Chapter 8

Resource Allocation
What is resource allocation all about?
According to Churchman1, management is responsible for allocating
resources in order to achieve an organization’s purpose.
“In organizations, the decision-making function is the responsibility of
management. In order to execute its responsibility, an organization’s management
requires information about the resources available to it and their relative
effectiveness for achieving the organization’s purpose. Resources are acquired,
allocated, motivated and manipulated under the manager’s control. They include
people, materials, plant and equipment, money, and information.2“

The notion of an “organization’s purpose” may be somewhat vague,

especially if the organization is large. In such cases, the purpose is
sometimes stated as its vision, or goal. The vision or goal is attained
through the achievement of multiple (often numerous, and competing)
objectives. We have defined rationality as the achievement of objectives—
in this context, the objectives that will best fulfill the organization’s purpose.
An important question then is, how does an organization rationally allocate
its resources in order to achieve its goals?
We have often asked executives how resources are allocated in their
organization. After an initial blank stare, we hear responses like — “The
executive committee meets and....”, or “We use the budget contained in the
five year plan....”, or “We start with last year’s resource allocation and make
adjustments based on....”. These responses beg the question of how resource
decisions are made in some ‘rational’ way that is based on their relative
effectiveness for achieving the organization’s purpose(s) or objectives.
In order to make resource decisions in such a rational way, an
organization must do the following:

1
“Measurement for Management Decision: A Perspective”, Richard O. Mason and E. Burton Swanson,
reprinted from the California Management Review, Vol 21, No 3 (Spring 1979), p14
2
Ibid.
236 Decision By Objectives

Identify / design alternatives (e.g., alternative R&D projects, or operational

plans for alternative levels of funding for each of the organization’s
departments)
Identify and structure the organization’s goals into objectives, sub-objectives,
sub-sub-objectives, and so on.
Measure how well each alternative contributes to each of the lowest level sub-
objectives
Find the best combination of alternatives, subject to environmental and
organizational constraints.

The Silverlake Project

Before looking at these activities in detail, a look at the underlying
conditions at one organization that was able to successfully make ‘rational’
resource decisions will help people in your organization relate to the
difficulties of traditional approaches and, hopefully, welcome some
improvement in your current resource allocation process.
In the forward to The Silverlake Project: Transformation at IBM3, Tom
Peters wrote:
“This is a remarkable tale. IBM had an amazing success with its AS/400 mid-range
family of computers. ... Disarray is too kind a word for IBM’s position in the
growing, important mid-range computer business in 1986. Competitors were
attacking from every point on the compass. A major project aimed at righting the
ship was an expensive fiasco, and had to be canceled. ... Twenty-eight months
later, a relatively neglected development lab in Rochester, Minnesota was the talk
of IBM. Two years later, the same group had added the prestigious Baldridge
quality prize to impressive gains in market share and profitability. The enormity of
the shift is hard to overstate. (Tally the revenue the group generates and you’d have
the world’s second largest computer company - behind IBM, of course.)”

The process was not an easy one however, as evidenced from some of
the participant’s observations:
“The skirmishes started from day one.”

“What made it even more frustrating was that everyone seemed to have a legitimate
claim for funding ahead of someone else. But in reality, there just wasn’t enough
money to fulfill everyone’s wish list.”

3
Bauer, Roy, A., Collar, Emilio, and Tang, Victor The Silverlake Project: Transformation at IBM Silverlake
Project: Transformation at IBM. Oxford University Press, New York, 1992.
Chapter 8—Resource Allocation 237

“Our quandary was a quintessential one. Like so many enterprises, we had to cope
with the untenable demands of satisfying virtually unlimited needs on a very
limited budget.”
“The real problem, though, was that each organization was evaluating and
committing to their piece-parts with no real understanding of how their decisions
were affecting the whole. And although we had a consolidation process for the
piece-parts, we were unable to convincingly demonstrate that the final result would
balance the needs for market share, technology leadership, and for being affordable
and competitive. We had no methodical, objective basis for making these tradeoffs,
especially in a way that served our overall business objectives. Consequently, we
had trouble explaining our decisions to those affected or to those we answered to in
a credible, defensible way. Indeed, for as rational as most managers want to be,
when it comes to allocating resources, they frequently lack the methodologies for
making systematic decisions. So they’re forced to act by decree. Or whim. Or,
worse, they wind up, like a practiced old Capitol Hill pol, making stopgap attempts
at appeasing the sundry, and often competing, interests found in most
organizations.”

“So one of two things happened. We wound up taking funds from everyone, right
across the board. Or we’d simply cancel a part of the Systems Plan. ... But no
mater what form it took, this give and take ... caused no small amount of
consternation, especially for those of us on the giving end.”
“Victor Tang, who was in charge of planning, and Emilio Collar, who oversaw
market analysis, watched as Furey, Schwartz, and other general managers struggled
with such decisions. They figured there had to be a better way. They viewed this
struggle as an issue fundamental to strategic decision-making, one only
compounded by the vast complexity of global markets. It simply begged for a
more rigorous and systematic process. So together they embarked on an approach
for setting priorities as the basis for allocating resources... To their way of
thinking, the only way to make sound decisions for allocating resources was to
create a priority ranking for each and every one of the line items themselves.”
“But what would be the basis for such a ranking? Markets? Technological
considerations? Financial objectives? It actually had to be done on the basis of all
three considerations. Not only that, we would have to take all three into account in
a balanced way - one that would accomplish the broadest goals as well as the
narrower business objectives...”

Does any of this sound at all familiar? What do you do at your

organization? Here is what they did at IBM Rochester:
238 Decision By Objectives

“Thankfully, Collar found a model for helping us make our priority-setting

decisions - a methodology to render the ranking process more objective and
systematic. It also allowed us to take any number of criteria into consideration. In
short, it enabled us to deal with our situation in all of its complexity. It was called
the Analytic Hierarchy Process (AHP),
“...as part of AHP, you, the decision maker, get to build the hierarchy. You
establish the goal, the criteria, and the options. In so doing, you can actually bring
ideas, anecdotal experience, even emotion into the process.”
“Automated in the form of a relatively inexpensive software program called
“Expert Choice”, AHP is an extraordinarily powerful decision-making tool. It
brings structure to a decision-making, yet it’s flexible because you get to design a
hierarchy of goals, criteria, and options customized to the particular problem at
hand. It can be used with groups, and, as a collaborative effort, it can bring
consensus to decision-making. It allows you to quantify judgments, even
subjective ones. It also forces you to consider the interdependencies of your
criteria to meet your goals. AHP pinpoints for you where the impacts are the
highest or inconsistent. You can play “what-if” games with it...”

“AHP allows you to set priorities by taking several factors into consideration—
factors that interplay and affect each other. In building the hierarchy, you can have
goals and sub-goals. You can have several layers of criteria. You can also deal
with several layers of options. In other words, you can address extraordinarily
complex situations, ones with multi-dimensions that have interconnections every
which way.”

“AHP became the template we imposed on our efforts to rank the line items of the
System Plan so that we could earmark funds in a much more methodical, rationally
defensible way.”
“All this had very real consequences. In the past, we would sometimes chase
markets simply because some highly placed executive decreed we should. Usually
these decrees were based on an anecdotal experience with a particular customer or
industry. But with our priority ranking in hand, it became easier to fend off such
unjustified dictates. ... For the first time, we could confidently articulate what
businesses we were in and, more importantly, which ones we were not.”
“We had come up with a process for setting priorities and, thus, for making
decisions on earmarking resources. It’s a rare organization that doesn’t find itself
in need of some similarly systematic process. In today’s competitive global
marketplace, where almost everyone is finding themselves having to do more with
less, figuring how much money should be spent where may be one of a manager’s
most difficult tasks. But the problems and pain can be obviated by setting
Chapter 8—Resource Allocation 239

priorities. Its simple: Before you decide how to budget, you’ve simply got to know
what’s most important, especially in reaching your organization’s overarching
goals in a holistic, balanced way. As straightforward as that sounds, however, it’s
hardly a simple task. Making decisions today depends on taking into consideration
any number of interdependent goals, criteria, and options. As we proved though,
relying on a model can make the job of setting priorities to allocate resources much
more methodical and objective and, therefore, much more credible and defensible.”
The details of how they did it are not presented in The Silverlake Project,
although the authors do discuss some specifics of what they did. We will next
present the details of several approaches to allocating resources that can be used to
fit in with any organization’s objectives and constraints. Before looking at the
details, it is important to once more focus on the big picture—what needs to be
done when allocating resources and what the consequences of a rational, systematic
resource allocation methodology can be. Bauer, Collar, Emilio, and Tang, authors
of The Silverlake Project, write:
“Making tradeoffs is a fact of organizational life, especially in an era of doing more
with less. So priorities have to be set. But those priorities must be determined on
the basis of the enterprise’s overall objectives. Resource decisions need to be made
holistically, that is, with their consequences to the entire enterprise and all its parts
in mind.”

“Setting priorities - priorities that will serve as a guide to resource decisions—

shouldn’t be a matter of guesswork. It must be done through a process that’s as
systematic as possible. And one which produces repeatable results. This is
precisely what we did—not only in allocating resources, but ultimately, in
determining the shape of the entire Silverlake Project.”

Methodology Overview
There are a variety of ways to achieve a systematic, rational, and
defensible allocation of resources that will provide a competitive advantage
to an organization. The methodology discussed below is quite flexible and
can be adapted to a wide variety of situations and constraints. As outlined
above, the methodology consists of the following steps4:
1. Identify/design alternatives

2. Identify and structure the organization’s goals and objectives

4
Keep in mind that these ‘steps’, are part of a ‘process’, in which iteration is extremely important.
240 Decision By Objectives

3. Prioritize the objectives and sub-objectives

4. Measure each alternative’s contribution to each of the lowest level sub-objectives

5. Find the best combination of alternatives, subject to environmental and

organizational constraints

Identify/design alternatives
Expertise in the art and science of identifying and/or designing
alternatives lies in the domain of the decision makers, who have many years
of study and experience to bear on this task. Our goal here is not to tell
them how to do this, but, instead, to help them better measure and
synthesize in order to better capitalize on their knowledge and experience.
Even if alternatives have already have been identified, e.g. R&D project
proposals as responses to a request for proposal (RFP), it might be possible
to augment or redesign these as part of the ‘process’. For example, if, after
one ‘iteration’ of the resource allocation methodology for allocating funds to
internal R&D project proposals, it may be to the organization’s benefit to
make known the ‘preliminary’ allocation as well as the details of the
evaluation so that proposors can revise their proposals, and, in the process,
improve their contribution to the organization’s objectives. Of course, there
are rules that must be employed in this context to insure fairness. For
example, government agencies may, by law or regulation, have to limit the
process to one iteration with no opportunities for the proposors to ‘improve’
their proposals. Such laws and regulations sacrifice ‘quality’ for ‘fairness’.
While rules, laws or regulations that limit feedback and iteration are
intended to make the process ‘fair’, the tradeoff between ‘fairness’ and
‘quality’ of results should be carefully considered when deciding on the
‘rules’ for the resource allocation process.5
Instead of deciding which alternatives to fund and not to fund, as in the
case of R&D project selection, a more common resource allocation activity
is the periodic allocation of an organization’s basic budget. Here, the
alternatives are not which departments to fund, but instead, at what level

5
The choice of procurement rules is itself a multi-objective decision that should be addressed before the actual
procurement process begins.
Chapter 8—Resource Allocation 241

should each department be funded. Each department, or organizational

unit, can design their operation at alternative levels of funding, e.g., equal to
last years funding, 10% above last years funding level, 10% below last years
funding level, 20% above, etc. If an organization faces a budget cut, say
10%, for example, cutting each organizational unit by 10% (sometimes
called across the board cuts) may seem fair, but it is not competitive. It
stands to reason that in today’s fast changing world, in order to be
competitive, some units should be cut perhaps 50%, 80% or entirely, in
order to increase the budgets of other units by significant amounts. The
resource allocation methodology presented below makes this a practical
approach.

Identify and structure the organization’s goals and objectives

The main message of this book is that decisions must be made on the
basis of achievement of objectives. And so it is with resource allocation
decisions. As Bauer, Collar, Emilio, and Tang advise in The Silverlake
Project, “priorities must be determined on the basis of the enterprise’s
overall objectives. Resource decisions need to be made holistically, that is,
with their consequences to the entire enterprise and all its parts in mind.”
And so the entire enterprise’s goals and objectives must be addressed. This
may not seem easy to do in a large enterprise or organization; however, it
can be done the same way that large organizations are typically organized,
that is, hierarchically.
Most organizations already have a statement of goals and objectives
(sometimes organized as values, goals, and objectives). A good bet is that
they are already structured hierarchically. Where are they kept? In some
organizations they might be framed and hung on the walls to help remind
employees and inform customers of the organization’s values, goals and
objectives. In other organizations they might be in ‘the blue book’ or ‘the
black book’ that is referred to from time to time. It is important to study
these goals and objectives to see if they are ‘living’ or long forgotten soon
after they were drafted. If they are still living, and up to date—taking into
account the fast changes in business and society, (locally, nationally, and
globally), they will serve as the driving force behind the allocation of
242 Decision By Objectives

resources. Otherwise, they must be revised, brought back to life, and

continually examined to keep them current. Once people realize that the
enterprise’s overall objectives will serve as the basis for the allocation of
resources, there will be a keen interest in keeping them current and relevant!
The hierarchy of objectives, sub-objectives, sub-sub-objectives, and so
on, must be broad enough to encompass every existing or desired activity
that is part of the resource allocation process. If not, proponents of an
activity will not be able to show where and how much the activity can
contribute to the organization’s objectives. Once more, we need to be open
to iteration since it might not become obvious that specific objectives were
overlooked until the activities are rated in a subsequent step of the process.
Top level executives understand and can best make judgments about the
relative importance of the main organizational objectives, and possibly the
sub-objectives. They know very little about the detailed alternatives vying
for funding. Lower level management and operational personnel can best
make judgments about the relative importance of the lowest level sub-
objectives and about how much contribution each alternative contributes to
the lowest level sub-objectives. They often don’t appreciate or understand
top management’s strategic direction or change in direction. The systematic
approach presented below makes it possible to synthesize knowledge,
experience, and insights across many levels within a large organization,
something that has not been possible to do in the past. Without such a
synthesis, it is virtually impossible to achieve a rational, competitive
allocation of resources.

Prioritize the objectives and sub-objectives

The relative importance of the objectives and sub-objectives must be
established in order to make a rational allocation of resources. Neglecting to
do so is a mistake. Assuming that all the main objectives are equally
important is a mistake. The pairwise comparison process and team methods
discussed in this text provide a straightforward, informative, and reliable
way to prioritize the organization’s objectives. Remember, this is a
‘process’. After establishing priorities for the organization’s objectives, and
subsequently deriving a ‘preliminary’ allocation based on these priorities,
Chapter 8—Resource Allocation 243

the priorities and judgments that served to derive the priorities should be re-
examined and revised as necessary. The prioritization of the organization’s
objectives during the resource allocation process leads to another important
benefit. In top management’s quest for excellence and response to shifts in
direction brought about by changes in the environment and competitive
forces, what better way is there to convey their priorities to the organization
at large? If these priorities are not conveyed, it is almost inevitable that
individuals or departments which formerly provided valuable services to an
organization, will, because of ignorance of changes in the organization’s
primary objectives or their relative importance, someday be surprised to find
their contributions are no longer of value.

Measure Alternatives’ Contributions

Having prioritized the organization’s objectives and sub-objectives, the
next step is to evaluate how much each proposed activity (or each possible
level of funding for each activity) would contribute to each of the lowest
level sub-objectives. This could be done by a pairwise comparison process,
but because there will normally be many, possibly hundreds or thousands of
activities or levels of funding, the ratings approach is customarily used.

Find the Best Combination of Alternatives

After prioritizing the organization’s objectives and sub-objectives, and
rating the contribution of the competing activities to the lowest level
objectives, we have ratio scale measures of the relative contribution of each
alternative to the overall objectives of the organization. We claim that there
is no way to rationally allocate resources without such measures and suggest
that the derivation of such measures for a large, diverse organization without
the approach detailed here is almost impossible to accomplish! We will
next consider two different basic situations in which we seek to find the best
combination of alternatives for the allocation of resources. In the first
situation, which we will refer to as Discrete Alternative Resource
Allocation, each project/activity is a separate or discrete unit. (There may
be a variety of dependencies between these units.) For example, we might
be allocating resources to discrete proposals in response to an RFP. In the
244 Decision By Objectives

second situation, which we will refer to as Activity Level Resource

Allocation, each project or activity is represented at one or more levels of
funding but an additional constraint is that each project can be funded at
only one level. (Here too there may be a variety of dependencies between
projects,).6 For example, each department of an organization may design
several alternative levels of funding; the resource allocation methodology is
to find the levels for each department that produces the best overall results
for the organization. For each of these two situations we will look at two
different approaches, one that focuses on the maximization of benefit/cost
and the other that focuses on the maximization of benefit.

Discrete Alternative Resource Allocation

George Washington University Academic Computing Advisory

Committee
We will illustrate the allocation of resources to alternative
projects/activities with a rather small, but typical example. The Academic
Computing Advisory Committee of the George Washington University had
$15,000 available to fund proposals relating to the use of computers. After
publishing a short RFP, eleven proposals were received. The total amount
requested for all the proposals was $34,430. The committee, consisting of
nine faculty members had to decide which of the proposed projects to fund.
Can you imagine getting nine faculty members to agree?
After discussing ‘criteria’ or, what we prefer to call ‘objectives’, the
following Expert Choice model was developed:

6
After looking at these two situations, it will be easier to understand the general situation which is simply a
hybrid of the two.
Chapter 8—Resource Allocation 245

Evaluate ACAC Proposals

Abbreviation Definition
ADDVAL adding value to existing resources
ALTFUND Availability of alternative sources
EXTREME
FACULTY Contribution to Faculty
HW/SW Hardware / Software
LONGTERM Long term benefit
MODERATE
NO
PROBABLY
SIGNIFIC
SLIM
SOME
STUDENTS Contribution to students
TAD
YES

(Priorities shown are 'Local' -- relative to parent node.)

Figure 1 – Decision Hierarchy

Through a process of discussion and pairwise comparisons (the

committee used keypads to enter individual judgments, which were then
aggregated) the priorities of the objectives (shown in Figure 1) were
derived. (A complete discussion of group or team decision-making begins
on page 115). Since the committee’s focus was on computer hardware and
software, and since some of the proposals were only peripherally related to
hardware and software the HW/SW objective was judged to be the most
246 Decision By Objectives

important. Expected long-term benefits to the University were second most

important, followed by contribution to student and faculty activities
(learning, research etc.) The objective of adding value to existing resources
was next in importance, followed by the objective of funding projects for
which there was little likelihood of alternative funding sources. Note that
the priorities possess the ratio level of measurement. The priority of the
HW/SW objective is 2.99 times the priority of the second most important
objective, long-term benefits to the University. Had an ordinal scale of six
items been used, the most important objective would have a value of 6, the
next most important objective a value of 5, and the ratio would have been
only 6/5 or about 1.2.
Chapter 8—Resource Allocation 247

Evaluate ACAC Proposals

Abbreviation Definition
ADDVAL adding value to existing resources
ALTFUND Availability of alternative sources
EXTREME
FACULTY Contribution to Faculty
HW/SW Hardware / Software
LONGTERM Long term benefit
MODERATE
SIGNIFIC
SOME
STUDENTS Contribution to students
TAD

(Priorities shown are 'Local' -- relative to parent node.)

Figure 2 – Intensities

The intensities below each of the objectives were also prioritized with
pairwise comparisons. For example, a project that is judged to make an
extreme contribution to students will receive a priority for that contribution
of about 63 times that of a project that makes only a ‘tad’ of a contribution
to students, as can be seen in Figure 2. This is quite different than an
ordinal scale where the ratio would be 5 to 1 instead. The intensities and
their scale can be different under each of the objectives. The intensities for
the alternative funding sources is shown in Figure 4.
248 Decision By Objectives

Evaluate ACAC Proposals

Abbreviation Definition
ADDVAL adding value to existing resources
ALTFUND Availability of alternative sources
FACULTY Contribution to Faculty
HW/SW Hardware / Software
LONGTERM Long term benefit
PROBABLY
SLIM
STUDENTS Contribution to students
(Priorities shown are 'Local' -- relative to parent node.)

Figure 3 – Alternative Funding

After the objectives and intensities are prioritized, each of the alternative
projects is rated with respect to each of the lowest level objectives, as shown
in Figure 4.
Chapter 8—Resource Allocation 249

Figure 4 – Combined Ratings

The entries in the columns representing the objectives (STUDENTS,
FACULTY, etc.,) are either one of the intensities for that column (for which
we derived is a ratio scale value) if all nine faculty members agreed on the
rating, or a decimal value which represents the average of the intensities of
the nine faculty members. The values in the Total column are ratio scale
measures of the contribution each project is expected to make to the
organization’s objectives, or, as Bauer et. al assert:
“Before you decide how to budget, you’ve simply got to know what’s most important,
especially in reaching your organization’s overarching goals in a holistic, balanced way.”

In this simple example, there are only six ‘overarching goals’ or

objectives, each represented by a column in the matrix. In a typical large
organization there may be hundreds of columns, structured hierarchically.
What does the total value .684 for Proposal One in Figure 4 mean? By
itself nothing, but it means a great deal when compared to the total values
for the other alternatives—Proposal One contributes about 8.7 times as
much as Proposal Six, for example. Because the totals are ratio scale
numbers, we can normalize without changing the ratios. The totals in Figure
4 are normalized such that an ‘ideal’ alternative, one that rated best in every
column would have a value of 1.0. Figure 5 shows the two other
normalizations: one normalized so that the best alternative is 100%, and
another normalized so that the priorities add to 1.0. Any of these three
250 Decision By Objectives

Figure 5 – Percent of Maximum

normalizations of the projects’ expected7 total benefit could be used in
determining the best allocation of resources. How the numbers are used
depends on the circumstances and constraints of the situation. If, for
example, each project could be partially funded and would produce benefits
in proportion to the amount funded, we might decide to allocate the $15,000
to the projects on the basis of the normalized priority column, so that Project
One would get 10.9%, Project Six 1.3% and so on. However, the
assumptions here are that each project requires a given amount of funding
(expressed as Costs in Figure 4) in order to deliver the benefits in the total
column—that is, we can not fund a project at some fraction of its stated
cost8. How then should we decide which projects to fund? We will look at
two approaches, depending on whether we are interested in achieving the
highest benefit/cost ratio or the highest benefit.

7
We refer to the benefits as ‘expected’ because the ratings were made on the basis of what the projects would
be expected to contribute. A more formal ‘expected’ contribution could be derived by including scenarios
between the goal and top level objectives in the AHP model and making judgments about the relative
likelihood of each of the scenarios.
8
Mechanically, it is just as easy, perhaps even easier, to perform the resource allocation allowing for partial
funding of projects.
Chapter 8—Resource Allocation 251

Figure 6 – Benefit Cost Ratios

Benefit/Cost Ratios—Sort and Allocate

Given the benefits and the costs, we can easily calculate the benefit/cost
ratios, and sort from high to low as shown in Figure 6.9 We can then
allocate funds to the projects starting with the project with the highest
benefit/cost ratio and continuing until the $15,000 budget is used up. This
will produce the largest benefit/cost ratio while funding as many projects as
possible within the budget constraint. Looking at the cumulative cost
column, we see that if we were to do this, we would fund Projects 8, 4, 3, 2,
1, 7 and 10. Projects 9, 11, 6 and 5 would not be funded.
Figure 7 contains a benefit/cost efficient frontier graph produced with
Team Expert Choice. Notice that the curve is concave, signifying
diminishing marginal benefit as additional projects are selected with lower
benefit/cost ratios.

9
The b/c ratios have been multiplied by 106.
252 Decision By Objectives

Figure 7 – Efficient Frontier

Before adopting this methodology, we should see when it makes sense

to seek the highest cumulative benefit/cost ratio as our objective and when it
doesn’t. If we assume that an organization will consistently follow the
philosophy of maximizing its cumulative benefit/cost ratio, then over time,
they will also maximize their benefits10. This statement also assumes that
there will be other comparable opportunities for additional resource
allocations in the future, so whatever portion of the budget that is not
expended will provide benefits commensurate with those activities currently
being funded. However, these assumptions are not always valid and it is
dangerous to take the maximization of benefit/cost ratios for granted. An
example illustrating the danger of using a benefit/cost maximization will be
presented after looking at the benefit optimization methodology.

10
As compared to any other allocation that produces a lower benefit/cost ratio and, in the long run, allocates
approximately the same total amount.
Chapter 8—Resource Allocation 253

Maximizing Benefits -- Optimization

Another method for allocating resources seeks to find that combination
of projects or activities that maximizes the total benefits without exceeding
the given budget.11 This problem can be formulated as a zero-one integer
mathematical programming problem, sometimes referred to as a knapsack
problem. Most spreadsheet programs today contain algorithms for solving
such problems.12 We will illustrate the solution to the preceding resource
allocation exercise using the Solver tool in Microsoft’s Excel.
Mathematically, the optimization problem formulation is as follows:
Maximize .109X1 + .096X2 + .119X3 + ...... + .103X11 (Benefits)
subject to: 2000X1 + 1622X2 + 1515X3 + ...... + 5000X11<= 15000 (Cost)
where: X1, X2, X3, ...... X11 >=0
and X1, X2, X3, ...... X11 <=1
and X1, X2, X3, ...... X11 are integer.

The following steps will setup and solve the problem with Expert
Choice and Microsoft’s Excel:

I-a) From Expert Choice’s ratings module, select the first three columns:
(Alternatives, Priorities, Costs).
Then do an Edit Copy.
I-b) Start Excel and do an Edit Paste to the Excel spreadsheet.
I-c) Add a column with the heading DV’s for the decision variables.
The decision variables will be adjusted by the algorithm to be either 0, if a
project is not to be funded, or a 1 if the project is to be funded. Put the value
of 1 in each element of this column (for illustrative purposes only).
I-d) Add another column with the heading F.Benefits (for funded
benefits). Enter a formula in the row corresponding to the first alternative
in this column that multiplies the decision variable cell by the priority cell

11
Compared to the ambiguity of what maximizing the cumulative benefit/cost ratio really accomplishes , the
objective of maximizing benefits is straightforward, understandable, and doesn’t depend on questionable
underlying assumptions.
12
Problems involving hundreds of alternatives might require special purpose optimization programs.
254 Decision By Objectives

for this alternative. If, for example, the decision variable is in cell D3 and
the priority in cell B3, the formula would be:
=D3*B3
I-e) Copy this formula down to the other cells in the F.Benefits column.
I-f) Add another column with the heading F.Costs (for funded Costs).
Enter a formula in the row corresponding to the first alternative in this
column that multiplies the decision variable cell by the cost cell for the first
alternative. If, for example, the decision variable is in cell D3 and the cost
in cell C3, the formula would be:
=D3*C3
I-g) Copy this formula down to the other cells in the F.Costs column.
II-a) Select the cell below the last alternative row in the F.Benefits
column and click the Summation symbol on the Excel toolbar, then press
Enter. Repeat this step for the F. Costs column.
II-b) Select the range of cells in the Decision Variable Column (all
should be 1 right now). Do an Insert Name Define and enter DVS (for
Decision Variables).
II-c) Move to the Total F.Benefits Cell.

The spreadsheet should look something like Figure 8.

Figure 8 – Funding All Alternatives

Chapter 8—Resource Allocation 255

If every project were funded, the Total Funded Benefit would be about 1
(shown as .999) and the Total Funded Cost would be $34,430. This is
obviously infeasible.
II-d) Click on the Tools menu.

If Solver is not listed as a choice, select AddIns and select Solver-Add In.

Figure 9 – Specifying Solver Parameters

Select Solver under the Tools menu. A Solver Parameters dialog box (see
Figure 9) should appear.

Parameters must be set for three things: The Target Cell, The Decision
Variables (By Changing Cells), and the Constraints.
III-a) The target cell is the cell we wish to maximize, in this case cell E-
14, the total funded benefit. Since we pre-positioned to this cell, E-14
should already be selected. The Max button should also be selected by
default.
III-b) From the Solver Parameter dialog box, type DVS (the name for
the decision variables range) in the By Changing Cells: box.
III-c) The constraints are added next, as follows:
256 Decision By Objectives

Figure 10 – Constraint Dialog Box

Figure 11 – Decision Variables <=1

Click the Add button and the Add Constraint dialog box (Figure 10) will
appear:

1) Type DVS in the left box (Cell Reference); select <= in the middle
box; and type 1 in the right box.
Then click Add to add another constraint.
2) Type DVS in the left box; select >= in the middle box; and type 0 in
the right box (see Figure 11). Then click Add to add a third constraint.

3) Type DVS in the left box; select INT in the middle box; and then
click Add to add a fourth constraint. (Note: Later versions of Solver have a
‘Binary’ option which can be used instead of steps 1-3 to specify that the
decision variables must be integer values of 0 or 1).
Chapter 8—Resource Allocation 257

4) With the left box active, click on the Total F.Costs in the
Spreadsheet; select <= in the middle box; and type 15000 in the right box.13

Figure 12 – Specified Parameters

Then press OK. The Parameters should look like Figure 12.
IV-a) Click on Options, select Assume Linear, then OK.
IV-b) Click on Solve. (If a message box appears stating the maximum
iteration limit was reached, select Continue). When the Solve Results box
appears, click the OK button.
The Spreadsheet should look like the Figure 13.

13
Instead of entering $15000 as a constant, it can be entered into a cell which is referenced in right hand side
of the constraint dialog box.
258 Decision By Objectives

Figure 13 – Solver Results

The results are exactly the same as those when we sorted by benefit/cost
ratio (which you can easily do in this spreadsheet as well). Is this true in
general? Usually, but not always. It will be true if, the available funding (in
this case $15,000) is exactly equal to one of the values in the cumulative
cost column of the sorted benefit/cost display. If this is not the case, the
organization might be better off (realize a greater total benefit) by replacing
one or more higher benefit/cost ratio projects with two or more lower
benefit/cost ratio projects.
Chapter 8—Resource Allocation 259

We can perform the above optimization for different values of available

funding and plot the results. To do this, we first add a cell to the Excel
spreadsheet representing the limit on available funding and modify the
constraint to reference this cell instead of the $15000 as we did earlier. See
Figure 14 and Figure 15.
If we first enter a very large limit for the funds available, we get the

Figure 14 – Cell With Funding Limit

solution shown in Figure 16 in which all proposals are funded for a total
cost of $34,430. Next, changing the amount available to be 1 dollar less
than 34,430, we get the solution shown in Figure 17 in which all but
Proposal five is funded for a total cost of $26,930.

Figure 15 – Constraint for Funding Limit

260 Decision By Objectives

Figure 16 – Solution for Very Large Limit

Figure 17 – Solution for Limit of 34,429

If we continue to change the amount available to be one dollar less than

the total cost of the previous solution, we will generate the series of points
shown in Figure 18. Given a level of available funding, the best solution is
that point at, or to the left of the available funding level.
Chapter 8—Resource Allocation 261

Figure 18 – Optimal Benefits vs. Available Funding

The yellow points in Figure 18 correspond to the points on the Benefit/Cost
efficient frontier in Figure 7. The points in benefit optimization plot that do
not correspond to any points in benefit/cost efficient frontier will be
‘hidden’ when making a resource allocation using benefit/cost. This may or
may not be significant. Suppose, for example, $1300 of funding was
available. The benefit maximization method yields a benefit of .122 at a
cost of $1280, while the benefit/cost method yields a benefit of .092 at a
cost of $565. Which solution is better, the solution produced by the method
that maximizes the cumulative benefit/cost ratio or the solution produced by
the method that maximizes total benefit? Is the 56% increase in costs
262 Decision By Objectives

justified by the 25% increase in benefits? If the amount expended doesn’t

matter14, then any increase in total benefit is an improvement and the
solution that maximizes total benefit is better. However, the benefit/cost
ratio of the maximum benefit solution is smaller, and this may seem ‘sub-
optimal’ to those who have been taught to maximize the benefit/cost ratio.
The danger of maximizing the benefit/cost ratio when benefit is more
appropriate is well illustrated with the following example.
Consider a situation where you can invest in either of two business
opportunities: (A) you can invest one dollar and get ten dollars back, and
(B) you can invest 50 dollars and get 100 dollars back. However, you have
only 50 dollars to invest. Both the benefit/cost solutions and the maximum

Figure 19 – Optimization vs. Benefit/Cost

14
As an aside, there are organizations that prefer to spend as much of their budget as possible to prevent
having the budget decreased the following year.
Chapter 8—Resource Allocation 263

benefit solutions as a function of the amount to invest are plotted in Figure

19. While all three solutions points are achieved following a maximum of
benefits, only the two points shown in yellow are achieved following a
maximization of benefit/cost. In particular, if 50 dollars is available for
investing, the maximization of benefit solution is to invest $50
(recommending that $50 be invested for the $100 return), while the nearest
point in yellow to the $50 for the maximization of benefit/cost is at $1
(suggesting that $1 be invested for a $10 return).
A variation of this situation is that there is no limit on available funding, but
you must choose either (A) or (B), but not both. If you had to make (and
possibly justify) the decision based on benefit/cost ratios, you would choose
the first proposition, because a benefit/cost ratio of 10 is better than a
benefit/cost ratio of two. However, the maximization of benefit solution
(shown in Figure 20) is to choose the second proposition wherein you

Figure 20 – Optimum is Large Investment

would walk away with a 50-dollar profit rather than a 9-dollar profit.
Clearly, he benefit/cost ratio approach produces the wrong answer. In
general, whenever unexpended funds from the resource allocation are not
relevant, the benefit/cost ratio approach is not applicable.
Another example of the inappropriateness of maximizing the
benefit/cost ratio is related to the knapsack problem designation given to this
type of optimization. Suppose a camper is planning to take a knapsack of
books on a trip. The knapsack can only hold a certain volume of books and
264 Decision By Objectives

each book has a different expected benefit15 to the camper. Which books
should the camper take so as to maximize the total benefit? There will
likely be some slack or unused volume in the knapsack., but slack is not one
of the campers objectives. The camper ‘s objective is clearly to maximize
the value of the books. The ratio of the total benefits of the selected books
to the amount of volume used is really not of any significance. If, after
tentatively selecting the books, the knapsack were replaced with one a bit
larger, the camper might be able to remove a previously selected book and
replace it with a larger book that has more value but a lower benefit/volume
ratio. This would increase the campers overall objective, maximizing
benefit. The increase in occupied volume from replacing the smaller book
by the larger is of no consequence so long as the new volume constraint is
not violated.
The problem can be made more complex by introducing other factors
such as weight and /or cost. These factors can be considered as objectives
and/or constraints. For example, suppose the camper had a budget of $X for
books. This constraint can easily be added to the mathematical
programming problem along with the volume constraint and an optimal
solution (combination of books) can be found. This optimal will maximize
the total value of the books without exceeding either the volume or cost
constraints.
Would cost also be an objective in addition to a constraint? It depends
on the situation. If the camper were deciding which books to take from his
existing library, cost would not be an objective. If the camper were told he
could spend up to but no more than $X to purchase books, then cost would
not be an objective. If the camper were given $X from which he could
purchase books, with any remaining funds retained by the camper, then cost
would be an objective.
The assumptions necessary to formulate the resource allocation to fit a
real world situation can best be identified by focusing on objectives. In
particular, answering the question of whether cost is an objective as well as
a constraint can best be answered by asking if cost, or the unexpended funds
of the resource allocation, have significance to the decision makers. The

15
The expected benefits can be derived with an AHP model.
Chapter 8—Resource Allocation 265

following are three examples where cost (or unexpended funds) would not
be considered as an objective:
Military campaign: In a military campaign, the objective is to get
maximum military effectiveness; dollars not spent are virtually meaningless.
Would military decision-makers be concerned with unexpended money if
that would in any way decrease their effectiveness (even though the overall
benefit/cost ratio is decreased)?
Small business competition: Suppose a small business in a competitive
environment has extended its credit to limit. The managers feel they need
to be as competitive as possible. Any increase in competitiveness is worth
the extra cost, provided money is available. Having funds unexpended is
really not part of the objective.
Government department’s budget: money not spent can’t be spent the
next year16.

Cost as a constraint and/or an objective:

When allocating resources, it is usually obvious when cost is a
constraint – e.g., an organization is constrained to work within a given
budget (although they can often try to justify increases to the budget). It is
not so obvious, when allocating resources, if cost is also an objective.
What is cost anyway? A primitive society without money will engage
in trade with barter. An advanced society has markets that, in equilibrium,
equate cost with value. Two people trading hand made goods in a primitive
society might take the following factors (or objectives) into account when
contemplating a trade:
1. production time
2. value of component parts
3. tangible benefits
4. intangible benefits (e.g., artistic value)
5. value for future trades (e.g., have monetary value, such as art).

16
An interesting twist to this situation: not only is any money not spent not carried over to next year, but the
under-spending is likely to result in a decrease in next year’s budget. Now cost again becomes an objective,
but the objective is to spend as much as possible!
266 Decision By Objectives

In a sense, the value (or cost) of something being traded is a synthesis of

objectives. More advanced societies attempt to translate this value into
dollars with monetary and market systems. But there may not be a market
to place a value on a proposed new weapon system. Or, even if a market
does exist, the market value may not represent the value to an individual or
organization. Hence the need for organizations to be able to synthesize
multiple objectives when allocating resources. It does not necessarily
follow that the ‘market value’ or cost of an alternative is commensurate with
the value of that alternative to an organization.

How to proceed if costs (or unexpended funds) are an objective as well as

a constraint:
Cost can be included in the hierarchy so that the total benefit of each
alternative includes the objective of low cost. (Judgments would be made
comparing the relative importance of cost and the other factors, e.g.,
author’s reputation, interest in the subject of the book, etc.; intensities
defined and prioritized for the price objective, and ratings made for the cost
of each book.) This is analogous to including cost along with other
objectives such as style, comfort, and performance when choosing a car.
Don’t include cost in the hierarchy (because sometimes people find it
easier to make comparisons about benefits without having to consider
cost17. Instead, optimize over a range of total budget constraint values,
plotting the maximum benefit vs. cost for each of the solutions. Such a plot
is called an “efficient frontier”. The decision-maker examines this efficient
frontier and decides which point is most desirable. This approach is
preferred by those who feel that decision-makers sometimes find it easier to
make judgments about objectives without considering costs.
Still another way to address cost as an objective is to add the percentage
of unused budget as a variable in the objective function. The coefficient of
this variable must either be estimated (which may be difficult to do) or
treated as a parameter.

17
Although when selecting a car, people don’t seem to have any difficulty with low price as one of the
objectives.
Chapter 8—Resource Allocation 267

Flexibility of Benefits Optimization Approach:

The benefits optimization approach to resource allocation is extremely
flexible. Constraints can be added to fit almost any managerial need. For
example, The Air Force Cost Center added constraints to guarantee that
each division received some minimum amount of funding. The Northeast
Fisheries Center added constraints to regulate the rate of change in budget
cutbacks so that members of the organization would not feel that the rug was
about to be pulled out. They added a constraint for each division that
guaranteed that the division total funding would be at least X% of the prior
years funding, and solved the optimization of benefits for different values of
X. The larger the value of X, the smaller the maximum will be. The smaller
the value of X, the more flexibility there is to increase funding in areas that
contribute more to the organization’s objectives. Management then
examined the plot of total benefit versus guarantee percent to determine
what the percent should be and implemented the corresponding resource
allocation. Resource allocation must be a dynamic, iterative process. By
examining intermediate results, and modifying either or both of the AHP
hierarchy or optimization formulation, an organization can express its
unique musts and wants to find an allocation of resources that is rational and
makes sense.

Constraints for dependencies

Two mutually exclusive activities i and j can be represented by a
constraint such as:

Xi + Xj <=1

This can easily be extended to three or more mutually exclusive activities:

Xi + Xj +Xk + …<=1

A constraint to specify that one from a set of activities must be selected can
be modeled by including a mutually exclusive constraint as above, plus a
constraint of the form:
268 Decision By Objectives

Xi + Xj +Xk + …=1

Constraints representing Synergy

Saaty and Peniwati18 have illustrated how constraints can be used to
represent both positive and negative synergy. Consider the formulation
presented above:
Maximize .109X1 + .096X2 + .119X3 + ...... + .103X11 (Benefits)
subject to: 2000X1 + 1622X2 + 1515X3 + ...... + 5000X11<= 15000 (Cost)
where: X1, X2, X3, ...... X11 >=0
and X1, X2, X3, ...... X11 <=1
and X1, X2, X3, ...... X11 are integer.

Suppose a group of activities (for example A1 and A2) taken together

contribute more than the sum of the activities taken separately. Define the
combined activity A12 to be A1 and A2 together. Suppose this activity were
added to the Expert Choice model and the derived priority is .250. This
would represent a positive synergy since individually activities A1 and A2
contribute .109 + .096 = .205. A negative synergy, perhaps due to overlap
of contribution, would be indicated by a priority for A12 of less than .205.
The optimization can be formulated by adding a decision variable, call it
Z1,2, where a resulting value of Z1,2 =1 means that the combination of A1 and
A2 is to be included. However we must add additional constraints to assure
what Saaty refers to exclusion and duplication.
Exclusion means that if an activity is performed exclusively (alone), no
other combination containing the activity can be performed. Thus, for each
activity that is a member of a combination, the decision variable
corresponding to the activity performed alone plus all other decision
variables representing the combinations does not exceed 1.
Duplication assures that for each combination, there is no other
combination of variables that duplicates the combination.

18
Saaty, T.L. and Peniwati, K. “The Analytic Hierarchy Process and Linear Programming in Human
Resource Allocation”, Proceedings of the Fourth International AHP Symposium, Vancouver, 1996.
Chapter 8—Resource Allocation 269

Thus, for the previous example, we would have:

Maximize .109X1 + .096X2 + .119X3 + ...... + .103X11 + .250 Z1,2 (Benefits)

subject to:

2000X1 + 1622X2 + 1515X3 + ...... + 5000X11 + 3622 Z1,2<= 15000 (Cost)

and constraints for exclusivity:

X1 + Z1,2 <=1 for activity 1 with combination Z1,2

X2 + Z1,2 <= 1 for activity 2 with combination Z1,2

and a constraint to prevent duplication:

X1 + X2 <= 1
to prevent duplication of combination Z1,2.
where: the decision variables are non negative integers:
X1, X2, X3, ...... Z1,2 >=0
X1, X2, X3, ...... Z1,2 <=1
X1, X2, X3, ...... Z1,2 are integer.
270 Decision By Objectives

Another, perhaps simpler representation of constraints representing

Synergy
Identify each set of activities with synergistic effect (either positive or
negative). For each set, define up to 2n -1 decision variables representing all
meaningful combinations of the activities. The cost and benefit of each
possible combination represent the synergistic cost and benefit respectively.
Only one constraint must be added for each set of synergistic activities – the
sum of the 2n variables must = 1. This approach results in more decision
variables but fewer constraints. To apply this technique to the example
above, lets represent the combinations of activity 1 and activity 2 by the four
decision variables Z0,0, Z0,1, Z1,0, and Z1,1, where the first subscript refers to
activity 1, the second subscript refers to activity 2, and a 0 for the subscript
means that the activity is not included while a 1 means that it is included.
So, for example, if Z1,1 is 1, then both activity 1 and activity 2 are to be
undertaken. The formulation using these variables would be:
Maximize .109Z1,0 + .096Z0,1 + .250 Z1,1 + .119X3 + ...... + .103X11
(Benefits)
subject to:
2000 Z1,0 + 1622 Z0,1 + 3622 Z1,1 + 1515X3 + ...... + 5000X11 <= 15000
(Cost)
and the additional constraint:
Z0,1 + Z1,0 + Z1,1 <=1
where the decision variables are non-negative integers:
Z0,1, Z1,0, Z1,1, X3, …. X11 >=0
Z0,1, Z1,0, Z1,1, X3, …. X11 <=1
Z0,1, Z1,0, Z1,1, X3, …. X11 are integer.

A more elaborate example

Suppose, as a more elaborate example, an additional synergistic
combination of A2 and A3 produces a combination represented by decision
variable Z2,3, with a benefit of .275. Furthermore, an additional synergistic
combination of A1, A2 and A3 produces a combination represented by
decision variable Z1,2,3 with a benefit of .400.
Chapter 8—Resource Allocation 271

Maximize .109X1 + .096X2 + .119X3 + ...... + .103X11 + .250 X12 + .275 Z1,3 + .400
Z1,2,3 (Benefits)

subject to:

2000X1 + 1622X2 + 1515X3 + ..... + 5000X11 + 3622 X12 + 3137 Z1,3 + 5137
Z1,2,3<= 15000 (Cost)

and constraints for exclusivity:

X1 + Z1,3 <=1 for activity 1 with combination X12

X1 + Z1,2,3 <=1 for activity 1 with combination X14
X2 + Z1,3 <=1 for activity 2 with combination X13
X2 + Z1,2,3 <= 1 for activity 2 with combination X14
and a constraints to prevent duplication:
X1 + X2 <= 1 to prevent duplication of combination X12.
X2 + X3 <= 1 to prevent duplication of combination X13.
X1 + X2+ X3 <= 1 to prevent duplication of combination X14.
X12 + X3 <= 1 to prevent duplication of combination X14.
Z1,3 + X1 <= 1 to prevent duplication of combination X14.

where the decision variables are non-negative integers:

X1, X2, X3, ...... , Z1,3, Z1,2,3 >=0
X1, X2, X3, ......, Z1,3 , Z1,2,3 <=1
X1, X2, X3, ...... Z1,3 , Z1,2,3 are integer.

The alternative formulation for this more elaborate example would be:
Define 23 -1 or seven decision variables representing the combinations of
activities 1, 2 and 3, where Z1,0,0 represents doing activity 1 but not two or
three, Z 0,1,0 represents doing activity 2 but not 1 or 3, Z 1,1,0 represents doing
activities 1 and 2, but not 3, and so on. The formulation would be:

Maximize .109Z1,0,0 + .096Z0,1,0 + .119Z0,0,1 + .250 Z1,1,0 + .275 Z1,0,1 + .400 Z1,1,1...... + .103X11
subject to:
272 Decision By Objectives

2000 Z1,0,0 + 1622 Z0,1,0 + 1515 Z0,0,1 +3622 Z1,1,0 + 3137 Z1,0,1 + 5137 Z1,1,1 ..... + 5000X11 <=
15000 (Cost)

Z1,0,0 + Z0,1,0 + Z0,0,1 + Z1,1,0 + Z1,0,1 + Z0,1,1 + Z1,1,1 <=1.

where the decision variables are non-negative integers:

Z1,0,0, Z0,1,0,…Z1,1,1, X4,…X11 >=0
Z1,0,0, Z0,1,0,…Z1,1,1, X4,…X11<=1
Z1,0,0, Z0,1,0,…Z1,1,1, X4,…X11 are integer.

An example with two sets of synergistic activities.

Suppose activities A1, …. A100 were being considered. Further suppose
there are two sets of synergistic activities, the first set including activities
A1, A2, A3, and A4 and a second set including activities A8, A9, and A10.
For the first set, suppose the following combinations of A1 through A4
are possible: each activity by itself, or the following synergistic
combinations (either positive or negative synergy):
A1 with A2
A2 with A3 with A4

For this first set, define a subset of 24 - 1 or 15 variables:

Z1,0,0,0, which, if set to 1, signifies A1, but not A2, A3, or A4
Z0,1,0,0, which, if set to 1, signifies A2, but not A1, A3, or A4
Z0,0,1,0, which, if set to 1, signifies A3, but not A1, A2, or A4
Z0,0,0,1, which, if set to 1, signifies A4, but not A1, A2, or A3
Z1,1,0,0, which, if set to 1, signifies A1 and A2 but not A3, or A4
Z0,1,1,1, which, if set to 1, signifies A2, A3, A4, but not A1

• These variables are included in the objective function and functional

constraints with the appropriate benefit and cost coefficients. Additionally,
a constraint:
Z1,0,0,0 + Z0,1,0,0 + Z0,0,1,0 + Z0,0,0,1 + Z1,1,0,0 + Z0,1,1,1 <=1 insures that only one of the combinations will be
selected.
Chapter 8—Resource Allocation 273

For the second set, suppose the following combinations of A8 through A10
are possible: each activity by itself, or the following synergistic
combinations (either positive or negative synergy):
A8 with A9 with A10
A9 with A10.

For this first set, define a subset of 23 - 1 or 7 variables:

W1,0,0, which, if set to 1, signifies A8, but not A9, or A10
W0,1,0, which, if set to 1, signifies A9, but not A8, or A10
W0,0,1, which, if set to 1, signifies A10, but not A8, or A9
W1,1,1, which, if set to 1, signifies A8, with A9 and A10
W0,1,1, which, if set to 1, signifies A9 with A10 but not A8.

These variables are also included in the objective function and functional
constraints with the appropriate benefit and cost coefficients. Additionally,
a constraint:
W1,0,0 + W0,1,0 + W0,0,1 + W1,1,1 + W0,1,1 <=1 insures that only one of the combinations will be
selected.

Constraints for different types of ‘costs’

Additional constraints can easily be added for other ‘costs’. For
example there may be one constraint for dollar costs, another for personnel,
and still another for space. Each such constraint can also be specified for
multiple time periods.

Orders of Magnitude Considerations

Care must be taken when allocating resources to alternatives that differ
in cost by several orders of magnitude. If for example, the cost for some
alternatives are in the 1-10 thousand dollar range while the cost for others
are in the hundreds of millions of dollars, then the benefit measure must be
capable of spanning a commensurate range ratio. This would not be the case
with a narrow hierarchy (one that doesn’t encompass a wide range of
objectives) or intensities that do not span ranges that can adequately reflect
the difference in contribution of very inexpensive alternatives and very
expensive alternatives. One approach, then, is to construct an evaluation
274 Decision By Objectives

hierarchy that is both broad -- so that expensive alternatives can be

recognized for their contributions over a broad range of sub-objectives while
inexpensive alternatives will be recognized only for the more narrow areas
to which they contribute. The evaluation hierarchy should also contain
numerous rating intensities so as to span as wide a range of measure as
possible, perhaps even two orders of magnitude.

It may be difficult or impractical to do this in some circumstances, so

another approach, and one that is similar to the traditional way organizations
allocate funds hierarchically, is to divide the alternatives into categories
from low cost to high cost and allocate separately within each category. For
example, if projects are being considered that range in cost from 5 thousand
to 8 million dollars, the projects can be organized into categories 1-10K,
10K-100K, 100K-1M and 1M-8M. The question then arises as to how much
of the total budget is to be allocated to each category. This can be addressed
various ways. One would be with a hierarchy to derive priorities for each
category’s contribution to the organizations objectives. The total funding
can then be multiplied by the resultant priorities to determine how much will
be allocated within each category. Another way would be to allocate
funding to each category based on the ratio of the total cost of the projects in
each category to the total costs of all projects, and multiply the funding
amount in each category by the ratio of the total funding available to the
total cost of all projects.

Summary of B/C Ratios vs. Optimization of Benefits:

B/C Ratios:
Cost is an objective because it goes into B/C ratio.
If cost (or unexpended funds) is not an objective, then B/C ratio should not
be used.
examples: Military, Competitive business; government agency
Pros: Easy to Solve.
Cons: Can lead to wrong answers if applied blindly (assumptions difficult
to verify).
Chapter 8—Resource Allocation 275

Can only deal with one constraint.

Optimization of Benefits:
Cost may not be an objective, or
Cost is an objective. If so, can do one of the following:
1. include cost as an objective in the hierarchy, or
2. consider cost as a constraint and optimize benefits over a range of total
cost constraint values, plotting the maximum benefit vs. cost for each of the
solutions. Such a plot is called an “efficient frontier”. The decision maker
examines this efficient frontier and decides which point is most desirable; or
1. add percentage of unused budget as variable in optimization the objective
function.

Pros: Easy to verify assumptions

Can include many types of constraints
Very flexible
Cons: Computationally difficult
May have to explain why replacing a higher b/c alternative with a
more costly but lower b/c alternative as budget is increased, is in the
best interests of the organization.

Comparison of results from maximizing b/c ratio and maximizing

benefits:
• Results are often quite close
• Identical results if budget limit corresponds to a cumulative cost break
value
• Drastically different results are possible, especially in situations where
cost, or unexpected funds are not important.
276 Decision By Objectives

Activity Level Resource Allocation

Whereas in discrete alternative resource allocation, each alternative
(e.g., research proposal) is discrete from (although possibly related to) the
other alternatives, activity level resource allocation considers one or more
levels of funding for each alternative (e.g., department). For example, each
department in an organization can be considered for funding at current
funding levels, +/- 5%, +/- 10%, etc. A plan is developed specifying the
type and extent of activities each department would perform at each level of
funding. An AHP model is used to evaluate the expected contribution to the
organization’s objectives for each alternative level of funding. A prototype
of an activity level resource allocation problem will be used to illustrate a
heuristic incremental benefit/cost approach as well as a maximization of
total benefit approach.
Consider an organization with four departments, Marketing, Production,
Operations, and Maintenance. Each department submits plans for how they
would operate with a bare minimum number of people, with the current
number of people, and with a moderate increase from current funding level.
The organization has six major missions, which we will refer to as Mission
A through Mission F. To keep the example simple, we will assume that the
objectives within each mission include responding to day to day needs and
activities, total quality efforts, and long term development. An Expert
Choice model with mission priorities derived from pairwise comparisons is
shown in Figure 21.
Intensities were defined for rating how much each alternative level of
funding for each department would contribute to each of the objectives
within each mission. A portion of the RATINGS model is shown in Figure
22. Notice that each ‘line item’ is a level of funding within a department.
Departments are identified by a common alphanumeric prefix before the
delimiter, in this case a colon. The costs are shown in thousands of dollars.
The Total column represents the un-normalized total expected contribution
of that line item (department at a specified level of funding) toward the
organization’s overall objectives. Objectives for which no contribution is
expected at a specified level of funding are left blank.
Chapter 8—Resource Allocation 277

Personnel Allocation

Abbreviation Definition
DAY2DAY Responding to day to day needs and activities
L.T.DEV Long term development
MISSIONA
MISSIONB
MISSIONC
MISSIOND
MISSIONF
MISSONE
TQM Total quality efforts
(Priorities shown are 'Local' -- relative to parent node.)

Figure 21 – Organizational Objectives

Figure 22 – Activity Level Ratings

278 Decision By Objectives

Benefit/Cost Ratios
Because only one level can be selected for each department, the
maximization of cumulative benefit/cost ratios is not nearly as
straightforward as for the discrete resource allocation situation. The
following heuristic approach has been found to provide a reasonably good
allocation. We begin by funding each department at its lowest level of
funding, thereby producing a combination with the lowest total benefit and
lowest total cost. We look for the best choice of department to increase to
its next level of funding. We make the choice by choosing the department
with the highest incremental benefit/cost ratio from its present level of
funding. We continue doing this until we exceed the total budget, and then
go back to the last level of increase.
This method will not necessarily result in the maximum benefit subject
to the total budget constraint, but it is likely to be quite close. One of the
practical difficulties in implementing the method occurs when a
department’s contribution fails to follow a diminishing marginal rate of
return curve. Theoretically, the law of diminishing returns states that the
incremental benefit from an increase in one unit of cost should be a non-
increasing function. If this assumption is violated, then a relatively low
incremental benefit/cost ratio at some level of funding can ‘hide’ a higher
incremental benefit/cost ratio from being seen by the algorithm. To avoid
this difficulty, a recursive procedure is used to remove the lower
incremental benefit/cost level(s) from consideration.
The initial lowest cost allocation consists of the lowest levels for each
department as shown in Figure 23, Figure 24, and Figure 25.
The values in the Benefit column in Figure 23 are normalized
equivalents of the total values in Figure 22 – normalized so that, if
each department were funded at the maximum level, the total would be 100.
The >>> symbol in Figure 25 suggests that if additional funds beyond the
$535K are to be allocated, then the Maintenance department should be
increased to a higher level of funding.
Chapter 8—Resource Allocation 279

Figure 23 – Each Activity At Lowest Level

Figure 24 – Efficient Frontier

280 Decision By Objectives

Figure 25–More Maintenance If More Funding

Suppose we didn’t know this, but asked the Department managers to
meet and decide. Suppose they decided to increase the Operations
Department to four people. The results would be as shown in Figure 26 and
Figure 27. The total benefit has increased from 11.57 to 22.7919 and the
total cost has increased from $535K to $610K.

Figure 26 – Seat of the Pants Decision

19
Since the benefit scale is a ratio level of measure, we can say that this is about a 100% increase in total
benefit.
Chapter 8—Resource Allocation 281

Figure 27 – Below the Efficient Frontier

Figure 28 – Incremental Benefit Cost Allocation

282 Decision By Objectives

But if we followed the algorithm, we would have increased funding in

the Maintenance Department from 3 people to 6 people20. The results are
shown in Figure 28, Figure 29, and Figure 30.

Figure 29 – Back on the Efficient Frontier

Thus, increasing funding in the Maintenance Department from 3 people

to 6 people, rather than increasing the funding in the Marketing Department
from 1 person to 2 people, would result in almost double the benefit (45.64
vs. 22.79) for only a slightly extra cost ($625K vs. $610K). The >>>
symbols in Figure 30 indicate that the next increase should be in the
Marketing Department. Following the algorithm would result in the points
shown on the efficient frontier plot in Figure 29.

20
Staffing levels of 4 and 5 were skipped because the incremental benefit/cost for this department did not
follow the law of diminishing returns.
Chapter 8—Resource Allocation 283

Figure 30 – Change Indicators

Maximizing Benefits -- Optimization

The problem formulation for the maximization of benefit approach that
we used for the discrete alternative resource allocation extends easily to the
activity level resource allocation case. We want to find that combination of
activity levels that maximizes the total benefits without exceeding the given
budget.21 This problem, is formulated as a zero-one integer mathematical
programming problem, sometimes referred to as a multiple choice knapsack
problem. The formulation for the prototype activity level resource
allocation presented above (refer to Figure 22 and Figure 23) is as follows:
Maximize 2.648X1,1 + 16.595X1,2 + 3.014X2,1+ 12.602X2,2 + 19.712X2,3 +
+..... + 3.079X4,1 + 14.773X4,2 + 20.172X4,3 + 37.142X4,4
(Benefits)22

subject to: 80X1,1 + 160X1,2 + 140X2,1+ 210X2,2 + 280X2,3 +

21
Here, as in the discrete alternative resource allocation case, maximizing the cumulative benefit/cost ratio
really is more straightforward and understandable than the incremental benefit/cost approach. Furthermore, it
doesn’t depend on questionable underlying assumptions. However, the computation complexity, even with
high speed computers, is sometimes a concern.
22
The coefficients of the objective function correspond to the normalized benefit column of Figure 23
(normalized so that the total benefit is 100 if each department were funded at the maximum level).
Alternatively, the un-normalized values in the total column of Figure 22 could have been used.
284 Decision By Objectives

Figure 31–Activity Level Optimization

+..... + 90X4,1 + 120X4,2 + 150X4,3 + 180X4,4 <= Budget (Costs)

where: X1,1, X1,2, X2,1, ...... X4,4 >=0

X1,1 + X1,2 =1

X2,1 + X2,2 + X2,3 =1

X3,1 + X3,2 + X3,3 + X3.4 = 1

X4,1 + X4,2 + X4,3 + X4.4 = 1

and X1,1, X1,2, X2,1, ...... X4,4 <=1
and X1,1, X1,2, X2,1, ...... X4,4 are integer.
Chapter 8—Resource Allocation 285

contains an Excel spreadsheet in which the optimization is performed.

The values correspond to the optimal solution for an available budget of
$800K. Columns A, B, and C were copied and pasted from the Expert
Choice Ratings sheet. The level decision variables (designated as LDV’s)
are contained in Column D. Column E contains formulas for Funded
Benefits. For example, the formula in cell E3 is =D3*B3. Column F
contains formulas for Funded Costs. For example, the formula in cell F3 is
=D3*C3. Cell G16 contains the available funding, presently set at $800K.
Cells B19 through B22 each contain the sum of the decision variables for
the respective activity. For example, cell B19 contains the formula:
=SUM(D3:D4).
The solver parameters are shown in Figure 32. The target cell, E16, is
the sum of the funded benefits. The By Changing Cells are the level
decision variables. The five constraints are as follows:
1.B19:B22 = 1 insures that one and only one level will be funded for
each alternative.
2.F16<=G16 constrains total spending.
3.The level decision variables are less than or equal to 1.
4.The level decision variables are integer.
5.The level decision variables are non-negative.

Figure 32 – Optimization Parameters

286 Decision By Objectives

Figure 33 and Figure 34 show the resulting maximum benefit as a

function of cost. The levels indicate the level of funding for each of the four
activities (departments). For example, 2,1,1,4 in Figure 33 means that the
first department is funded at level 2, the second and third departments at
levels 1, and the fourth department at level 4. The yellow dots in Figure 34
correspond to the incremental benefit/cost solutions discussed above. Given
a level of available funding, the best solution is that point at, or to the left of
that funding level. Like the discrete alternative resource allocation funding
problems, the difference between the maximum benefit and maximum
benefit/cost solutions for activity level resource allocation problems may or
may not be significant. If the solution point at or to the left of a given level
of available funding is a solution point to both methods (a yellow dot) then
there is no difference. Otherwise the difference is the difference between
the benefit at the first maximum benefit solution point and the first
maximum benefit/cost solution point to the left of the available funding. For
example, if $650K of funding is available, there is no difference. However,
if $600K is available, maximizing benefits produces almost three times the
benefit than maximizing benefit/cost ratio (28.67 vs. 11.57).
Chapter 8—Resource Allocation 287

Figure 33 – Benefits, Costs, Activity Levels

Resource Allocation Summary

The need to synthesize information and judgments is a difficult, if not
insurmountable problem when allocating resources in medium and large size
organizations. High level executives can best judge the relative importance
of the organization’s objectives while operational and technical personnel
can best judge the performance of the alternatives with respect to these
objectives. Knowledge and judgments about both objectives and
alternatives are required in order to rationally allocate resources. One might
say that this is just a communications problem, but it is more than that. No
matter how many meetings and discussions are held, it is virtually
impossible to combine the requisite information and judgment without the
ability to properly measure and synthesize. The methodologies presented in
this chapter make this possible.
288 Decision By Objectives

The AHP process is used to derive priorities for the organization’s

objectives as well as the alternatives that compete for resources. Once
priorities are derived, either a benefit/cost or maximization of benefit
approach can be used. The former offers some computational benefits
whereas the latter is more robust and flexible. We have presented
methodologies for discrete alternative resource allocation as well as for
activity level resource allocation. The former is particularly useful when
considering new activities, such as proposals for research and development.
The latter can easily be applied in any organization that does periodic
resource allocation. It is particularly useful as an alternative to ‘across the
board cuts’ when an organization is faced with budget cutbacks23 since
‘across the board cuts’ will eventually lead to the demise of an organization
that faces competition in a changing environment. Instead, this rational

Figure 34 – Optimum Benefit Vs. Cost

23
Although resource allocations during stable or increasing budgets do not concern management as much as
cutbacks, the need to gain and maintain a competitive advantage should make this activity just as important as
when budgets are being reduced.
Chapter 8—Resource Allocation 289

resource allocation methodology will help identify those departments that

should receive increased funding even when the overall budget is being
decreased. The methodologies are extremely flexible and can be
implemented piecemeal -- yielding incremental improvements over the
organization’s current resource allocation “methodology”. Improvement
may come just from prioritizing the organization’s objectives, and / or from
evaluating the expected contributions at different levels of funding for each
department.
Implementation of the complete methodology, including the
combinatorial optimization problem will require iteration so that the
intermediate ‘solutions’ can be questioned, challenged, and modified by:
•changing the AHP model and / or
•changing judgments and / or
•adding or modifying constraints in the optimization.

The ‘final’ solution should not only be mathematically ‘optimal’, but

intuitively appealing to those who will have to live and compete with the
resulting levels of funding.