make the firm profitable from the beginning. Apple started with
minimal financing and became profitable, as did Microsoft.

Profit is habit-forming.

To ensure viability, profit must be the company’s principal goal from its
inception.

If a hardware product is to be appealing to the ultimate buyer/user, it must be

usable (i.e.,include the necessary software) rather than having to be built up by
the buyer /user.

Therefore, relationships with other firms (called ”strategic partnerships”) are

critical in order to finish the product, make it useful, and distribute it. These
relationships typically follow one of three patterns: (1)the start-up and another
company jointly market a system; (2) a third party resells the system in an OEM
relationship; or (3) the system manufacturer becomes the system supplier and
provides specialized software. Sun’s revenues reflected the first two strategies,
whereas Apollo primarily adopted the third approach.

”Open systems” are more appealing to users than proprietary systems.

Given Sun’suniversity origin, the company operated to provide what appeared

to be an ”open system,” whereby criticalsoftwareinterfaceswereavailable from
multiple vendors. Sun developed its Network File System (NFS) and made it
available to all suppliers and, more recently, did the same with its processor
architecture.SPARC was also licensed to a number of vendors. Only in this way
can a lasting new computer class be established,because it has multiple sources
of supply.

Sun retained a number of critical interfaces,including its windows and network-

sharing software. Sun may have failed to sponsor and make the standard in
 Observations About Founder Involvement 329

windows and most likely will have to follow the X Window standard, which
was sponsored by MIT. Apollo created a richer, but proprietary network,
initially in the style of the traditional industry. Thus, users entering the Apollo
environment, Domain, appeared to become locked in.

When designing products, a manufacturer must either make the standard or

follow it.

If a company tries to make the standard and fails, it gets to implement the system
twice-the first time the company’s way and the second time according to the
standard.

The rule that says “Only back experienced people” should not be followed
blindly.

A very smart team, even though inexperienced, may perform significantly

better, be more dedicated, and remain together longer than a team of mercenaries
that’sdoneitonceor twicebefore. Ifnothingelse,byhavingayoungerteam than
Apollo, Sun has survived longer and developed a substantially more interesting
and vibrant culture that produces creative products.

OBSERVATIONS ABOUT FOUNDER INVOLVEMENT

Additional insightscan begained by contrastingtheexperienceof founder-controlled
companies with the experience of companies whose founder has relinquished
control, as shown in Table 12-2. (A case in point is the fact that, in 1990, Apollo’s
founders have left the firm, whereas Sun’s founders have all remained involved.)
The following are some observations about the impact that the founder’s
departure or continued participation can have on a high-tech venture:

To create a company that will be successful over the long term involves a
lifetime, not just a five- to ten-year commitment on the part of the founder.

In 1990, three of the vital, working founders of Sun (Bechtolsheim, Joy, and
McNealy) remain with the firm and are still bachelors. The founders are key
leaders in various parts of the company and are able to help propagate the
culture in a rapidly growing organization.

When the founder responsible for the leadership and technical integrity of
products leaves a company, it is likely to flounder and enter a state in which it
sustains its user base but doesn’t set off in any new directions.
330 Technical Workstations

Table 12-2. Founder Control in Various High-Tech Companies.

Firms Whose Founders Have Firms Whose Founders Have
Retained Control or Remain Involved Relinquished Control
AMD, Cypress, Intel, National Fairchild, Texas Instruments, Motorola
Conner, Priam, Quantum, Seagate CDC Disks, Maxtor, Shugart

Compaq, Dell Altair, Apple, Commodore, Convergent

DEC, HP, Tandem, CCI, DG, Harris, Prlme, SEL, Wang
Silicon Graphics, Sun Apollo
Alliant, Convex Floating Point Systems
Microsoft Lotus

Oracle Relational Technology

Cray Computer CDC, Cray Research
Amdahl, Honeywell, IBM, Unisys
3Com, Ungerman-Bass (now with Tandem)

Certainly, Lotus changed when Kapor left, and HP changed when Hewlett and
Packard turned over the management to Young. In 1990, HP introduced a new
management concept by appointing Morton as co-CEO.

With a few notable exceptions, those companies that retain the founder enjoy a
continuity of culture.

Founder-controlledfirms tend to remain at the forefront, reflecting the founder’s

involvement, provided the founder has business and market savvy and is
technically competent.

A counterview of why an entrepreneurial founder leaves is that the company

may for some reason have failed to meet his or her expectations.

Failure can involve lack of product or market, team, leadership, etc., and thus,
the founder leaves as a reaction to the firm-a push, rather than the lure of a new
venture. This scenario is particularly common among “chronic entrepreneurs,”
who continually drive to found new companies.

Multiple, competitive companies are always created as each new technology or

computer class is formed or as each new product type is introduced.
 Stardent and the Elusive Graphics Supercomputer 331

Table 12-2illustrates the multiplicity-of-companiesprinciple.Examples include

Univac (now Unisys since it merged with Burroughs)and IBM in the early 1950s;
TI and Fairchild, which started the semiconductor industry; DEC and SDS (now
defunct) for minis; Cray and CDC for supercomputers; Apple, Altair and
Commodore; Apollo and Sun for workstations; Alliant and Convex for
minisupercomputers; and Ardent and Stellar, which created graphics
supercomputers.

SILICON GRAPHICS

Silicon Graphics started up in the early 1980sto provide 3-D graphics terminals that
were connected to minis and mainframes. By 1985, when more powerful 68000
microprocessorsbecame available,they were expanded into 3-D workstations. The
basic design for the Geometry Engine chip, which transformed polygons into a 3-D
space, was completed in 1982by SGI founder Jim Clark while at Stanford. Clark, an
alumnus of the University of Utah, acquired the basic design ideas from his own
thesis work and the powerful, highly pipelined Evans and Sutherland displays. In
1986, SGI adopted the MIPS architecture for its computational engine and began
supplying a range of products from low-cost, diskless 3-D workstations to multi-
processor workstations and computational servers.

STARDENT AND THE ELUSIVE GRAPHICS

SUPERCOMPUTER

In January 1986, Stellar and Ardent (originallycalled Dana) started up in Belmont,

Massachusetts, and Sunnyvale, California, respectively. The basic plan of both
companies was to create a new computer class that would have substantially higher
computing and graphics performance than existing workstations.
Ardent started with a $12.5 million first-round investment based on a plan put
together by its seven founders during a month-and-a-half-long, accelerated, self-
funded seed stage. Stellar obtained roughly the same level of funding based on a
plan that had been in gestation for roughly a year.
Ardent’s ten-page business plan, dated January 20,1986, identified a number of
risks that might jeopardize the plan. Table 12-3shows the assumptions Ardent made
in its plan regarding each of these potentially problematic areas and contrasts those
assumptions with the actual outcomes.
In 1987, Ardent secured the second round of funding from Kubota Ltd., which
took on the following functions:manufacturing the products in Japan, marketing in
332 Technical Workstations
~~

Table 12-3.Assumptions Made in Ardent’s Business Plan Versus Actual Outcomes.

Risk Assumption Outcome
Staffing Would take 1-3 months. Took 6 months.

Large-scale integration (LSI) Looked hard, risky. Not a significant

problem but was a
bit more costly than
anticipated.

Software Would be complex. Did more than the plan:

Operating system
( O / S ) ,compiler, and
visualization
software exceeded
plan.

Hardware Was not mentioned. Required almost twice

the originally
anticipated
engineering effort.

Customers 70% would be OEMs. OEM strategy failed to

materialize; had to
build a sales force
oriented toward end
users.

Applications Would be necessary. Took much time and

was costly. Users
wanted more.

Offshore manufacturing Was provided for Occurred and was

in the plan. much better than
expected with
Kubota.
Product cost Would be $50,000. Was over $80,000. Too
far off plan. Major
perturbation.

Other First (beta) shipment

was 5/88, not 7/87.

Japan a n d Southeast Asia, and developing software for mechanical design and
manufacturing.
As a graphics supercomputer, Ardent’s first product, Titan, was significantly
more difficult to developthan anticipated, resulting in almost twice the development
 Stardent and the Elusive Graphics Supercomputer 333

and product costs. However, the real flaw was introducing a first product that failed
to demonstrate clear performance superiority, as discussed in Chapter 8. Although
a second set of products overcamethis shortcoming,the inadequacy of Ardent’s first
product inflicted very painful damage on the firm. According to Allen Michels,
Ardent’s founder and CEO: “You get one shot.”
Rarely does a start-up get a second chance if it misses with its first product. The
first product must be designed, sold, and delivered correctly. The only reason the
company survived to merge with Stellar in October 1989was that Ardent’sJapanese
partners had a basic faith in the firm and the market. Poduska became president and
Michels chairman, and in 1990, Michels and Sanders left the organization.
The following are some lessons that can be learned from Ardent’s experience
with its graphics supercomputer:

Although a start-up can count on a long-established company to be lethargic

and noncompetitive,a high-growth firm less than ten years old is unlikely to let
a start-up enter its space.

Stardent underestimated the competitivenessof Silicon Graphics by classifying

it with established companies such as Apollo, DEC, HP, and IBM.

When attacking a walled city, a start-up shouldn’t telegraph its intentions to the
inhabitants by distributing promotional coffee cups, towels, and T-shirts. And
it should avoid attacking two walled cities at once.

When developing a high-risk, state-of-the-art product that has never before

been created, a start-up’s founders would be well advised to anticipate project
surprises (by a factor of 2) in schedule, resource requirements, product cost, and
specifications.

When a new venture is entering a crowded market where the niches are small
and perhaps hard to find, it helps to have a strategic partner with resources.

Ordinary venture financing would probably have given up on Ardent after

three years. Kubota is determined to be successful in technical computing and
is prepared to invest for a long-term gain.

A start-up must not skip the seed stage, especially if it is building a product that
the world has never seen before.

Detailed planning is essential. With a proper seed stage, Ardent would have
either cut the product to fit the funding or made a realistic product plan.
334 Technical Workstations

DEC, HP, IBM, ETC., AND THE PERFORMANCE WARS

Mainframe and minicomputer companies have come to realize the importance of a

distributed-workstation environment by watching Sun grow to over $2 billion in
1990 and take large fractions of their markets. By 1990,DEC and HI?share less than
half the market, and as the year has progressed, their share has dwindled even
further. HP’s divisional structure is organized to build and sell competitive 68000-
based workstations. In 1985, HP began using its own RISC chips, and it bought
Apollo in order to have a larger installed base of customers. DEC took longer to enter
the market, given its need to have VAX-on-a-chip processors as its workstation
engine. In 1989, DEC introduced a high-performance UNIX workstation using the
MIPS RISC chip as its ”Sun killer,” followed by a faster version in the summer of
1990.Sun responded with its SPARCstation,and Data General entered the fray with
a high-performance Motorola 88000-based workstation. In 1990, Evans and
Sutherland made its first entry into the workstation market with a high-performance
3-D workstation based on the MIPS microprocessor to challenge its progeny, SGI.
In the late 1980s,IBM introduced several technical Unix workstations in the RT
series (IBMs RISC architecture) and one based on the 68000.By 1990,none of IBMs
products was powerful enough to appeal broadly to the technical and software-
development community. Andy Heller, former president of IBMs Workstation
Division, offered the following observation about the effect of having a long
gestation period before a product finally becomes available: “Technology is like
fish-the longer they stay on the shelf, the less desirable they become.” In February
1990,IBM introduced a range of UNIX-based workstations and servers that is likely
to retain a product edge for one to two years. These products are based on IBM’s
next-generation,superscalar RISC technology, which provides substantially higher
performance for technical applications than ordinary RISC processors.
Given the push for performance as the differentiator,all companies have chosen
different strategies for achieving supremacy. The multiprocessor approach intro-
duced by Apollo, Stardent, and Silicon Graphics allows an arbitrary amount of
power to be placed in a given computer. With several firms-such as DEC, Data
General, and various Intel 486- based workstation companies-introducing such
products in 1990, the 1990s will truly see the advent of parallel processing.
Some further observations on the performance wars:

Large companiesbuild every conceivableproduct as an advanced development

project, often years ahead of any start-up.

These efforts rarely affect products and are often the basis of a start-up such as
Apollo.
 Whatever Happened to JAWS? 335

When a new market opportunity appears, entrepreneurs should go for it with

a start-up.

Even if large, established companies are working on products, or have devel-

oped prototypes, in the area in question, these firms are lethargic and uncreative,
and the likelihood of their producing a good product is very low in the short
term (and sometimes even in the long term).

Even if its products are mediocre, a large company will have a high market
share.

Large firms have a very big sales force and do not like to lose control of their
relationship with a customer.

Inevitably, all large companies will enter every large market area because their
ego forces them to do so.

This tendency represents opportunity, not competition, for a new venture. The
large firms will either buy the start-up’s products or buy the company itself, and
if they don’t, their competitors will.

IBM always enters every product and market area and becomes a major winner
(e.g.,minicomputers, PCs, workstations, word processors).

Rarely does IBM fail, as it did in home computing with the PCjr. When it does
fail, it learns from its mistake and tries again. At some point, IBM will be a highly
successful home-computer supplier.

WHATEVER HAPPENED TO JAWS

(JUSTANOTHER WORKSTATION)?

With the introduction of the microprocessor and Ethernet, and the existence of a
standard operating system, UNIX, over fifty more companies designed new
workstations during the 1980s, among them Adage, Cadmus, Celerity, Lexidata,
Lundy, MassComp, Megatek, Mosaic, Raster Technologies, Ridge, Sanders,
Symbolics,Tektronix, and Vector General.The situation was similar to the late 1960s
and early1970s, when ninety-two firms began making minicomputers (see page
175).Sincethere were fewer workstation start-upsthan minicomputer start-ups, one
might conclude that either it was harder to create a workstation company (true),the
perceived market was smaller (true), or less capital was available (not true).
Entrepreneurs and the venture capital community may finally have learned from
336 Technical Workstations

experience. (Although the plethora of start-ups to make minisupercomputers and

supercomputers in the 1980s for a very small market shows that the lesson about
“too many start-up companies chasing a small market” was either poorly learned or
soon forgotten.)
In October 1988, Steve Job’s start-up (NeXT)entered the market with a 68000-
family-basedarchitecture.The workstationofferedmany incrementalimprovements
over the Macintosh, including a built-in Ethernet, large CD ROM, signal processing
for audio and fax input/output, the use of UNIX (through the MACH operating
system), and NeXT Step, an improved graphical user interface.In October 1990,the
second-generationproducts were announced, with RISC workstation performance,
true color, and built-in video with compression. Most important, the second
generation attracted numerous application programs for laboratory and desktop-
publishing use, including multimedia and video. NeXT products, with sufficient
marketing, could be a substitute for products from Data General, DEC, and HP/
Apollo, as well as Apple.

THE CLASH WITH THE PCs AND THE APPLE MACINTOSH

The IBM PC and its hundreds of clones present a threat to the workstation. The
evolving capability of Microsoft DOS-with Windows 3.0 providing an easy-to-use,
multitasking environment-will make the PC competitive with the UNIX-based
workstations in the mid-1990s. Furthermore, the PC has accumulated almost a
decade of software packages that system developers and users support.
What differentiates the personal computer from the workstation? Not much.
However, workstations are already in the large-screen, multiprogrammed, dis-
tributed-processing space, but with more capability and more power, and they are
less expensive. Intel 486-based PCs have comparable processing power. When PCs
become available with large, high-resolution screens and the ability to carry out
multiple tasks in parallel and also provide transparent distributed computing so
that the network acts as a single system, workstation growth will be limited.
The future of the Macintosh is cloudy. It has only evolved in the most obvious
fashion to use faster 68000 processors, larger memory chips, and bigger screens. It
has a cadre of loyal users and software developers, although with the introduction
of Windows 3.0 to provide a ”Maclike” interface on IBM PCs, it is fundamentally
doomed to a niche, with a rapidly dwindling market share and only modest or
declining growth. Authors will continue to use Macintoshes until they switch to the
PC, where a wider range of nonproprietary hardware, including portable comput-
ers, awaits them at much lower costs.
Although the Macintosh local area network, Appletalk, is inexpensive, it has
only a tenth the bandwidth of Ethernet, the network most commonly used for
 Conclusion 337

workstation and PC distributed computing. Almost no capability exists for distrib-

uted computing, and because the Macintosh, like the PC, is locked into a particular
CISC (complex instruction set computer) architecture, the Motorola 680x0 series,
the prospects for an evolution in power are minimal. Because of the bus structure
used in the MAC I1 series, those computers are more expensive than workstations,
which have no external buses.
The moral: If Sun and other workstation vendors can establish really good
channels of distribution, such as retailers, for their products, users will understand
that workstations are a bargain compared to large PCs. Alternatively, PC makers
(including Intel) may rise to the workstation challenge and begin providing pow-
erful, low-cost PCs that challenge workstations.

THE 1990s: TECHNOLOGY STRIKES AGAIN

In the course of creating a workstation environment that would be competitive with
a single shared system, a number of technologies were developed. One of those
technologies, the X Window system, was developed to allow windows on one
system to access another system using the local area network. In this fashion, a user
can merely have a window terminal connected to an LAN that provides access to a
time-shared system. Thus, stillanother approach that would becompetitive with the
workstation is the return of time-sharing in a fashion that’s completely transparent
to a user.
The ebb and flow among centralized versus fully distributed computing is a
recurring theme in computation. Just as minis, PCs, and workstations (which made
every user a computation-center director and system administrator) were an al-
ternative to centralized computing, X Window terminals connected to servers
represent a return to centralized computing services.
The group that developed UNIX at Bell Laboratoriesis using a new system, Plan
9, that is based on users connected via highly interactive, person-serving terminals
that access central facilities for data and computation. Of course, such a system
requires an excellent, high speed, worldwide communication network. Other
technologies will also influence the evolution of the workstation, including high-
definition TV and compressed digital video, enabling the development of ”multi-
media” computing.

CONCLUSION

Workstationcompaniesareintroducinglower-pricedstations to be used as personal

supercomputers for high-performance and 3-D applications. By 1995, the relentless
338 Technical Workstations

reduction in the cost of memory, coupled with increasesin processing performance,

will force the workstation into the personal computer market space by default for
nearly all but a few high-performanceapplications.Bill Joy of Sun Microsystems has
predicted the “300M workstation” by 1992and the “3G workstation” by 1995.Each
of these workstations will provide 100 million and 1,000million (giga-)instructions
per second,respectively; carry out 100million and 1,000million(giga-)floatingpoint
operations per second; and have 100megabytes and 1gigabyte of primary memory.
Intel’s 1989announcement of the 80860 microprocessor and graphics processor
chip as a ”Cray on a desk” could have a great effect on technical computing, because
it means a single chip can deliver both computational and graphics performance. In
this way, the PC and the technical workstation could be merged by any development
team capable of assembling a few chips together. This confluence of technology to
make a technical workstation/PC using both the 486 and 860 chips would pit the
UNIX world against the PC (DOS and OS/2) world. On the other hand, OS/2 is
evolving toward a full-scale,general-purpose, multiuser operating system that will
no doubt support both the UNIX and DOS programming environments. Thus, an
evolutionary path for most PC users appears to exist.
In short, technology evolution beats revolution every time it can get there fast
enough and do the job.
Chapter 13

THE FUTURE

The future can be viewed from several perspectives. Chapter 8 extrapolated from the
past to providea product overviewof fifth-generationcomputing.Using this perspective,
the evolutionof hardware continuesto be the dominantforcein the creation of computer
hardware products. The availabilityof hardware at a given price and performancelevel,
in turn, paves the way for new software and applications.
The first section of this chapter presents the Intel view of the future. This is a good
starting point, because Intel is the leading supplier of microprocessors, and hence, its
plans ure the future.
In The Age of Intelligent Machines, Ray Kurzweil(1990)provides an excellent over-
view and compendium of intelligent machines. His book includes a time line for the
future, a portion of which is presented in the second section of the chapter. Ray's view
is highly optimistic, because he sees the world in terms of possibilities.
That section is followed by one that offers my own, more conservative views. It
summarizes what can be expected during the next decade, beginning with basic
physical technologies (semiconductors,magnetics, etc.), and includes the application
alternatives that 4 1 be made possible by new systems.
The last two sections of the chapter describe my view of the changes that can be
anticipated in the mainframe, minicomputer, and supercomputer industries when
cheap,powerful,and moreubiquitousdesktopand otherformsof distributedcomputing
become available, as well as my ideas on worldwide competition.
These various perspectives on the future should be helpful to readers interested in
identifying new business opportunities.

339
340 The Future

INTEL'S VIEW OF THE FUTURE

Intel provides the largest fraction of the worlds computing power, thanks to the
omnipresenceof the Intel/Microsoft PC (aka the IBM PC). Because such a leader plays
a dominant role in shaping the future, it is important for readers to understand Intel's
view of what that future holds.
Intel characterizesthe rapid change in computingperformance as a functionof time,
as shown in Table 13-1, which compares microprocessors with larger machines of
constant performance. Note that the first 8080 microprocessor,introduced in 1974, was
roughly equivalent to the IBM 704 mainframe, introduced twenty years earlier. Over
time, the microprocessor rapidly evolved to the point where it was only four years
behind the large IBM mainframes.In 1990,the combined 80486/80860microprocessor
is equivalentin power to the largest minicomputerfrom Digital EquipmentCorporation
(DEC).Although the microprocessorslisted sold for a few hundred dollars at most, the
processor portion of all the "mainframes" shown in the table sold for between $100,000
and $2 million.
Figure 13-1,from Intel, shows how the computingpower of a single microprocessor
has crossed over to exceed the power of DEC's minicomputersin 1988 and is projected
to cross over and exceed the power of IBMs mainframes in 1996.Figure 13-2,from the
Gartner Group, points out the incredible disparity in the number of instructionsthat can
be processed per second per dollar of hardware versus time for various computer
classes, from mainframes to workstations,PCs, and pocket computers.For example,in
1983,a mainframe costing $2 million provided about 4 million instructionsper second
(i.e.,mips), for a cost-effectiveness of 2 instructions per second per dollar. Eight years
later, in 1991,a relativelylow-cost,$2,000 personal computer delivering2 mips provides
about 1,000instructions per second per dollar.
Although the evolutionaryline of performance shown in Figure 13-1appears to be
continuous, the actual performance of the 80x86 microprocessor series has evolved in
discrete steps. Table 13-2illustrateshow the clock frequency doubled over a three-year
period, giving a 26 percent per year improvement, for each of the Intel products. The
table includes the clock frequency not only at the time each microprocessor was
announcedbut also at the end of the first, second,and third years thereafter, sincedesign
and process improvements often allow new versions of a microprocessor to be made
available during the next several years after the first model is introduced.
Not only are microprocessors becoming more powerful, but the total number of
chps (microprocessorsand support chips) required to build a PC is rapidly declining,
as shown in Table 13-3.
Although the cost of a particular computer will be substantiallyreduced by talung
it from 170chips down to 1chip in less than a decade, greater functionality is alsobeing
incorporatedinto a given PC. Figure 13-3,from Intel, shows the variousparts of a system
 Intel's View of the Future 341

Table 13-1.Yesteryear's Mainframe Is This Year's Microprocessor.

~ ~

Mainframe or Minicomputer Delay

-

Intel Year Micro Model Year Model (In Years)

__ -

1974 8080 1954 IBM 704 20

1977 2-80 3962 IBM 7094 15

1981 8086 1973 PDP11/45 8

8088 1975 PDP 11/70 6

1984 80286 1977 VAX-11/780 7

1987 80386 1984 (1982)" VAX 8600 3 (5)"

1989 80486 1985 IBM 3090 4

1990 80860 1986 IBM 3090/VF 4

80486/ 80860 1990 VAX 9000 0

'The project was two years late to market

1000

1976 1980 1984 1988 1992 1996 2000

Figure 13-1. Performance Versus Time for Mainframes, Minis, and Selected Intel Micro-
processors. (Adapted from a figure provided by Intel Corporation.)
342 The Future

-
100000

Workstation
aJ
L
cd
B 0000
2cd
G
cr
0
L
cd
=
0 1000
U
L
aJ
n
zm
MV 100
e
P,
U
c
8
\%
m
10
E
.
0
I
Y
0

U
?
H
z 1
1983 1985 1987 1989 1991 1993 1995 1997 1999
Figure 13-2. Estimate of Instructions Processed Per Dollar Versus Time for Various
Computer Classes. (Courtesy of the Gartner Group.)

that have been integrated into PCs of a particular size over the course of time. Each of
the following options evolved in a similar fashion, going from 100 chips' when first
introduced to a single chip three to seven years later:

The bus-interconnection scheme for controlling the chip set

Graphics, with evolving speed, resolution, and functionality

_.
1. This is the number of chips that can be placed on a printed circuit board for a PC option
 Intel’s View of the Future 343

Table 13-2. Intel Microprocessor Clock Frequency (in Megahertz) Versus Time.
Years After Announcement
Microprocessor 0 1 2 3
8086 5 6 8 10

80286 6 8 10 12

80386 16 20 25 33
80486 25 33 50 -

Communicationsmodems and local area networks (LANs)

Audio output

Video output

Speech output

Whether Intel will still be the dominant supplier of computers in 2001 remains to
be seen, given the formation of consortia of suppliers to build chips and competitive
systems based on MIPS’s R-series, Motorola’s 88000, and Sun Microsystem’sSPARC
architectures.Furthermore,if Microsoft’soperating-systemsoftwarebecomes available
outside of the Intel architecture, it could enhance the position of the three alternative
suppliers.The fourth computer generation,beginningin 1978, has evolved based on de
facto company-supplier standards from Intel and Microsoft. Having a sole-source
monopoly for the microprocessor and operating system almost defies the open-
architectureprinciple. Will the fifth generationcontinueto be definedby these company I
standards?
On the other hand, the supercomputer (Cray X architecture), mainframe (370
architecture), and minicomputer (VAX) and PC (80x86) machine classes have been
dominated by a singlearchitectureand supplier.Workstation architectureis still up for
grabs in 1990based on the use of the SPARC,MIPS, and IBM RISC (reduced instruction
set computer)for executing the UNIX operating system, with no single manufacturer
having more than a 50 percent market share.

SECONDARY AND TERTIARY MEMORIES AND

PAPER COSTS AND SIZES (CIRCA 1990)

The decline in the number of support chips and the cost of microprocessorsdiscussed
above has been matched by a continued rapid decline in the cost of memory.
344 The Future

Table 13-3. Number of Chips in a Personal Computer Versus Time.

YeaY Number of Chips in a Personal Computer (sans Memory)

1984 170
1987 70

1990 10

1993 1

Functional Integration

Desktop

c’ Portable
.-M
E
4
8
iij

Handheld
FOUR
FUNCXION P R O G R A m I N G COmAT,BLE
32-BIT SPEECH
SERIAL
LINK FLASH CUI SATELLITE
DISK LINK

1980 Today 2000

TimelFunctions
Increasing Functions Drives Integration
Decreasing Size

Figure 13-3. Evolution of Size and Weight and Increased Functionality of Personal
Computers with Time. (Courtesy of Intel Corporation.)

Chapter 5 discussed the sharp drop in the cost of primary memory (the fastest memory
most easily accessed by the processor), but the cost of secondary and tertiary memories
has also declined rapidly. Modern storage media are cheaper than paper, as indicated
in Table 13-4, w h c h gives the storage capacity (in megabytes) and the price per
megabyte of various memories in 1990.
 Ray Kurzweil’sView of the Future 345

Table 13-4. Cost Per Megabyte and Memory Size for Secondary Memories.
Media Cost ($)/Megabyte Megabytes Stored
8 mm tape 0.005 5000

4 mm (DAT) tape 0.01 1200

IBM 3480 tape 0.04 400

Write Once Read Memory 0.10 3200

CD ROM 0.20 620

Pulp paperback 2.00 2

5.25” magnetic disk 5.00 150-1200

Paper at $0.02/page 5.00 1.0-0.004

RAY KURZWEIL’S VIEW OF THE FUTURE

In The Age oflntelligent Machines, Ray Kurzweil(1990)offers his view of the future in the
form of the following time line:

Early 1990s A profound change in military strategy arrives. The more developed
nations increasingly rely on ”smart weapons” that incorporate elec-
tronic copilots, pattern recognition techniques and advanced tech-
nologies for tracking, identification and destruction.

Continuous speech systems can handle large vocabularies for spe-

cific tasks.

Computer speeds of 100 MIPS.

Application Specific Integrated Circuit (ASIC) technology makes

writing chip programs as easy as writing software.

Mid-1990s A multi-hundred-billion-dollar computer and information process-

ing industry is emerging, together with a generation of ubiquitous
machine intelligencethat works intimately with its human creators.

Significant progress is made toward an intelligent assistant, a deci-

sion support system capable of a wide variety of administrative and
information gathering tasks. The system can for example, prepare a
feasibility report on a project proposal after accessing several data
bases and talking to human experts.
346 The Future

Reliable person identification using pattern recognition techniques

applied to visual and speech patterns, replaces locks and keys in
many instances.

Accomplished musicians, as well as students learning music, are

routinely accompanied by cybernetic musicians.

AI [artificial intelligence] technology is of greater strategic impor-

tance than manpower, geography, and natural resources.

Late 1990s Documents frequently never exist on paper because they incorpo-
rate information in the form of audio and video pieces.

Media technology is capable of producing computer-generated

personalities, intelligent image systems with some human charac-
teristics.

1999 The several hundred billion dollar computer industry is largely

intelligent by 1990 standards.

2000 Three dimensional chips and smaller component geometries con-

tribute to a multi-thousandfold improvement in computer power
(compared to a decade earlier).

Chips with over a billion components appear.

The world chess champion is a computer.

2020-2070 A computer passes the Turing test: which indicates human-level

intelligence.

COMPUTERS IN 2001

Since it takes roughly a decade for a technology to move from the laboratory to common
use, every technology that will be employed in products by the year 2001 should be
operating in a research laboratory in 1990. Extrapolating from today’s research-stage
technology, this section of the chapter examines some of the developments that can be
expected by the turn of the century.

2. A person in communication with such a device could not tell whether he or she was communicat-
ing with a machine or with another person.
 Computers in 2001 347

It is unlikely that any circuit technology such as optical switching will replace
semiconductors, despite AT&T's impressive 1990 demonstrations that photons can
perform logic operations.Although announced as a phenomenon,it is not yet operating
as a system component in a research laboratory.Memory is critical, too, followedby the
mastery of microfabricationofoptoelectronicsystems.Thebest useof opticalcomputing
would be in the design of a switching system in order to effectively utilize fiber-optic
links. This will occur perhaps a decade before an optical computer is developed.
Molecular computing is still in a preresearch proposal phase.
The mixing of electronics and biological processing to form the "biochip" has
begun. Thesechps can be used as sensorsand effectors.Thus, microelectronicswill play
an ever-increasing role as a bioengineer's component.
Semiconductor densities will continue to increase on their current trajectory.
Various estimates predict semiconductorchip densities of over 400 million transistors,
giving at least 100 million bits per chip (the Intel model) by the year 2000. Since the
1-kilobitchipof 1972,memoryhas quadrupledeverythree years-aphenomenon known
as Moore's law (1975version).If growth continuesat this rate, the use of 4-megabit chips
in 1990 would extrapolate to the use of 1-gigabit chips in 2002. Similarly, because a
microprocessor with a cache requires about 4 million transistors, the largest chips will
contain two to four microprocessorsby 1995!Thus, the era of parallelism will be forced
by technology.
The gains in magnetics and electro-opticstorage are almost as impressive.In 1989,
the most cost-effective disk was the 5.25-inchdisk, which held 1.6 gigabytes and cost
roughly $1,500. Assuming disk densities of all types continue to double every three
years (an increase of 26 percent per year),such a drive would store over 25 gigabytesby
2001. Semiconductoradvocateswho continue to predict the demise of magnetics, to be
replaced by semiconductors, will not see much replacement, except for notebook- and
pocket-size computers. Magnetic disks offer an advantageby providing permanency.
The vast amount of memory available to a user for transport on his or her person, at
home, and in central and regional facilities is staggering.
If Seymour Cray continues to design computers, his clock rates may reach 4
gigahertz. This projection is based on a 1988 objective of the Cray 4 to achieve a
1-gigahertzclock in 1992using galliumarsenide semiconductors.These high clock rates
are leading to some really impressive computational capabllities. The Cray 4 was
announced as a multiple,vectorprocessor computerwith sixty-fourprocessorsand was
projected to reach 128gigaflops. In 1990,NEC's SX-X44 supercomputerprovided over
22 gigaflops when utilizing all four processors at a clock rate of over 500 megahertz.
At a more mundane level, the performance gains of "plain old processors"
implemented as one chip are going to continue to be spectacular.Thanks to the simple,
348 The Future

pipelined RISC architecture, a processor is as easy to build as a ”test” chip for the
semiconductor process. In the decade starting with 1985,performance has grown at the
rate of 60 percent each year. Some of this performance growth has been due to
architectural and design improvements, and some has been due to clock-speed im-
provements. Because the clock in a chip only needs to be propagated over a very small
distance, clock-speed improvements have occurred at a rate nearly twice that of the
larger Cray machines (26percent a year versus 14percent a year).The industry has come
from a clock of 200 kilohertz in 1971to 33 megahertz in 1989,and a speed improvement
of 26 percent per year would mean that the 33-megahertzmicroprocessorwould reach
a speed of over 500 megahertz in 2001. However, as these microprocessors increase in
speed to 50 megahertz,multichippackaging is required, so the rate of increasemay slow
slightly due to the need for increased clock lead lengths.
Like microprocessor clocks, local area networks have also improved in speed at a
rate of 26 percent per year, a factor of 10 per decade. Ethernet operated at 10 megabits
per second in 1980. FDDI operates at 100 megabits per second in 1990, and most
certainly, 1-gigabit fiber networks demonstrated in 1990 will operate as networks in
2001.
Computer backplane buses have also improved at a rate of 26 percent per year. In
1970, Digital’s 2-megabyte-per-second Unibus was the industry standard. In 1980,
Motorola’s VME bus operating at 20 megabytes per second was the standard. In 1990,
various flavors of the Futurebus standards operate at 200 megabytes per second.

COMPUTER ARCHITECTURE6)

By extrapolation, evolution will carry us into the twenty-first century doing things
pretty much the same way we were doing them in 1990.This would mean that for high-
performance technical computing, we’ll be using faster, multiprocessor and
multicomputer systems based on the computers we’re just beginning to see from the
plethora of start-ups aimed at developing high-performance computers based on
parallelism. Although it is nontrivial to exploit the parallelism inherent in these
structures, much progress is being made. The languages will have to change, but only
after people start understanding the simple, parallel machines of the early 1990s that
utilize only a modest number of processors.Just a few languages transparently support
parallelism, and it’s hard to imagine that users will switch very rapidly, given their
conservatism and their interest in running dusty decks of Fortran (and Cobol).How-
ever, they must change in order to get the most benefit from parallel computers.
The Cray YMP introduced in 1988operates at near its peak speed of 2.5 gigaflops
on the Linpackbenchmarkwhen using alleight processors.Although Alliantparallelized
programs first with its eight-processor FX 8 in 1985,the Cray YMP benchmark heralds
 Computers in 2001 349

the acceptanceof parallelismin scientificcomputingand the beginningof a new, parallel

era.
NEC’s SX-X44 four-processor supercomputer provides a workload capacity of
about three times the Cray YMP processor. Furthermore,the peak speed, using just one
processor, is about 1.5times lugher than the Cray YMP using all eight processors!
By the turn of the century, parallel programming will become routine, either by
evolving 1990 computers to be more transparent and easier to use, or by using explicit
parallelism.Data-flow languageconstructscouldbe used to provide implicitparallelism
if languages and training evolved more rapidly. In no case can compilers recognizethat
a problem may be inherently parallel. Users who program must be retrained in order
to exploit parallelism, and computer science has yet to modernize its curriculum to
include supercomputers. Otherwise,the ”Do what I mean” (DWIM)paradigm would
have to be ”Do what is right” (DWIR),and that would be a superhuman computer.

MASSIVE PARALLELISM

The high-performance,general-purposecomputers will have a high degree of parallel-

ism, with each computer consisting of hundreds of processors in a multiprocessor with
a shared memory. Several thousand computers that communicate via a fast switching
networkwillallowmulticomputers.In addition,computersthat have a singleinstruction
stream to process data in a massive number of processing elements can be built,
patterned onThinlungMachines,Inc.’sConnectionMachineidea.Thesecomputerswill
most likely provide a teraop, or 1 million-million operations per second, of
computing power. DARPA (DefenseAdvanced Research ProjectsAgency)has funded
two such efforts, by Thinking Machines and Intel, to achieve high speed. Researchers
and funding agenciesbelieve a computer capableof 1teraop is feasibleby the year 1995.
It is unclear how much parallelism personal computers will have. This is because
a single chip might have 100 megabytes of primary memory. Therefore, our whole
personal database and all the programs we use today may be carried in our wallet, or
in a pencil-likedevice. The interface to humans and to other systems is a crucial factor
if the computer is to be widely used by a billion people.

THE VON NEUMANN ARCHITECTURE MODEL

Three alternative computer structures-single instruction massive data, multiproces-

sors, and multicomputers-are all simple extensions of the basic von Neumann archi-
tecture model of instruction fetch and instruction execute.
350 TheFuture

Neural-network computing is definitely not encompassed in von Neumann's

model. Already, neural networks are being employed in a wide array of pattem-
recognition applications, from speech synthesis to image recognition. Neural-network
computing elements are likely to be used in conjunction with basic, general-purpose
computers for such tasks as signature or handwriting analysis.
Systolic processin$ maybe used for signal- and image-processing functions at the
computer's interface, although it is unlikely that special-purposeprocessors can com-
pete with the rapid evolution of the RISC microprocessor.

OPERATING SYSTEMS AND THE HUMANINTERFACE

Everyone hopes to see the advent of operating systems that are substantially more
robust and easier to use than systems of the 1980s. Communication via voice and
handwriting will most certainly become possible. The Apple Macintosh, which intro-
ducedXeroxPARC'smethod ofcontrolbymeansof iconsinsteadof tediouscommands,
still has a long way to go in terms of ease of use. Many organizations are working on
better graphical user interfaces, including all the workstation companies, the Open
Software Foundation, and desktopmanager firms such as Visix. The NeXT desktop is
perhaps the simplest and most complete with multimedia support. The Microsoft
Windows interface for the PC is likely to evolve to reach parity with the Macintosh in
terms of ease of use.Nevertheless, there is stillroom for start-upsto continuedeveloping
products that improve the human interface.

PERIPHERALS

Basedonextrapolationfrompastprogress,betterinput/outputofall typeswillcertainly
exist. In particular, better human interaction is critical. The quality of screen images will
risedramaticallyand at a lower cost when highdefinition television arrives. At the same
time, lowcost, small liquid-crystal screens of all types are appearing just as the basic
RCA patent on liquid-crystal displays expires. Higher resolution will also mean the
production of realistic, three-dimensional graphics and images.
Hands and arms, legs, and much better eyes will become the most important new
peripherals. These transducers will permit the construction of a whole new class of
autonomousrobots that can carry out simple tasks in the home, office, laboratory, and
factory environments. The most useful, practical, and mobile robots are unlikely to be
built by the year 2001, however, since such robots would need to "see" and "under-
stand" the environments in which they operate.

3. The "pumping"of data through an array of processing elements.

 Computers in 2001 351

PCs of nearly all sizes, from pocket to desktop, will have to include a fax and voice
input/output. Chips from Intel (DVI) and C3 for compressing video by a factor of 10-
1,000 will ensure video integration with most computers by 2001.

HANDWRITTEN TEXT AND SPEECH

By the turn of the century, computersshould be able to read text and communicatewith
us by voice, but unless more progress is made in speech understanding and computers
come to have the ability to learn, they will still "feel" much like 1990 computers. HAL,
in the movie 2002: A Spuce Odyssey, may not be a bad model for how a computer might
communicate,but a device with all of HAL'S capabilities seems several decades away.
The most useful device I could envision would be a personal assistantnotepad and
database that would not only do everything today's PC does but also accept commu-
nication via voice, handwriting, and perhaps keyboard. Later on in the twenty-first
century, such a device could listen and translate. It would plug into a conventional
phone or be a cellularphone. Thiswould representthe ultimateevolutionof the plethora
of virtually useless devices for the wallet, wrist, and pocket sold for $10 to several
hundred dollars through specialty stores and airline magazines.
Computers would provide a more extensive public interface without the humans
who currently operate systems in airlines, banks, insurance offices,and stores. Today,
we stand in line to talk to these interpreters. Many of us would rather communicate
directly with the computer system, when it can be made easy and powerful enough to
use.

THE UNBOUNDED NETWORK

Cellular radio networks will make fully distributed computing available for use in any
location without the need to plug in to a network. Such a network would allow any
computer to call any other on a totally space-independent basis for computer and fax
messaging,data access,and performingvarious tasks such as reserving cars or ordering
merchandise.
The U.S. government's High Performance ComputingInitiative aimsat creating a
network that will exploit the bandwidth inherent in fiber optics by the end of the decade
to enable the research establishment in academe, government, and industry to com-
municate. Such an evolution would allow information to be transferred between
machines at gigabit rates.
Fax traffic now constitutes about one-tenth of all telephone traffic and is growing
rapidly. If specializeddigitalfax network providers came into existence,then computers
might utilize such networks.
352 The Future

The IntegratedServicesDigitalNetwork (ISDN)system,which was supposed to be

available on a worldwide basis by 1985, might come into being to function as a “data
highway” that would be available everywhere, including homes. The normal ISDN
service (basicrate) provides mere 64-kilobit-per-secondchannels. Although this speed
is adequate for documents of the 1980s, it is not sufficient to transmit the many color
graphicalobjects required for remote, interactive computing.Primary-rateISDN (1.544
megabits per second T1 carrier in the United States) or higher speeds, such as T3 (45
megabits per second) communication, are needed for the technical marketplace. A T3
network is likely to be the common mechanism for the interconnectionof computers on
a wide area basis by the end of the decade.The SynchronousOpticalNetwork (SONET)
being installed by the world’s telecommunications companies is also a promising
development. By all accounts, Europe and Japan will lead the United States in high
performance wide area digital networks.
The availability of higher data rates and/or improved data-compressionsystems
should result in the advent of home videophones using highly compressed video.
Workstations will be the first to provide built-in computer and videoconferencing
capability, because they can rely on fast local and wide area networks.

THE COST AND SIZE OF COMPUTERS

Computerswill range in cost from a few dollarsfor thosethat can fit in someone’spocket
to a hundred million dollars for those that are central to an enterprise or a laboratory.
Whereas simulationstops at the molecularlevel today, these large machines will be able
to simulate more of the universe, including interactionswithin the atom. Every person
or organizationspendssomeconstant fractionof their budget for computing,just as they
do for food or other necessities.Furthermore,the industry is segmentedinto companies
that supply computers at every price level, from a few dollars to tens of millions.

THE DISAPPEARANCEOF PAPER

Computers with laser printers have enabled every PC user to become a publisher and
continue the exponential increase in the production of printed matter. Paper must
inevitably start to disappear as the archive-storagemedia, simplybecause the quantity
of available information is becoming so vast that computers will be essential for
tracking, finding,analyzing, and perusing it. This means that all information, including
the enormous amount coming in fax form, will ultimately be captured and encoded,
stored, and shipped around electronically,potentially minimizing the use of printed
media. Printing should be reserved only for those occasions when the information
contained in the machine cannot be used directly on-line.However,given the rise in the
 Computers in 2001 353

number of notebook-size computers, it will become feasible to start making nearly all
information,such as books and periodicals, available in computer-readableform.
On the other hand, the amount of paper will rise rapidly as the ease of computer
printing and fax proliferates among everyone, including schoolchildren.There will be
a continued increase in the use of paper as the primary medium for applications
requiring portability, such as entertainment (books and newspapers), information
distribution(advertisements),and even the disseminationof information(memoranda)
within organizations.
Electronic mail, though widely used within corporations, has failed to limit and
substitute for fax because of the computing and communications industry’s failure to
create the necessary standards for interoperability.However,by 2001, one would hope
for a common dial-up network for electronicmail that’s as easy to use and accessibleas
fax.
With very large memories, and the ability to view and peruse information elec-
tronically, an environment could be envisioned in which there is much less paper and
information is normally stored and viewed electronicallybecause the image quality is
higher than with paper. This scenario could become a reality by the early twenty-first
century.

ARTIFICIAL INTELLIGENCE

In the past, artificial intelligence has been defined as a set of techniques not served by
the mainstream of computer science. AI includes efforts to address very difficult
problems such as understanding images, vision, natural language, and speech. It also
includes the study of robotics and how robots can function autonomously. As an area
of computing (e.g., expert systems) matures and becomes widely understood by the
academiccommunity,it is a s s d a t e d into the mainstreamof computer science.During
the late 1980s,expert systemsusing knowledgebases matured for practical, though not
common, use.

EXPERT SYSTEMS

Many types of expert systemshave been built for giving adviceand for doing diagnosis
and design, as described in Chapter 9. Writing programs to solve problems for a
particular applicationdomain, such as mechanicalcomputer-aided design (MCAD),is
not per se a new development, because much of the programming industry has been
performing this function since computers were invented. Before the advent of “expert
systems,” however, programs were written in standard procedural programming
languages, such as Cobol and Fortran. The ”official” expert-systems programs are
354 The Future

written using a rule-based approach, either by extending the LISP language with an
inference mechanism shell (e.g.,KEE or ART) or by using a new and unique language
for rule-based programs (e.g., Prolog, OPS, or NEXPERT).
Unlike commercial and technical applications, which evolved rapidly because
Cobol and Fortran standards were established by 1960, rule-based expert systems are
still in the ”sandbox” stage. Dozens of proprietary and unique languages exist for
writing these programs. Rule-based programming is unlikely to grow very rapidly
because the lack of standard languages will slow the industry’s maturation. Thus,
potential users cannot be trained easily by computer-engineeringor computer-science
departments.Furthermore,programs written on one system cannotbe used on another
system, and large systems cannot be built up from other systems.
If rule-based expert systemsare to become a mainstay programming technique, all
the major languages-including C, Cobol, and Fortran-must have extensions that
includerule-based,inferenceprogramming.By the year 2001, the proprietary languages
will become completely extinct if the industry is to grow and mature.

GREAT NEW APPLICATIONS

Although it is difficult to predict how the vast increase in processing power will affect
science and engineering generally, its impact on the following specific areas is clear.

Animation

Large-scalecomputers,with their ability to compute realisticscenes, offer an alternative

to traditional filmmaking techniques. The evolution of multimedia will permit the
computer generation of cartoonlike actors that can be used within desktop publishing.
Only in the twenty-first century will it become possible to compute real actors and real
scenes.

Commercial and Transaction Processing

Although the preceding examples are from the technical marketplace, given a vast
increase in transaction power, a revolution of equal proportions appears to be possible
in the area of commercialand transactionprocessing.l k s canbe accomplishedthrough
higher degreesof parallelism, obtainedby using small disks, multiple fast microproces-
sors, and large memories.By2001, a computer that never failsand deliversover a teraop
is feasible.It will most likelybe the center of enterprisecomputingand largetransaction-
processing networks used for banking and air travel.
 Computers in 2001 355

Computational Chemistry, Including Biochemistry and Materials Science

Molecular modeling and computational chemistry have made possible the interactive
design of molecules, using large-scalecomputers.

Computational Fluid Dynamics (CFD)

CFD, the basis of aerospace engineering, is also useful in designing buildings and
automobiles. Horst Simon, of NASA's Ames Research Center, estimated that carrylng
out a simulation of a vehicle requires about 1015operations (5million grid points, 50,000
iterations,and 5,000 operations per point per iteration),using about 200 million 8-byte
words of primary memory and about 10 gigabytes of disk storage. Table 13-5 shows
several design activities, together with the turnaround time and computer speed
required to perform each one.

Image Processing

Various disciplines, ranging from radiology and surveillanceto weather forecasting,

rely on the interpretation of hgh-resolution photographs and other image and signal
sources.High-performance computers are finallymaking the use of digitalimages and
image processing feasible, and the availability of satellite image data is transforming
everything from military intelligence to urban geography.

Mathematics as a Generic Capability

The late 1980ssaw the emergence of severalprograms that will have a long-term effect
on the way computers are used. Programs such as MathCAD, Mathematica, and
Matlab can deal with much of the college-level mathematics used in science and
engineering.Theseprogramsarealsothebestway totrainallstudentsaboutmathematics,
beginning with algebra in the upper elementary school. Fundamental mathematical
competence is far more important than the ability to do spreadsheet programming.
Malung various forms of mathematicsunderstandableto and usable by a much larger
fraction of society could cause more change than spreadsheets, word processing, and
electronic mail.

Mechanical Engineering

Over the last decade, a revolution occurred in semiconductors and digital-system

design, with the advent of the ability to design and model large digital systems
356 TheFuture

Table 13-5.TurnaroundTime and Computer-Speed Requirementsfor Selected Design

Activities.

Turnaround Time Computer Speed

Design Activity (in Hours) (in Gigaflops)
Proof of concept 1000.00-100.00 0.3-3.0
Design 10.00-1.00 30.0-300.0
Automated design 0.10-0.01 3,000.0-30,000.0

accurately.Furthermore, fabrication information can be sent electronically to factories

for the direct production and testing of chips and physical interconnections.Thus, no
physical prototyping is required, which at least halves the time between design and
production.
Computers have been used to design all sorts of mechanical structures, from
automobilesto spacecraft,and to performa wide range of activities, fromdrafting to the
analysis of designs, including crash simulation for cars. A designer can render high-
quality images and show the objects in motion with the help of video. A vast increase
in available computer power should provide mechanical engineers with sufficient
computing capability to revolutionize mechanical design.
Under this design paradigm, it is possible to perform everyfacet of product design,
including the design of the factory in which the product will be produced, without a
prototype.Sucha developmentcould transformmechanicalengineeringand industries
relying on the manufacture of physical parts. For starters, product quality would
improve.Thebigimpact would comefroma drasticreductionin product-gestationtime
and in the size of the organization,which could be decreased by a factor of 5. Reducing
the size of design teams has been shown to increase both product eleganceand quality.
In a totally computerized design and manufacturing environment, a team of no more
than fifty could design a perfect car or other vehicle from the ground up, in under two
years.

Personal Computers in 2001

The largest computerswill continue to be used to explorethe forefront of applications-

applicationsthat willlaterbecomefeasiblewithaPC. Ardent'sgraphicssupercomputer,
introduced in 1988 at a price of $100,000, is an excellent model for the home computer
of the year 2001 at a price of $2,000. Three-dimensional phenomena, from molecules to
 Computers in 2001 357

galaxies, will be simulated at high enough speeds to transform modem science from
experimental to computational-simulation-based.
This paradigm shift will transform every facet of science, engineering, and math-
ematics, starting with the fundamentalnature of education. Every home will have an
unlimited laboratory in which to conduct experiments.As a result, children will learn
in new ways, such as by using the Writing to Read program introduced on the PC,
simulating all kinds of scientific experiments, and doing programming to develop
logical thought. An educational revolution rivaling the replacement of the slate by the
notebook is possible.

Robots for the Home

Given the eleven-year rule, useful robots for the home will probably not be available in
2001.Industrial and research robots are quite dumb, hard to program, and of limited use.
Without significantadvancesin vision, planning, and understandingcommon sense,as
well as some general ability to learn, it is hard to believe that robots will be very mobile
(and hence very useful) around the home in the near-term future. Vision processing to
rival human capabilitiesis estimatedto require on the order of 10l2operationsper second
(one teraop) of computing power.

General Technical and Scientific Applications

With the vast increase in computing power reaching over a teraop, it will be possible to
simulateverycomplexsystemsfor suchdisciplinesasatmosphericscienceandchemistry.
This highly advanced simulationability is needed in order to understand phenomena
such as acid rain and the erosion of the ozone layer.

Visualization and Virtual Reality

Visualization is simply the ability to present a graphical view of a problem in order to

enable a human user to gain insight into possible solutions.High-performancework-
stations provide the ability to generate photo-realistic models of all physical systems.
Virtual-reality environments take the form of a head-mounted, three-dimensional
display that changes in response to the movement of the user’s head, together with
gloves or joysticks, all of which allow the user to position himself or herself within a
three-dimensionalspace. Thus, the user can navigate through any three-dimensional
area (in effect, ”be anywhere”) and look at anything at any scale, ranging from a
358 The Future

molecule to a galaxy. Laboratories and start-ups have formed to provide interactive,

virtual-realityenvironments such as these.

INDUSTRY RESTRUCTURING
Given the increase in microcomputer power and the reduction in price, as shown in
Figure 13-2, an incredible performance-per-unit-price difference among the various
computer classes must cause a major restructuring of the computer industry. Such a
discontinuity alone could trigger a major recession, and the shift away from the
minicomputer is most certainly behind the 1990 changes in the mini-based New
England economy.Furthermore,portable operating systemsand applications,both for
UNIX and for MS-DOS and OS/2, have transformed hardware systemsinto high-tech,
commodity products, with many suppliers providing essentially the same product.
Althoughthe price of all the practicalcomputersused today will be reduced to near-
zero in 2001, parallel computers with more capability will come into being to maintain
constant prices. In terms of quantity, however, the greatest growth will be in simple
wallet, pocket, briefcase, and desktop computing, provided such devices can be made
easy to use. All of these computer types are built with the same manufacturing
techniques used in consumer electronics.
Figure 13-4 (by the Gartner Group)illustrates this growth scenarioby showing the
evolutionin the number of computers sold in the various classes.Note that mainframes
are projected to decline from their current population of fifty thousand. Workstations
are the fastest-growing segment in the early 1990sand approach the installed base of all
the shared minicomputer and microprocessor-based systems, including all local area
network servers. The Gartner Group projects the emergence of powerful pocket-size
personal computers that will nearly equal the installed base of the PC by 1999.
The message for start-ups is clear: very small, low-cost, new fifth-generation
products present opportunities for new ventures, provided they concentrate on quality
manufacturing, unique and important applications,and innovative marketing.
For the minicomputer and mainframe classes, the need to solve much larger
problems involving more data will result in the emergence of more radical and
specialized parallel computing for use in specialized local area network servers.
Network computing is finally coming into being. The VAX computing environment,
conceivedin 1979, was formed as a hierarchy of three stylesof computing,starting with
mainframes operated as a central service, connected to distributed departmental
minicomputersoperatedby specific groups, and terminating with personal computers
and workstations operated by individuals. A modern follow-on to this structure is
predicated on theevolutionfroma collectionof centralizedcomputersto fully distributed,
networked personal workstations interconnected via high-speed local area networks,
as shown in Figure 13-5.
 Industry Restructuring 359

1OOOOOO2

- PCS

B
I
I
0
Y
v)

.-E
L
v)
W
Y
a
!
8
c
0

a"C
Q
v)
a
0
5
L
0
L
W
a

zE
1983 1985 1987 1989 1991 1993 1995 1997 1999
Figure 13-4. Estimate of the Number of Computers Installed in the Various Classes
Versus Time. (Courtesy of the Gartner Group. Reprinted with permission)

CHANGES IN THE MAINFRAME AND MINICOMPUTER INDUSTRY

It is hard to believe that mainframes and large minis will benefit from the need for
distributed computing,because of the considerabledisparity in both performance and
price/ performancebetween mainframesand microprocessor-basedcomputers,as was
shown in the Intel scenario. Mainframes and superminicomputers have been slow to
evolve, since they use expensive packaging for ECL. In contrast,complementarymetal
oxidesemiconductor(CMOS)technologyruns faster,because all the componentsare on
360 The Future

Departmental time-sharing (1965-1985) Personal computers with

terminal emulation (1978-1985)
X.25. Tenninal
SNA emulators

ROe Files Rint . .. Communications.

nets. terminals
Network
intercamen
n
Personal mnputer
(wilh terminal

I Operating system, languages I l l i

Job ... Job

0 . .

X.25. Terminal

a SNA emulators r,

-
Realtime

canmunications
server server server gateway intercormect
server
I

Figure 13-5. Evolution of Computing, from a Centralized Computer to Fully Distributed

Personal Computers andworkstations.

a single chip. Mainframe and minicomputer complex instruction set computer (CISC)
architectures are more complicated and require arcane implementations that have
historically had a gestation time of more than five years. Getting all the fast circuitry on
one c h p results in hardware that goes faster, runs cooler, is more reliable, and costs next
 Industry Restructuring 361

to nothing. Differencesin disk technology-from the large, expensive disks used in disk
farms to the 5.25-inchformat-also exacerbatethe shift.Isn’t it clear that expensive and
slow has to lose to cheap and fast, especiallyif the development times and costs are also
substantially less?
Even the large, die-hard users with lots of code locked up in “code museums” are
beginning to understand that they must switch to an open-systems form of computing
based on standards in order to get a better computingenvironmentand at a substantially
lower price.
The following CMOS micro-based alternativesall offer performance that equals or
beats the old line of expensive mainframes and superminicomputers at negligible
prices, thereby ”niching” the existing hardware:

PCs and multiprocessor PC servers from Compaq, etc.

Workstations and servers from Sun, etc.

Uniprocessor micros from Altos Computer Systems, MIPS, Motorola, NCR,

Pyramid, etc.

New network servers from Netframe, etc., using the Intel chips

Multiprocessor servers and minis using multiple microprocessors from Arix,

Digital, Encore, Sequent, Silicon Graphics, and Solborne

Specialized technical computers from Intel and NCUBE, as well as a plethora of

Transputer-based companies

Front-end and back-end database servers from Teradata, etc.

For example,the Stardent 3000 is a quad vector processor multiprocessor delivered

in November 1989 that sells for a tenth the price of a VAX 9000 yet almost equals it in
performance. The DEC 9000, promoted as DEC‘s first mainframe, was delivered
roughly one year after the Stardent 3000. The difference in the product-gestationtime
was a factor of 3 (twoyears versus six years to design each product). Digital claimed that
it spent almost a billion dollars developing the 9000 (a factor of nearly 50 times what
Stardent spent, excluding the microprocessor). Probably a thousand people were
involved in developing the VAX 9000, whereas less than fifty people developed the
Stardent3000. Is it any wonder why America is losingits competitivenessand why start-
ups must form?
It seems logical to anticipate a massive and continued shift in the industry‘s
structure during the next decade, not growth for traditional minicomputer and main-
frame products. Both of these markets have remained flat at about $15billion per year.
362 The Future

Of the original BUNCH, Burroughs and Univac combined to make UNISYS, which is
distributing UNIX products; NCR has switched to UNIX and the PC using Intel
microprocessors; Control Data Corporation (CDC) has downsized its mainframe
business and become a distributor of MIPS products; and Honeywell sold its business
to Bull, a builder of UNIX productsusing the MIPS architectureand products. The lower
revenuesand reduced profits (orincreased losses)at Alliant,DEC, DG, Encore,Hewlett-
Packard, MassComp (now part of Concurrent),Prime, and Wang in the last five years,
resultingin company downsizing,are not solely the fault of managementbut rather are
the result of a fundamentaltechnology shift. It is unlikely that these firms understand
or know how to address the shift. IBM appears to understand and adapt to a changing
market. On the other hand, the mainframe code museums hold most of the world
corporate data as hostage.
Users are beginning to realize that proprietary architectures utilizing the wrong
technologies mean higher cost and lower performance. Furthermore, key software
suppliers have switched to writing applications for standards-based platforms.
This industry shift, whch has been apparent since the mid-l980s, could, by itself,
cause a major recession in the 1990s, as observed earlier. At the very least, the shift will
be as dramaticas the movementof the textile mills fromNew England when new fabrics
entered the market. Like the mills, New England's computer companies have begun a
steady decline, which tracks the declining market for minicomputers.Kendall Square
Research, a Boston-area firm building a large multiprocessor using CMOS technology,
characterizes the inevitable decline of these companies in Figure 13-6.

CHANGES IN THE U.S.SUPERCOMPUTERINDUSTRY

When it was introduced in 1975, the first successfulvector supercomputer, the Cray 1,
provided power for general-purpose computation by having the highest scalar per-
formance, together with a fast vector processor to provide an additional tenfold
performance increase for problems involving matrix operations. In the 1990s, RISC-
based workstationscomputingat near-supercomputerspeed are more cost-effectiveby
a factor of 10 to 20 for mostly scalar programs. Workstations such as IBMs R6000
perform adequately on "average" programs, giving an overall performance difference
of only a factor of 3 or 4 between the workstation and the supercomputer. At the other
extreme,massivelydata parallelcomputersfromIntel,NCUBE,andThinkingMachines
offer higher peak speedsthan the supercomputerat one-fifth its price for highly parallel
and tuned, vector processor applications.
The supercomputeris a protected speciesbecause of its use in defense,even though
many of the functionsit performs could be carried out just as well by alternativemeans.
Furthermore,it has become a symbol of competitiveness,sincethe Japanesehave begun
 Industry Restructuring 363

Figure 13-6. Fundamental Restructuring of the Computer Industry Brought About by

Standards and Commodity Hardware. (Adapted from a figure provided by
Kendall Square Research Group.)

making supercomputerswith greater peak speeds than Cray Research or Cray Com-
puter. A number of companies that tried to build supercomputers-including CDC's
ETA, Denelcor, and Chopp-have withdrawn from this very small and overcrowded
market. SupercomputingSystems,Incorporated,another Cray Research spin-off firm,
is based on the traditional,vector multiprocessor supercomputerarchitecture.DARPA
is funding Tera Computer Corporation to design a large-scale supercomputer.IBM is
enhancingthe scientificcomputing capability of its mainframeline.One, or at most two,
of thesemulti-hundred-million-dollardevelopmente f f o r t s d succeedandbe profitable.
The amount of technical computing done with supercomputersper se is not very
large, for many reasons, including high initial cost, complexity,training requirements,
difficulty of use, responsiveness, and lack of visualization.It simply costs more to run
most jobs on a supercomputerthan on fast workstations or even personal computers.
Minisupercomputers are a lower-cost alternative that has taken some fraction of the
364 The Future

supercomputer’spotential users. Thus, the U.S. supercomputerindustry is being held

to a billion-dollar level by:

Minisupercomputersthat are at least as cost-effectiveand provide high through-

put. Minisupers built from CMOS microprocessors should prove to be the most
cost-effective.

Specialized,massively parallel computers such as the Connection Machine, which

uses thousands of processing elements, and the Intel Supercomputer,built from 64
to 1,024 CMOS microprocessors,which provides exceptionally high performance
at lower cost for vector applications.

Personal computers, ordinary technical workstations, personal supercomputers,

and supenvorkstationswith a few microprocessorsthat perform the same tasks at
one-tenth to one-third the speed, but at five to ten times the cost-effectiveness.

Competitive machines from start-ups.

Japanese supercomputers.

The U.S. supercomputerindustry will declineby the year 2001, because it will have
been “niched‘ at, as described above.Japanesesupercomputerswill supply the largest
fraction of traditional supercomputingcalculations due to the commitmentby Fujitsu,
Hitachi, and NEC. All three are vertically integrated and can develop the critical
circuitry and packaging on which the supercomputeris based. TheJapanesetend to see
thesupercomputeras the drivingforceor leading edgeforbuilding the fastesthardware
in much the sameway asautomobilecompaniessee buildingracing carsasadvertisement.
Furthermore,all of these companiescan sustain any losses that are inherent in a small,
but very-high-technologymarket.Intheunited States,onlyIBMisverticallyintegrated-
but rarely does IBM sustain a loss just for show or to maintain its market presence.

WORLDWIDE COMPETITION

Provided that the U.S. venture capital community doesn’t change significantly to
become much more conservative and cease funding high information technology, the
United States should retain its lead in imentions simply because it has such a fine ma-
chine for training a small number of creative engineers and scientists. Start-ups
everywhere will continue to invent and to bring fundamentally new technology and
products into existence.
 Worldwide Competition 365

Larger US. and foreign companies will still be synergistic with start-ups to adopt
their inventions. The large firms that control over 90 percent of the engineering and
scientific talent will continue to evolve slowly and to reinvent, instead of channeling
creative energy into innovationby making the changes that would be required to do a
given product or technologyina new but evolutionaryway in order to make it dominate
a market.
But the bulk of innovation will come from outside the United States, because the
Japanese,for example, are not as plagued by the ”not invented here” syndrome, which
is endemicamong most Americanand many European engineers,who tend to reinvent
technology and products, often with poorer results than the original design.
With the economicunificationof Europe in 1992,a broader, easier-to-accessmarket
may emerge. On the other hand, European suppliers might become more competitive
in a world market. At the very least, Europe may take a more aggressive role as a global
funder of start-ups. In the 1980s, Europe became effective at managing research and
advanced development across national borders and between industry and academe.
However,given the laws supportinglifetimeemploymentin Europe, it is difficultto see
how start-ups will form very easily there. Thus, the larger companieswill most likely
continue to count on slow evolution and invention coming from research. These ad-
vances will probably be small, as in the past. The only way large European firms will
acquire big inventions will be to buy products from small companies or buy the
companies themselves.
China and India-which have the world’s largest supply of highly trained talent
with mathematical, engineering, and scientific slulls-are beginning to develop soft-
ware for worldwide consumption. Since the development of much software requires
minimal capital investment (often, only a PC), any country can become a significant
software supplier because of the inherently ”low barrier to market entry.”
Japan will continue to excel in innovation and to become the dominant supplier in
every market it enters.Japan will take the plethora of hardware and softwareinventions
generated by the rest of the world and, by innovation,improve their quality, function-
ality, and performance, to substantially increase the size of the markets. In 1972, IBM
invented the 8-inch floppy and used it to hold diagnostic programs for its large disk
controllers. By the late 1970s, Shugart Associates started up and began the floppy
industry, whch evolved to the 5.25-inch-diameterfloppy with more start-ups.In 1982,
Apple adopted Sony’shighly innovativeand more durable 3.5-inch floppy, which has
become the standard of interchange since 1985, with evolving increases in density,
because the Sony floppy provided so much more than the first, simple inventionand its
evolution.
Because nearly all software products are developed by a method that is subject to
process control and quality standards, the Japaneseappear to generate softwarethat is
366 The Future

a factor of 10better in terms of quality, as measured in defects per thousands of lines of

code, and at a productivity rate that is a factor of 2 to 3 times higher, as measured in
thousands of lines of code per person per day. This development environment is often
called a software factory, and as with any U.S. factory, no one wants to work in one-
certainly not America's creative software engineers.
Thus, over the next decade, when software products become better defined by
standards (e.g., VHSIC's VHDL language for describing digital systems)and by well-
developed algorithms and paradigms in well-defined domains (e.g., spreadsheets,
electroniccomputer-aideddesign [ECADI forvery-large-scaleintegrationIVLSII design),
Japanwill dominate the softwareproductsmarketjust as it dominatesnearlyall markets
for physical goods. In any case, the software industry will remain only a small fraction
of the entire high-technology industry until all computers cost just a few hundred
dollars.
American engineers, coupled with the American MBAs who manage most US.
organizations,will ensure the continued decline of the information-processing indus-
try, because this deadly duo focuses on the human organization (and especially its
political structure), not on the technology and product. The remaining bastion of
American creativity, software products, may be eroded more rapidly than hardware.
The highly disciplined,process-engineeringnature of Japanesesoftwareengineeringis
antitheticalto theU.S.software-engineeringculture.As is the casewith the steel-making
industry, the old ways are too deeply ingrained in the culture to permit change. Thus,
other countries will study and profit from the considerable body of knowledge about
softwareengineeringaccumulatedand taughtby the U.S. academicestablishment long
beforetheUnitedStatesdoes.Here, asinotheraspectsof engineeringandmanufacturing,
the U.S. must switch its role from teacher to student and colleague.

CONCLUSION

This book has presented many technology and product ideas to stimulate the reader. I
seealmost unlimited possibilitiesfor productsextendingwell beyond the year 2001, just
by extrapolatingfrom the technology currently expected to be available. In this regard,
the Kurzweil time line establishes many wonderful goals.
If any of the new-product development scenariosare to become a reality, however,
it will most likely occur outside of the evolutionaryproduct development process that
is characteristicof establishedcompanies.Thus, entrepreneursand venture capitalmust
continue to exist. Hence, the opportunity for start-ups. If the reader is in a large
organizationand is trying to invent or even innovate,the challengeis to outperform the
start-ups,the Japanese,and the rest of the world-all of whom are trying to build better
products.
 Conclusion 367

The last caveat of this book is especially important if the reader has gotten this far
and is still determined to found a company:

Now that you've studied everything I know about technology, products, and start-ups;
persevered and mastered a great new technology; demonstrated that it can be useful in
a product that people are likely to buy; found a way to validate that there really is a market;
and decided to start a company; the easy part is over-but the fun is just beginning.
BIBLIOGRAPHY

ARTICLES

Bell, C. Gordon. 1984. “The Mini and Micro Industries.” Computer 17, no. 10 (Oct.): 14-30.
Carr, Robert. 1989. ”How to Build Better Programming Teams.” Soft*letteu 6, no. 4
(May 1).
Gomory, Ralph E., and Roland W. Schmitt. 1988. ”Science Products.” Science 240 (May 27):
1131-1 132,1203- 1204.
Grayson, Paul. 1989. ”How to Motivate Programmers.” Soft*letter 6, no. 4. (May 1).
Lampson, Butler. 1988. ”Personal Distributed Computing: The Alto and Ethernet Soft-
ware.” In A History of Personal Workstations, edited by Adele Goldberg, 291 - 344.
Reading, Mass.: Addison-Wesley.
Meindl, James D. 1987. ”Chips for Advanced Computing.” Scientific American 255, no. 10
(Oct.): 78-88.
Mendelson, H. 1987. ”Economies of Scale in Computing: Groschs Law Revisited.”
Communications of the ACM 30, no. 12 (Dec.): 1066-1072.
Rosenstein, James, Albert V. Bruno, William D. Bygrave, and Norman T. Taylor. 1989. ”Do
Venture Capitalists on Boards of Portfolio Companies Add Value Besides Money?”
Working paper for a 1989 study.
Tarter, Jeff, ed. 1989. ”Why Goliath Usually Wins.” Softdettev 6, no. 3 (June 15).

368
 Bibliography 369

REFERENCES FOR THE ENTREPRENEUR'S BOOKSHELF

One or more of the following symbols appear before a number of the entries in this section of the
Bibliography to indicate my recommendations for how these works can most profitably be utilized:

A reference to own, understand, and use.

C2 A reference to own and read.

i A reference to outline, whose expert advice constitutes sound rules by which

to operate.

Augustine, Norman R. 1987. Augustine's Laws. New York: Penguin Books.

Fifty-two tongue-in-cheek "laws" governing the production of high-technology,
expensive, and unreliable military products. For example, "By the year X, only one
airplane can be built because it will absorb the entire GNP." Contains many
unfortunate, but empirically derived, laws and conjectures explaining the military-
industrial complex. Evidence is given to indict the military-industrial establishment
for incompetence. An essential work for any company dealing with the military.

Baty, Gordon B. 1990. Entrepreneurship of the Nineties. Englewood Cliffs, N.J.: Prentice-
Hall.
An excellent start-up handbook to supplement White (1977).

Bell, C. Gordon, J. Craig Mudge, and John E. McNamara. 1978. Computer Engineering.
Bedford, Mass.: Digital Press.

Brooks, Frederick P. 1975. The Mythical Man-Month. Reading, Mass.: Addison-Wesley.

A classic, useful book on programming that's also enjoyable reading. Essential if the
start-up's technology is embodied in programs.

Burgelman, Robert A., and Modesto A. Maidique. 1988. Strategic Management of Technology
and Innovation. Homewood, Ill.: Irwin.
See also Roberts (1987).

R Card, David N., and Robert L. Glass. 1990. Measuring Software Design Quality.
Englewood Cliffs, N.J.: Prentice-Hall.

Cooper, Robert G. 1986. Winning at New Products. Reading, Mass.: Addison-Wesley.

Aimed at large companies. Contains some good advice and techniques for looking
at products that a start-up can also use for product positioning.
370 Bibliography

+ Davidow, William. 1986. High Technology Marketing: An Insider’s V i m . New York Free
Press.
A fine book of stories. I recommend spending about two hours to outline the
material and get the author’s advice. The sixteen rules (i.e., questions, just like the
Bell-Mason Diagnostic) presented in Chapter 11 (”Do You Have Marketing?”) are
worth understanding and following.

Q+ Davis, Robert T., and F. Gordon Smith. 1984.Marketing in Emerging Companies.

Reading, Mass.: Addison-Wesley.
Contains good insights and much good advice about marketing and selling.

Deal, Terrence E., and Allan A. Kennedy. 1982. Corporate Cultures: The Rites and Rituals of
Corporate Life. Reading, Mass.: Addison-Wesley.

DeMarco, Tom. 1982. Controlling Software Projects: Management, Measurement and Estima-
tion. Englewood Cliffs, N.J.: Yourdon Press, a Division of Prentice-Hall.

DePree, Max. 1989. Leadership Is an Art. New York: Doubleday.

An excellent book describing the culture of Herman Miller, Inc.

Drucker, Peter F. 1985. Innovation and Entrepreneurship. New York: Harper & Row.
A work that should be read rapidly and outlined if time permits.

a+Fairley, Richard. 1985. Software Engineering Concepts. New York: McGraw-Hill.

Presents concepts that should be understood if the start-up is engaged in software
engineering.

Gershman, Michael. 1990. Getting It Right the Second Time. Reading, Mass.: Addison-
Wesley.
Covers marketing dos and don’ts.

Gilder, George. 1989. Microcosm: The Quantum Revolution in Economics and Technology.
New York Simon & Schuster.

Gladstone, David J. 1988. Venture Capital Handbook. New York: Prentice-Hall.

An essential book on raising capital, with a good discussion of the business plan.

Goldberg, Adele, ed. 1988. A History of Personal Workstations. Proceedings of the History
of Personal Workstations Conference (Jan. 1986).Reading, Mass.: Addison-Wesley.

*Grove, Andrew S. 1983. High Output Management. New York: Random House.
An excellent book on how to manage and how to increase management productiv-
ity. It would certainly be great if everyone read and in some way practiced this kind
of management.
 Bibliography 371

Q+ Grove, Andy. 1987. One-on-One with Andy Grove. New York Putnam.
Presents questions and answers about management.

R+ Humphrey, Watts S. 1989. Managing the Software Process. Reading, Mass.: Addison-
Wesley.
Essential for software-engineering management.

R JIAN. 1988. BizPlanBuilder. Los Altos, Calif.: JIAN Co.

A template that can be used on a PC or Macintosh to write a business plan by
revising and responding to material contained in the template, including spread-
sheets.

C2 Juliussen, Karen, and Egil Juliussen. 1990. The Computer Industry Almanac, 1990. New
York: Simon & Schuster.
Contains very useful information about companies, organizations, markets, people,
and products.

R+ Kawasaki, Guy. 1989. The Macintosh Way: The Art of Guerrilla Management. Glenview,
Ill.: Scott, Foresman.
Presents many critical rules for marketing products.

Kotler, Philip. 1986. Principles of Marketing. 3d ed. Englewood Cliffs, N.J.: Prentice-Hall.
A traditional marketing textbook. Shows why MBAs can be replaced by a series of
computer programs.

Kurzweil, Ray. 1990. The Age of Intelligent Machines. Cambridge, Mass.: The MIT Press.

R+ Levitt, Theodore.1986.The Marketing Imagination. New York: Free Press.

The book on marketing. Reprints the classic "Marketing Myopia" from Harvard
Business Review (July- Aug. 1960).

*+ McKenna, Regis. 1985. The Regis Touch. Reading, Mass.: Addison-Wesley.

A great guide to all aspects of high-tech marketing.

+ McKenna, Regis. 1989. Who's Afraid of Big Blue? Reading, Mass.: Addison-Wesley

I recommend spending an hour to read and outline its two pages of advice.

!2 Nesheim, John L. 1988. Startup: Founding a High Tech Company and Securing Multi-Round
Financing. Saratoga, Calif.: Electronic Trend Publications.

Contains many details about what to do, along with numerous, clearly marked
rules. This book is expensive, however, and most start-ups are unlikely to spend the
several hundred dollars it costs.
372 Bibliography

Osborne, Adam, and John Dvorak. 1984.Hypergrowth: The Rise and Fall of Osborne Com-
puter Corporation. Berkeley, Calif.: Idthekkethan Publishing Co.
Recommended reading to see what can go wrong in stage IVb, as a product takes
off. Clearly illustrates the flaw of introducing a new product that can’t yet be
shipped while the company is still selling a product whose revenue is vital.

Peters, Tom J., and Robert H. Waterman. 1982. In Search of Excellence. New York: Harper
& Row.
Presents a good discussion of corporate cultures in large organizations based on a
survey of successful companies. Some ideas may be useful to a start-up.

Rifkin, Glenn, and George Harrar. 1988. The Ultimate Entrepreneur. Chicago, Ill.: Contem-
porary Books.
The story of Digital Equipment Corporation, a great role model for CEOs and for
establishing corporate culture. Ken Olsen founded DEC in 1957 and led it to become
the world’s second-largest computer company, staying in charge longer than any
other CEO.

Roberts, Edward B. 1987. Generating Technological Innovation. Oxford, England: Oxford

University Press.
Contains many case studies. Any large-company bureaucrat managing research and
development should understand its contents.

Rogers, Everett M., and Judith K. Larsen. 1984. Silicon Valley Fever: Growfhof High Technol-
ogy Culture. New York Basic Books.
Helps in understanding the culture of employees, customers, and investors-if the
start-up is doing business in Silicon Valley.

!2 Schlit, W. Keith. 1990. The Entrepreneur’s Guide to Preparing a Winning Business Plan and
Raising Venture Capital. Englewood Cliffs, N.J.: Prentice-Hall.
Worth owning. Contains lots of useful plan formats, definitions, and sources of
capital.

Shim, Jal K., Joel G. Siegel, and Abraham J. Simon. 1986. The Vest-Pocket MBA. Englewood
Cliffs, N.J.: Prentice-Hall.
Presents useful guidelines for understanding the subtleties of financial statements
and financial decision making. Assumes that the reader is familiar with basic
accounting principles.

Silver, A. David. 1985. Venture Capital: The Complete Guide for Investors. New York:
Ronald Press, John Wiley & Sons.
Explains how customers-i.e., investors-think when doing financing. The book
about the venture capital community. Describes how the financing of funds and of
companies works.
 Bibliography 373

Smith, Douglas K., and Robert C. Alexander. 1988. Fumbling the Future. New York
Morrow.
Story of Xerox's inventions in distributed computing and its attempts to enter the
information-processing business. Useful for understanding the management of
research and the technology-transfer process.

SZ Walker, John. 1987. The Autodesk File. Thousand Oaks, Calif.: New Riders Publishing.
A great book on starting a software company. I recommend reading it and using its
memos directly in managing a company.

White, Richard M. 1977. The Entrepreneur's Manual. Radnor, Pa.: Chilton Book Co
Presents a wonderful set of rules for understanding and managing all aspects of a
business (e.g., salespeople and how to close sales). Chronicles in an almost encyclo-
pedic fashion many aspects of a start-up. Also contains a good discussion of
building a plan.
INDEX

AI Corp., 232 manufacturing strategy, 144-45

Alexander, Robert, 319,373 plan, 238-39,331-33
Alintuck, Ronna, xiii, 230,297-99 price-performance, 154-55,193-96
Allen, Paul, 22, product position, 193-96
Alliant, 159 quality guarantee, 182
Alpha test stage. See Stages and Engineering, technology balance sheet, 126-27
project phases Articulate Systems, 279
Amdahl Corp., 191,235 Artificial intelligence, 18749,232-33,353-54.
Amdahl, Gene, 3,131,235,297 See also knowledge engineering
Amdahl's law, 198 Ashton-Tate, 279
American Research and Development (AR&D), AT&T. See UNIX
52 Atari, 231
Analytica, 238,272,276-82 Augustine, Norm, 369
Andor, 235 Autodesk, 42-3,211,234,296
Animation, 354
Apollo, 22,159,179,316-17,323-29 Backus, John, 104
business plan, 3 9 4 2 Backward integration strategy, 51
Apple, 22,50,120,183,234,350 Baker, Jim and Janet, 67,282-83
Macintosh, 336-37 Balmer, Steve, 22,
Applications. See Marketing, how used Basic rules. See Rules
Applied Materials, 224 Baskett, Forest, 297,325
Architecture and architects, 119-22,137,348-52 Baty, Gordon, 369
Ardent, 80. See also Stardent Bechtolsheim, Andy. See Sun Microsystems
architecture, 122 Bell, Gordon, 153,166,175,198-99,368,369
first graphic supercomputer, 193-96,317 Bell, Gwen, xiii
hiring, 123 Bell, Labs, 102, 104
incompatible architecture, 190 Bell-Mason Diagnostic, vii-xii, 239,251-71
lessons, 333 applying, 25456,267-70

374
 Index 375

basic rules, 267-70 (see also Rules) Capital sources, 67-69. See also Financing,
dimensions, 253,263-64 Financeability
relational graph, 25455,266-67 Card, David, 369
rules, 253-54,264-70 (see also Rules) Carnegie Group, 232
stages, 252-53,256-62 (see also Stages) Carnegie-MellonUniversity, 215-16,232,321-
Bentley, Tom, 126 23
Beta-testing.See Stages and Engineering, Carr, Robert, 124,368
project phases, Cash, 59-64. See also Flaws and Rules,
biochips, 374 buying time, 60
Birnbaum, Joel, 179 financeability, and control
Bitzer, Don, 233-34 Cirrus Logic, 289-90
BizPlanBuilder, 38 Gateway, 299
Blank, Tom. See MasPar Gensym, 306
Board of directors, 27-33. See also Flaws and MasPar, 314
Rules Ovation, 275
Cirrus Logic, 288-89 having too much, x, 63
function, 27-29 manufacturing needs for, 143
kitchen, 79 respect for, 23
meetings, 30 running out of, 60-61
selection criteria 27-29 Cashing out, 5-6,
size, 27-29 Casio, 231
Borland, 279-82 CEO. See Chief executive officer
Bosworth, Adam, xiii, 276-82 Channels of distribution. See Marketing
breakthrough flaws, 129-30,135,147 Chief executive officer (CEO), x, 11-19. See also
BREIT, 188-89 Flaws and Rules
Bricklin, Dan, 104,183 conflict resolution, 88
Brooks, Fred, 126,369 management ability, 81
Brown, Owen. See Sun Microsystems marketing role, 202
Bruno, Albert, 368 relationship to board of directors, 27-30
Burgelman, Robert, 369 requirements, 11-13 (See also rules)
Burkhardt, Henry, 131-32 wealth handicap, 11
Business plan, 34-58. See also Flaws and Rules, Chief operating officer (COO), 11-12
Apollo, 3 9 4 2 Chief technology officer (CTO).See technology
Autodesk, 4 2 4 3 balance sheet,
Cirrus Logic, 287-89 Chopp, 192
creation, 3 Christy, Peter. See MasPar
formats 35-36,37-39 Cimflex, 232
Gateway, 296 Cirrus Logic, 62,273,283-89
Gensym, 305 Clark, David, 106
Ovation, 274 Cobol, 174,
Poduska format, 35 Cocke, John, 104,179
purposes 34-35 Code museum. See Computer
spreadsheet model, 3,36 Coit, Steve, 29
Sun Microsystems,42-43 Communications, 351-52
vision, 36-39 Company culture. See Culture
Buyers and buying criteria. See Product and Company vision. See Vision, Plan
Marketing Compaq, 28
Bygrave, William, 368 Compatibility. See Computer
Complementary Metal Oxide Semiconductor
C language, 183 (CMOS). See Semiconductor
Cadence, 295 Complex instruction-set computer, 176-80
Calculator, 177 Components, new. See Technology, and Flaws,
Cane, Dave, 311 technology
376 Index

Computation, cost of, 93 management and communications, 77-78

Computational chemistry, 215,355 organizational design for, 78-79
Computational fluid dynamics, 355-56 planning importance, 76-77
Computer. See also Architecture and Product profitability link, 78-79
array processor, 194 Control Data Corporation (CDC), 50,100,134,
class consolidation, 166-67 198,233-34,237
class creation, 169-70 Convergent Technologies, 190,231
classes, 161-80, 194 Convex, 159,236
clock evolution, 181 COO. See Chief operating officer.
code museums, 174-78 Cooper, Robert, 369
compatibility, 190-92 Corporate research. See Research
dimensions of performance, 195 Cost predictability. See Manufacturing
evolution, 164,178-80 Cray Research, 110,343,34849,362-64
fault-tolerant, 159,177 Cray, Seymour, 3,104,183,347
functions, 160-62, Creativity and standards, 116
future, 186,337-38 CTO. See Chief technology officer
general purpose versus special products, Culler Scientific, 192
186-88 Cullinet, 235
generations, 171-80 Culture
graphics super, 194, company, 11,17,21,23,26
historical evolution, 171-80, 186 Silicon Valley, 26
level-of-integration,163 Customer Advisory Board (CAB),31,242
mainframe, 167,194 Customer applications profile. See Marketing
micro, 177,34044 Cydrome, 51,192
microprocessor, 159
minicomputer start-ups, 174-76 Data General, 148,279
minisupercomputers, 159,167,194 Davidow, Bill, 202,228,370
multi, 177, 18486 Davis, Robert, 370
neural computing, 350 Deal, Terrence, 370
notebook, 167 Defense Advanced Research Projects Agency
performance and price, 153-55,170,178-81, (DARPA),99-101,315,318,349,363
195,34045 Dell Computer, 18,130
personal, 167,176-78,186,33638,34043, Dell, Michel, 18
356-57 DeMarco, Tom, 370
pocket, 167 Department of Energy
product space, 160-61 research, 100-101
supercomputer, 167,194,362-64 DePree, Max, 370
superminicomputers, 159,167,194 Design Power, 192,224
systolic, 350 Design process, 116-17. See also Technology
wallet, 167 balance sheet
weight, 166-67 Design rules for supercomputers, 198-99
workstation, 167,177,194,316-38 Designing great products, 180-86
Computer aided instruction, 233-34 builder-is-user, 183
Computer and Business Equipment Manufac- cost and quality, 181-82
turers, 204 evolvability, 184-85
Computer Associates, 235 Dessault, CATIA, 192
Computervision, 211,234 Development stage 111,110-12
Concept stage I, 6-7,109-11. See also Stages Digital Equipment Corporation. See also Ken
Conflicts among team members, 21 Olsen
Conner, Finis 8,300 architecture, 120-22
Control, 76-84. See also Flaws and Rules experience, viii
input-output management, 77 incompatible products, 190
 Index 377

market map, 220-24 ETA, 71,192,237. See also Control Data

market restructuring, 34042,358-62 Corporation
minicomputer performance, 340-42 Exit criteria. See Cashing out
mocking up products, 189-90 Expert systems. See Knowledge engineering
organization, 78-79,330
PDP-5, PDP-8 origin, 174,183 Fairley, Richard, 370
PDP-11,184 Fault-tolerant. See Computers
printed circuit boards, 147-48 Features, functions, and benefits (ffb). See
product development, 102-3,361 Product
starting up, 52, 17476,328 Fifth generation, 178-80, 187. See also computer,
universities, 214-15 generations
VAX, 180,201,343,361 Financeability, 64-76. See also Flaws and Rules,
VAX strategy, 37,102,358,360 factors, 64-65
workstations, 317-18,327,334-35 valuation, 6547,290,
Dimensions (by which company is measured), Financing, 3,5-6,29
253,263-64 cram down, 45,69
Disk densities, 94-98 criteria for partner, 69-70
Disks, 73,171-76 dumb and easy money, 69
Distributors. See Marketing, examples. (see Chapter 11)
”Do” orientation, 19-20 family and friends, 68-69
DOS. See Microsoft lack of venture, 62-63
Dragon Systems, 67,128-29,272,282-83 laundered venture, 237
Drucker, Peter, 154,370 liquidity preference, 71
Duncan, Mark, xiii presenting case, 70-71
Dvorak, John, 371 rounds, 3,7-8,71-72
self-funding, 67
Early adopter. See Markets, emerging Silver’s laws, 66
Eckert, J. Presper, 171 sources, 67-69
Economy of scale, 154 time, 71
Einstein, Albert, 251,297 too much, 62-63
Elahian, Kamran, 285-86 Flaws
Electro-optical. See Disks board of directors, 30-31
Elexi Computer, 130,148,192 industry inexperience, 30
Ellison, Larry, 237 investor dominated, 30
Emitter coupled logic (ECL). See Semiconduc- reviews lacking, 31
tor running the company, 31
Encore Computer, 51,184-86, business plan, 44-54
Engineering. See also Technology and Technol- motivation, 49
ogy balance sheet multiple agendas, 50
design for manufacturing, 151 multiple companies in one, 50-54
hardware, 107-12 research required 4 6 4 7
management, 107-12 seed stage lacking 49-50
marketing role, 86, 88,202 sustaining product or technology 48-49
process, 107-12 unrealistic, 44-46,48
project phases, 109-12 cash, 61-63
schedule fantasy factor, 77,280,304 investor take-over, 62
software, 108-12 no concept funding, 61
specifications, 118 stuck in a stage, 61
Entrepreneurs, 11,49. See also CEO too much 62-63
Ethernet, 319-20 CEO, 13-16
Europe, 365 energy, 14,
Evans, Dave, 318 hiring ability, 15
378 Index

Flaws, CEO (continued) support plan, 248

integrity, 14 team, 23-24
intelligence, 14 ego conflict 24
management ability, 15 mercenaries 23
selling company, 16 respect ladting, 24
team building, 16 technology and engineering, 129-34
control, 79-82 breakthroughs required, 129-30
inadequate financial, 80-81 component missing, 131-32
major slip, 82 dehiring inability, 133
management inadequate, 81 hiring inability, 133
measure lacking or poor, 80 leaking ideas, 133-34
support lacking, 81 little or no technology, 130
unrealistic plan, 79-80 not-invented-here (NIH), 131
financeability, 72-73 pre-announcing product, 134
crowded product field, 72 research required, 126-29
dearth of capital, 72 Focus groups. See Marketing
investor disagreement, 73 Fortran, 174
investor fatigue, 73 Forward integration strategy, 51
manufacturing, 146-48 Founder involvement, 329-31
make/buy decision, 146 Franksten, Bob, la
marginal component or process, 147 Fredkin, Ed. See Gensym
over capaaty, 14647 Funding. See Financing
process breakthrough, 147-48 Future, 339-66
product flow erratic, 146
marketing, 230-40 Gallium arsenide, 179
always emerging market, 233-34 Gartner Group, 340,342,358-59
cost model wrong, 238 Gates, Bill, 18,22
emerging markets take time, 231-32 Gateway, 67,273,289-99
hires are wrong, 239-40 General Electric, 325
incomplete product, 237 General Instruments, 284
just another product of type x, 236 General Motors, 207,327
niche inadequacy, 236 Gensym, 233,273,299-307
no technology monopolies, 233 Gershman, Michael, 370
one customer, 236-37 Gilder, George, 370
pioneering product, 230-31 Gladstone, David, 37,65,370
pre-announcing, 238 Glass, Robert, 369
universities are rarely markets, 238-39 GO Corp., 124
university pre-announce, 239 Goel, Prabhu, xiii, 289-99
walled cities, 234-5 Gold Hills, 183,
who and why missing, 237-38 Goldberg, Adele, 320,370
product, 186-96 Gomory, Ralph, 88,368
incompatible products, 190-92 Gorman, Dennis, 13,
mock-ups aren’t products, 189-90 Grayson, Paul, 124-25
partners take product, 192 Grove, Andy, 22,125,370-71
specialized product, 186-87 Greed, I-2,7,44
specifications deficient, 192-96 Greata, Mike, 22
technology but no product, 187-89 Grillos, John, xiii, 62,69-70
sales, 247-48 Grosch’s Law, 153
hiring plan, 247 Gyration, 53-54
ineffective leader, 247
order gestation time, 247 Hackworth, Mike, 70,285-89
sales costs high, 247 Hammock, Jim, 11
 Index 379

Handwritten input, 350-51 microprocessors, 176-79,183,34942

Harrar, George, 14,372 Intellectual disciplines, 210
Harvard Center for Information Policy Intellicorp, 192,232
Research, 204-5 International. See Marketing, global
Hawkinson, Lowell, xiii. See also Gensym International Data Corp., 204-6,
Heller, Andy, 334 Interpersonal computing, 178,200
Hennesey, John, 179 Intimate computing, 231
Heuristics, 264. See also Rules Invention, 36445
Hewlett-Packard, 126,145,176,179, 183,319, Investors. See Financing and Capital sources
330. See also Apollo Iverson, Ken, 104
workstation, 316-17’334-35
High margins, importance of, 159 Japan
High Performance Computing Initiative, 351 engineering, 36546
Hill, Larry, 284 innovation, 365
Hillis, Danny, 307 market, 224
Hiring, 15,18,25,33,82,329-330 marketing, 227
manufacturing, 143,145,148-50 software quality, 117
pygmies, 15,123,133 Javelin, 238
technical 122-125,133,137-38 JIAN Company, 38,371
Hoff, Ted, 183 Jobs, Steve, 3,22,
How product is used. See Marketing Apple computer ownership philosophy, 183
Humphrey, Watts, 117,138, 371 industrial design, 126
interpersonal computing, 178,200
Iacocca, Lee, 227 multimedia, 156
IBM, NeXT, 336
architecture, 119-20 Joy, Bill, 215,338. See also Sun Microsystems
disks, 174,234 Juliussen, Karen and Egil, 371
dominance, 171,176,335 Just Another Workstation (JAWS),316,335-36
Fortran, 174
mainframe cost, 340,342 Kaiser, Bill, 295-96
mainframe market restructuring, 359-62 Kalb, Jeff, xiii, 307-14
personal computer, 102,167,177,336-38. See Kapor, Mitch, 22
also Computer, personal Kaufman, Sy, 60
pocket computers, 231 Kawasaki, Guy, 371
pre-announcing product, 134 Keeley, Bob, xiii, 263
reduced instruction set computer, 179, Kelvin, Lord, 251
research, 102, Kemeny, John, 104
supercomputers, 36244 Kendall Square Research, 131-32,36243
System/360 ..., 102, 166, 174 Kennedy, Allan, 370
universities, 214 Khosla, Vinod, xiii, 326. See also Sun
workstations, 317,33435,362 Microsystems
Image processing applications, 355 Kilby, Jack, 174
Implementation phase, 112. See also Engineer- Kingsley, Gordon, 294
ing, project phases Kitchen table. See concept stage
Information business map, 204-5 Kleiner’s Law, 71,311
Innovation, 365 Knowledge engineering, 187-89,222-24,232-
Integrated circuits. See Semiconductors 33,353-54. See also Gensym
Integrated Services Digital Network (ISDN), Kotler, Philip, 371
178,352 Kubota Corporation, 14445,331-33
Intel, 22,50,102-3. See also Computer, Kurtz, Tom, 104
personal, workstation Kurzweil, Ray, 12829,339,34546,366,371
future, 34044 Kvamme, Floyd, 227-28. See also MasPar
380 Index

Lacroute, Bernard. See Sun Microsystems time-to-market, 143,151

Lampson, Butler, 319-20,368 Manzi, James, 22
Larsen, Judith, 27,372 Market. See also Marketing
Lawyers buyer role, 207-8 (see also Marketing, who
on board of directors, 28 buys)
Leadership, 17. See also CEO driven, 86
Levine, Michael. See Gensym emerging, 214,230-34
Levitt, Theodore, 371 existing or new, 158
Lisp, 183,222 knowledge engineering market map,
Lisp Machines, Inc. 299-300 market map, 22&24,228,24142
Lockheed, 192 niche and nichemanship, 15940,236
Lotus, 22,28,71,279-80 professional discipline (see Market, user’s
Lucasfilm, 102 profession)
luck, 73 segmentation, 206-12
size, 2046,241
McCance, Henry, 294-99 substitution, 158
McCarthy, John, 104 surveys, 241
McKenna, Regis, 22,371 user’s organization, 208
McNamara, John, 153,369 user’s profession, 207-1 1 (See also Marketing,
McNealy, Scott, 22. See also Sun Microsystems who)
Magnetic technology. See memory user, 207 (See also Marketing, who)
Maidique, Modesto, 369 Market development stage IV. See Stages
Management. See also CEO, Team, Control Marketing, xi, 201-45. See also Market, Product
Bo-Peep, 81, application notes, 229
by objectives, 83 applications (see Marketing, how used)
CEO, 17-18, balance sheet, 225-30
control, 83-84 benchmarking, 229
engineering, 107-12,135,137-39 brand re-labelling, 219
team, 25-27, calibrations stage (see Stages)
Manufacturing, 140-52 channels of distribution, 218-24,248-49
breakthrough, 14743,150-51 Cirrus Logic, 288
cash requirements, 143 collateral material, 229
components, 147 communication, 212
cost predictability, 144 company killer, 202
design for,l51 competitive analysis, 229
disks, 141 competitive table, 219
hiring (see Manufacturing, staff require- conflict with engineering, 86,88,201-2
ments) customer applications profile, 214,241,243
inventory, 144,151 definition, 201-25
makelbuy, 14243,145 demographics, 212
offshore, 14445,149 direct mail, 219,229
printed circuit boards, 140,14748 direct sales, 219
process investments, 142,149 distributors, 219
quality levels, 141, 149, 151 early adopter (see Markets, emerging)
Sander’s principles, 141-45 expansion stage (see Stages)
semiconductors, 141 features, functions, and benefits (see Product)
software, 140 focus groups, 229
specifications, 118 global, 224
staff requirements, 143,145, 148-50 head of, 22f5-28
strategic partners, 149 hiring, 227,23940,24344
 how used, 203,212-13,216-17,222-24 Micrografx, 124-25,
industry (see Standard Industry Classifica- Microprocessor. See also Computer
tion code) bugs, 132
international (see Marketing, global) evolution, 92,176-80
key customers, 212 Microsoft, 18,22, 50, 67, 102,279
original equipment manufacturer (OEM), workstation clash, 317,336-38
219-24,323-28 Minicomputer start-ups, Minisupercomputers.
Ovation, 274-75 See Computers
plan, 228,242,244 MIPS Computer Systems, Inc., 52-53
press relations, 230, Mock-ups and models, 189-90
processes, 229 Modular Advanced Design, 52
product role, 202 Moorby, Phil, 292-96
public relations, 229,238-39 Moore, Bob, xiii. See Gensym
sales channels (see Marketing, channels of Moore, Gordon, 3,22
distribution) Moore’s Law, 90,347
sales support literature, 230 Morgan, Jim, 224
steady state (see Stages) Motorola microprocessors, 17740,318,321-23,
system integrators, 219-24 334-36
trade shows, 229 Mudge, Craig, 153,369
training sales, 229 Multicomputer. See Parallel processing
university role, 208-9,214-16,238-39 Multimedia, 156,233
value added resellers, 219-24 Multiple Instruction Multiple Data (MIMD).
what is sold, 2034,237 (see also Product) See Parallel processing
when purchased, 203 Multiprocessors. See Parallel processing
where purchased, 203,217-24 Multisite start-up, 45-46
who buys, 203-12,237,240
wholesalers, 219-24 NASA research, 100
why customers buy, 203,216-17,237-38,241 National Cash Register (NCR), 50
Markula, Mike, 22 National Science Foundation, 99,101
Mason, Heidi, xiii. (See also Bell-Mason NCUBE, 362
Diagnostic) NEC (Nippon Electric Corp.), 34749,364
MasPar, 273,307-14 Negative productivity, 133
Massachusetts Institute of Technology, 216, Nelson, David, xiii, 22,116, 166. See also Apollo
284,288,318,329 Nesheim, 5,38,66,371
Massive parallelism. See Parallel processing Neuron Data, 232
Mathematics applications, 355 New England. See Start-ups
Mathews, Karen, xiii NeXT, 155,336,350
Matsuda, Naohisa, 34 Nichemanship. See Market
Mauchly, John, 171 Nikolls, John. See MasPar
Mechanical engineering applications, 355 Nondisclosure contracts, 239
Meindl, James, 91,368 Norris, Bill, 233-34,
Memory Not-invented-here (NIH) syndrome, 131, 190-
cost, 96,34345 92,365
secondary, 90,93-98,34345,347 Noyce, Bob, 3,22,174
size, 94, 347
Mendelson, H., 153, 368 Objectivity, 63, 123-24
Michelman, Eric, xiii, 276-82 Offshore manufacturing, 144-45
Michels, Allen, 239,333 Olsen, Ken, 14,189
Microelectronics and Computer Technology wallpaper remover, 87,276
Corp. (MCC), 187 Oracle, 237
382 Index

Original equipment manufacturer (OEM). See opportunities, 169

Marketing pioneering, 158
Osborne, Adam, pocket computers, 231,
Ovation, 272-75 point, 197
position, Ardent, 193-96
Paper, 345,352,353 replacement, 158-59
Paradox, 279 special versus computer, 186-88
Parallel processing, 179-80,182. See also specification, 196
MasPar, Thinking Machines, substitution, 158
Patil, Suhas, xiii, 62,282-89 technology to product, 196
Patterson, Dave, 179 views, 152
Payoff of start-ups, 66 Product development stage 111. See Stages,
Peachy, Jim. See MasPar Productivity, 132
Peglar, Rob, 237 Professions. See Market, user’s profession
People, x, 10-33. See also Chapter 11 Profitability,
Performance and price. See Computer Cirrus Logic, 288-89
PERQ, 321-23 Dragon Systems, 28243
personal computer, see computer Gateway, 292-99
Peters, Tom, xii, 27,372 Gensym, 303-7
Phases. See Engineering, project phases, importance of, 4-5,23,76,328
Pixar, 102 Sun Microsystems, 325-28
Plan. See Business plan Pygmy hiring. See hiring
Poduska, Bill, xiii, 22, 35, 39-42,263. See also Pyramid, 179
Apollo
Poquet, 231 Quality
Press. See Marketing design, 142,181-82
Price-performance. See Computer guarantees, 182
Printed circuit board technology, 132,14748 manufacturing, 14142
Product, 152-200. See also Computer; Market; schedule conflict, 126
Marketing, software, 117
applications; Marketing, how used
buying factors, 153-57 Recording whiteboard, 87
commodity, 158 Reduced Instruction Set Computer (RISC), 48-
competitiveness, 197 49,101-2,17740,343,348,358-64, Seealso
completeness, 203 IBM
cost and quality, 156 MIPS, 52-53,17941
data sheet, 197 Relational graph
dimensions, 198 company, 254-55 (see also Analytica,
evolution, 169 Gateway, Gensym, MasPar, Ovation)
evolvability, 153,184-86 technology balance sheet, 115
features, functions and benefits, 196,216-17, Research
244 consortia, 187
goals, 196 corporate, 100,319-20
high margins, need for, 159 government funded, 99
invention, 169 individual, 1 0 2 4
just another product, 236 source for start-up, 98-99
lead time, 197 university role, 99-103 (see also Marketing)
manuals, 197 Ridge, 179
market requirements, 157-60,212,216-17 Rifkin, Glenn, 14,372
mock-ups and models, 189-90 Ritchie, Dennis, 104
new capabilities, 155-57 Roberts, Edward, 372
nichemanship, 159-60 Robots, home, 357
 Index 383

Rock, Arthur, 10,22,263 sustainability, 56

Rogers, Everett, 27,372 cash, 63-64
Rosen, Ben, 28 continue with company, 64
Rosen, Bryan, 321-23 execute the seed stage, 63
Rosenstein, Joesph, 28-29,368 three months supply, 64
Rosenthal, Dan, 279-82 write business plan, 63
Round, investment. See Financing CEO, 17-19
Rule-based programming. See artificial finance selling, 19
intelligence hiring key people 18-19
Rules intelligence,energy, ethics, quality, 17
basic, 268-70 management, general 19
board experience and expertise, 270 managing the seed, 17
board not decision makers, 270 selling vision and company, 18-19
board reviewing company, 270 start-up management, leadership 7 7
business plan, 268 control, 82-84
business plan integration, 268 financial control, 84
business plan realism, 269 founder capabilities, 82
cash for three months, 270 management-by-objectives,83
cash to complete stage, 270 minimal deviation to plan, 83
CEO intelligence,energy, quality, 269 objectives met, 83
CEO manager and leader, 269 seed accomplishments, 83
CEO recruiter, 269 financeability,74-75
CEO salesman for board, capital, exogenous factors, 75
customers, 269 founder commitment, 74
control by objectives, 270 imprimatur of experts, 74
control processes, 270 market for the plan, 74
control to plan, 270 plan and people, 74
finance is available, 270 seed success, 75
manufacturing inventory control, 268 valuation sanity, 75
manufacturing processes, 268 for measuring company, 253-54,264-70
manufacturing quality products, 268 manufacturing, 148-50
marketing plan, strategy, 269 cost and quality, 149,
marketing plan, tactics, 269 inventories, 150
product features, functions, and benefits, partners, 149
268 plan, 150
product next generation, 268 plant size and location, 150
sales team and model, 269 processes defined, 149
team "do" orientation, 269 product design, 150
team experience, 269 product introduction, 150
team, hiring ability, 269 strategy, 149
team is a team, 269 marketing, 240-45
technology balance sheet, 268 business and market plan together, 242
board of directors, 31-33, buying rationale, 241
customer and technology advisory board, customer application profiles, 243
32 customers and applications, 241
identified, 32 customers identified, 240
operational expertise, 33 market map at seed, 243
business plan, 54-57 market map, 241
all strategic elements, 54-56 market plan and details at seed, 244
detailed tactical plan, 57 market plan outline, 242
financing plan, 56 market size estimate, 241
next stage detailed, 56 market vision, 244
384 Index

Rules, marketing (continued) resources, 246

marketing head, 243 Salesman as CEO, 12
product specifications, 242 Sanders, Matt, rules for manufacturing, xiii,
product, 196-99 141-44
competitive lead, 197 Schedule fantasy factor. See Engineering
data sheet with features, functions, and Schedules, engineering, 117-18,125-26,136,
benefits, 197 138-39
follow-ons, 197 Schlit, Keith, 372
goals, 196 Schmidtt, Roland, 88,368
specifications, 196 Scott, Mike, 22
technology to product, 196 Seed stage II,6-7,109-11. See also Rules, Stages,
understanding requirements, 198-99 Semicomputer company, 272,285
user’s manual, 197 Semiconductor. See also Microprocessor
sales density, 90-92
channels of distribution, 248 evolution, 89-92,181,361
sales leader, 249 scale of integration, 172,17480
selling model, 24849 Semiconductor Research Corp. (SRC), 187
team, 24-27 Sharp, 231
attracting others, 25 Shim, Jal, 372
completely formed, 25 Shirley, Jon, 22
corporate culture, 26 Shoch, John, 19,60
development experience, 24 Short socks test, 16. See also Flaws, CEO
”doing” and ”managing,” 25 Shugart, Al, 60,234,365
hiring criteria and process, 25 Siegel, Joel, 372
manufacturing leader, 26 Single Instruction Multiple Data (SIMD).See
working together, 24 Parallel processing
technology and engineering, 134-39 Silicon Compilers, 182
chief technology officer, 137 Silicon Graphics, 191,236,317,327,330-31,334
development plan, 136 Silicon Valley. See Start-ups
experience of team, 138 Silver, A. David, 11,6546,372
hiring ability, 138 Silverberg, Brad, 277-82
key technologists, 137 Simon, Abraham, 372
limited breakthroughs, 135 Smith, Douglas, 319,373
product architect, 137 Smith, Gordon, 370
product development process, 138 Software Engineering Institute, 117,138
product schedule, 139 Software factory, 108-12
risks identified, 136 Software. See Engineering
technology balance sheet, 134-35 Sony, 365
technology-product transition, 135 Spector, Charles, 22,324
tools, 139 Speech typewriter, 128-29. See nlso Dragon
working proto or model, 136 Systems
Spinner, Arthur, xiii, 27-28,30
Sales, 245-49 &ages of company growth, 6-8,109-12,252-
channels of distribution (see Marketing) 53,25642
head of, 246 alpha testing IIIc, 110-12,259
hiring, 247 beta testing IIId, 112,214,259
marketing conflict, 246 concept I, 6-7,109-11,256,258 (see also
model, 228,24546,248 Chapter 11)
order gestation time, 246-47 designing and building IIIb, 259
plan, 24549 hiring and planning IIIa, 259
processes, 245 market calibration IVa. 26041
 Index 385

market development IV, 64,110-12,259-61 Symbolics,300

(see also Chapter 11) System integrators. See Marketing,
market expansion IVb, 261
market steady-stage IVc, 261 Tandem Computers 27-28
product development "6-7,259 (see also Tarter, Jeff, 160,368
Chapter 12) Taylor, Norman, 368
research (see Cirrus Logic) Team, 19-27. See also Flaws and Rules
seed 11, 6-7, 109-11,258 (see also Gateway) Team requirements, 19-23
steady-state V, 74,261-62 Technology, 85-1 13, I1 5. See also Computer,
transitions, 7-9,262 Engineering, Product, Technology balance
Standard Industry Classification(SIC) code, sheet,
2074,212, architecture, 104-5
Standards balance sheet, 11615,134
need for, 116,139 components, 1034
role of, 243 conflict with marketing, 86
Stanford University, 232 definition, 85
Stardent, 145,239,331-33,361 evolution, 88-90,181
AVS and Dore, 191 monopolies,
Start-ups progress, 88-106,
duplicating, 314-1 5 push, 86-88
government funded, 100, research role (see Research)
motivation, push-pull, 2 revolution, 88-90
New England, 72 (see also Dragon Systems, semiconductors, see
Gateway, Gensym, Ovation, Thinking sources of, 98-1 05
Machines) standards, 105 (see also Sun Microsystems)
performance, 1,66 transfer, 85-86,105-6,319-28
productivity, 1, university role (see Research, Marketing)
program for, 2,7-9 Technology Advisory Board (TAB), 31,242
Silicon Valley, 72,320 (see also Analytica, Technology balance sheet. See also Technology
Cirrus Logic, MasPar) and Engineering flaws and rules, 85,114-
success rate, 66 39
technology sources of, 98-105 architecture, 119-22,137
technology transfer, 105-6 chief technical officer, 119,137
tools for, 4 design process, 11617,138-39
university-based, 99-100, dimensions of, 11415,134
Steady-state stage V. See Stages engineering plan, 117-18
Stellar, 123, 145, 159,317,325 engineering specifications 118 (see also
Storage. See Memory Architecture,)
Strategic partners, 14345,149,192 engineering team and culture, 119,324,
Sun Microsystems future, 123
architecture, 120, management, 123-24,138
business plan, 4243, manufacturing specifications, 118
class of systems, 167 relational graph, 115
founder roles, 22,325-31 rules for, 134
start-up, 108 standards, 116
story of, 316-17,322,325-30,334,338 technical resources, 122-23,139
Supercomputer, minisupercomputer, and technology, 115
superminicomputer. See Computer Teknowledge, 232-33
Supercomputer design rules, 198-99 Telemarketing. See marketing,
Sutherland, Ivan, 318 Thinking Machines, 273,307,31415,349
Swartz, James, 13 Thompson, Ken, 104,201
386 Index

Three Rivers Computer, 321-23 Very-large-scale integrated circuits, (VLSI),

Time requirements, 21-22 definition, 172
Time-to-market, 143 Viatron, 189
Top-line problem, 48. See also Flaws, business Virtual reality, 357
plan Visicalc, 183
Transaction processing, 354-55 Vision, company, 36-37
Transistor companies, 173 Visix Software, 124,239,350
Transistors per chip. See Semiconductor Visualization, 357
Transistor-transistor logic. See Semiconductor Vitesse, 51-52
Transition among stages, 262 Von Neumann Architecture, 349
Trilogy, 71,129-30 VPL. 53
Trilogy flaw, 129-30
Turing test, 346 Walker, John, 4243,373
Two-in-a-box, 11-12 Wallpaper remover. See Ken Olsen
Waterman, Bob, xii, 27,371
UNIVAC, 171,174 WAVETRACER, 123,132,
Universities. See Marketing, Product, and What is sold. See Marketing
Technology Where purchased. See Marketing
UNIX, 101-2,104,180,215,319,322,326,335- White, Richard, 11,38,67,245,373
38 Who buys. See Marketing
Utah, University of, 318 Why customers buy. See Marketing
Workstations. See Computer
Vacuum tube companies, 173 Worley, Bill, 179
Valentine, Don 201,263 Wozniak, Steve, 22
Value Added Reseller WAR). See Marketing
van der Rohe, Mies, 198 Xerox, 238
Vanderslice, 22,324-25 Palo Alto Research Center, (PARC), 101-2,
Vaporware, 274 319-20
VAX strategy. See Digital Equipment Corpora-
tion Zilog, 189
venture capital return, 6546,70-71
ACKNOWLEDGMENTS

Page 66. Numbered lists. A. David Silver, Venture Capital: The Complete Guide for
Investors, copyright 0 1985by John Wiley & Sons, Inc. Reprinted by permission of
John Wiley & Sons, Inc.
Page 96. Figure 5-6. Courtesy of Askmar and Frank Ura, Hewlett-Packard. Re-
printed with permission.
Page 126. Summary of Brooks’ Law. Frederick P. Brooks, Mythical Man-Month,
copyright 01975 by Addison-Wesley Publishing Company, Inc., Reading, Massa-
chusetts. Reprinted with permission of the publisher.
Page 160.Extract. Copyright01989by Soft defter.All rights reserved. Reprinted with
permission.
Page 168.Figure 8-8. Courtesy of the IEEE Scientific SupercomputerSubcommittee,
from “The Computer Spectrum: A Perspective of the Evolution of Computing.”
Reprinted with permission.
Page 201. Epitaph. Reprinted with permission from UPSIDE Magazine, Volume 2,
Number 3.
Page 263. Quote by Arthur Rock. Levering, et.al., The Computer Entrepreneurs, New
American Library, New York, New York. Reprinted with permission.
Quote by Don Valentine. Reprinted with permission from UPSIDE Magazine, vol-
ume 2, Number 4.
Page 300. The Gensym story. Reprinted with permission from Robert Moore,
President of Gensym.

387