Beyond

Spend Analysis

Taking Procurement Analytics

to the Next Level Through
Predictive Analytics
About the Author

Michael Lamoureux, aka the doctor, is the Editor-in-Chief of Sourcing

Innovation (.com), a resource for sourcing, procurement, and supply chain
professionals who are interested in improving themselves and the overall
performance of their organizations. A regular contributor to Spend Matters,
he is a Computer Science PhD who has been heavily involved in the Sourcing
and Supply Chain Space since 2000 and the e-Commerce space since 1997.
As a freelance procurement consultant with extensive expertise in sourcing,
procurement, and supply chain processes, he aims to continually push
innovation in and beyond the supply chain space. With particular expertise
in analytics, modeling, and optimization, he is able to dive much deeper into
technology and core issues, striving to help businesses with their internal
knowledge transfer, positioning, and planning problems.

2
Part I

The Deficiencies of
Descriptive Analytics

3
 Part I: The Deficiencies of Descriptive Analytics

A
nalytics is the new hotness. But it’s not the same old
story that you might remember from twenty years
ago when Business Intelligence was all the rage or the
same old story from ten years ago when spend analysis
first hit the market big time. This is because modern analytics
systems have left the age of static reporting on stale OLAP
cubes and entered into an age of dynamic reporting on real-
time cubes. Why is this important?

To answer the question, we have to first explain

why descriptive analytics, once the revolutionary
offering of business intelligence systems, is
no longer sufficient in today’s fast-moving
economy. After all, if the technology was good
enough for almost a decade, why isn’t it good
enough today?

When they were first released, first generation Business

Intelligence tools like Cognos and Business Objects were
revolutionary. For the first time, an organization could build a
complete OLAP data cube on its spend data or performance
data, organized to its liking, and then create a cross-section of
views on this data that would allow it to get complete insight
into its spend, or a subset of its organizational performance
data, for a quarter or a year. Or it could build a cube on
organization-wide on-time delivery performance on the supply
side and sell side, determining not only how often it failed to
meet its delivery, but how often late deliveries on the sell side
corresponded to late deliveries on the buy side. And so on.

4
 Part I: The Deficiencies of Descriptive Analytics

In the first case, you could take last year’s invoice data, build a
cube that contained the spend amount, the products, the volume,
supplier, and location data, and then roll it up by category,
geography, and quarter, and see not only the total spend but
spend patters for the year across the entire organization and, for
the first time, do strategic sourcing with complete and accurate
spend and volume data on the categories being sourced. And,
you could get this insight typically within a week or two of
getting all of the data for the prior year in the data warehouse.

Generally speaking, once the data warehouse schema was

defined, an organization could export the data from all of the
source systems it used in its global operation and get the
data properly structured and appropriately normalized and
categorized in the data warehouse within a few months.

5
 Part I: The Deficiencies of Descriptive Analytics

For an average organization limited to an ERP and Excel, this

was a light year advancement in technology. Before these
systems, analysts would have to take exports from all of the
source systems, slice and dice them into categories, clean
them up and normalize the data in each category by hand, and
then assemble one sheet per category for each quarter or year
(depending on how many transactions there were as Excel had
a 1M row limit per sheet) and then create workbooks with roll
up summaries by quarter or year by category. An analyst could
slave away with the data for months before she had what she
needed just to identify the strategic sourcing opportunities
within a single category.

But now an organization could see their top categories, their

top products and services within each category, and their top
suppliers and how the spend breaks down by department and
geography, and slice ‘n’ dice it across these dimensions in near
real time on their global spend. They could see how it changed
quarter over quarter and month over month, infer some trends
that were only a few months out of date, and identify the top
categories where there was a majority of spend not sourced to
a few key suppliers.

Similarly, they could also identify those categories thought to

be strategically sourced to a few suppliers that were actually
sourced from a large number of (off-contract) suppliers and
those categories where spend was much higher than initially
thought. It was a great time to be a sourcing professional. Until
it wasn’t.

6
 Part I: The Deficiencies of Descriptive Analytics

Once an organization worked its way through its

top N categories that represented 80% of the
spend, signed contracts with the top M suppliers
that provided 60% to 80% of the products in the
top N categories, identified the biggest spending
departments and worked with the lead buyers in
those departments to teach them sourcing best
practices, they hit a brick wall.

If they wanted to dive into the next N categories and the

next M suppliers, look at ways to redefine categories and
supplier segmentation to create new opportunities, or redefine
geographies for greater local negotiating power, they were
out of luck. This would require defining a new OLAP cube, re-
mapping and re-categorizing the data by hand, defining new
views and reports, and basically starting from scratch. As with
the construction of the annual cube, this will take months and by
the time it’s done, the time left to do anything about it will be
minimal as it will soon be time to build next year’s standard spend
cube and begin the annual process all over again.

And this is the major deficiency of descriptive analytics systems.

All they tell you was what was. Not what is. And definitely not
what could be as the data is too far out of date to be useful. So,
what’s the answer? That’s what we’ll attempt to provide in this
series.

7
Part II

Market Intelligence:
The Foundation for
Opportunity Analysis

8
 Part II: Market Intelligence: The Foundation for Opportunity Analysis

I
n our last article, we talked about the deficiencies of the
descriptive analytics provided by the first-generation Business
Intelligence tools. We talked about the fact that while these
tools were revolutionary when they were first released, they
had their deficiencies, which became more prevalent as time
went on.

The ability to build an OLAP cube that could allow invoice data
or AP data for a year to be sliced and diced across categories,
suppliers, geographies, departments, and even component
products and services and then mapped quarter over quarter
and even month over month was quite powerful, even if it did
typically take a quarter (or more) to build the OLAP cube. For the
first-time sourcing professionals could get a relatively complete
picture of historical spend over the prior year and use that to
their strategic advantage in event selection, demand projection,
and negotiation.

9
 Part II: Market Intelligence: The Foundation for Opportunity Analysis

But once the organization had worked its way through its
top N categories that represented 80% of the spend, signed
contracts with the top M suppliers that provided 60% to 80%
of the products in the top N categories, identified the biggest
spending departments and worked with the lead buyers in
those departments to teach them sourcing best practices, they
hit a brick wall. The single OLAP cube could only do so much,
and as per our last post, due to the limitations of the first-
generation data warehouses and business intelligence tools,
building another one that supported a different set of views,
that may or may not be more useful, could take months.

And even if the organization did build a second cube and find
some potential opportunities on the historical data, there’s no
guarantee that the opportunities are still real six months after
the fact. First of all, the demand trend needs to be similar to
what it was six months ago. If demand has dropped sharply,
the best of opportunities will vanish overnight. And if demand
has risen sharply, while the opportunity might be much larger,
the demand satisfaction strategy will likely need to change
significantly.

But even if the trends are similar, that doesn’t

mean that the opportunity still exists, or that
it ever existed in the first place. In order to
appropriately qualify, calculate, and extract an
opportunity, even with the best of strategic
sourcing programs, one needs market
intelligence. Moreover, one needs market
intelligence that can be directly tied to the
categories in question.

10
 Part II: Market Intelligence: The Foundation for Opportunity Analysis

For example, there is only an opportunity if the current (expected)

market price is sufficiently less than the current (average) price
the organization is paying for the good or service, and only if
it possible to obtain the needed good or service in sufficient
quantity at that price. Sometimes the market price can only
be obtained if the product is bought in a sufficient quantity,
and other times only if small quantities are bought (because
traditionally limited demand meant that production lines were
optimized for smaller output batches). And sometimes the market
price is only for a commodity version of the product or service,
not for a customized or improved version.

This is yet another reason why an organization needs good

market intelligence. Without good market intelligence, not only
won’t it know what the market price is, but it won’t know where
the true market opportunities are. All analytics can do on its
own is identify spend and spending patterns, performance and
performance patterns, and where the biggest discrepancies
within the patterns are. For example, where the price paid for a
product is across the board. Where the demand sourced across
the supply base is erratic.

But with market intelligence, an organization

would know when the spend presented an
opportunity for savings through consolidation
or spend leverage, better performance through
shipping re-allocation (for better on-time delivery)
or lean transformation across the supply base, or
both through supply base re-allocation to more
strategic suppliers.

11
 Part II: Market Intelligence: The Foundation for Opportunity Analysis

So where do you find this market intelligence? And how do you

integrate it?

Price data for products and services can be obtained from

de-facto market indices, government contracts, and GPOs.
These days, on-line sites like Amazon and Alibaba can provide
an organization with a de-facto baseline index for commodity
products. This baseline cost data can be augmented with
cost data from government contracts and bids on government
contracts as this data is public. Finally, for deep insight into the
best pricing that can be obtained, an organization can tap into
the knowledge of a GPO that amalgamates millions, if not tens
or hundreds of millions, of spend against a product or service.

But not everything an organization buys is a commodity or

service for which there is a lot of pricing information readily
available. Many organizations buy custom products that there
is no public product pricing for. So, what does an organization
do? It builds should-cost models that allow it to understand
how much the product or service should cost. It builds the
product up as a bill of materials, energy requirements, labour
requirements, and the should cost is all of these component
costs plus a production overhead and a fair margin.

12
 Part II: Market Intelligence: The Foundation for Opportunity Analysis

For every raw material, somewhere there is a commodity index

that can be used to generate a raw cost. Similarly, it is possible
to pull a relatively accurate energy rate because, depending on
where the product is being produced, either the energy rates
are public or the market is deregulated, in which case there are
exchanges rates can be pulled from. Most governments maintain
information on labour rates for different industries and average
labour rates can easily be calculated. Finally, government and
third-party data sources can be consulted to determine average
plant overheads in any industry, and then all an organization
needs to do is pluck in a fair margin to complete a should-cost
model. The price will then be less than what the organization
is paying, indicating a good savings opportunity. About equal,
indicating the organization should stay on the same course, or
greater, indicating the organization has found a supplier that is
doing better than expected. In this last case, the organization
should invest in what should be a strategic supplier to ensure
costs stay down over time.

In other words, with a bit of seeking and integration, for every

primary product in every category, an organization can use
market intelligence to determine market pricing and compare that
to the average price being paid by the organization.

13
 Part II: Market Intelligence: The Foundation for Opportunity Analysis

But this is not the only market intelligence an organization

should be seeking. Sometimes organizations can save by
making appropriate performance improvements. As long as
it can identify areas where performance improvements are
possible in reasonable timeframes and at reasonable cost levels,
performance opportunities can be golden. And here’s another
area market intelligence can help.

If on-time delivery, either in-bound or out-bound, is an issue,

an organization can use freight marketplaces to determine
average shipping timeframes for various transportation modes
as well as determine which carriers consistently meet, or beat,
the average for each transportation mode. Similarly, if the
organization consistently seems to get stuck with too much
obsolete inventory that has to be sold at a loss, the organization
can source data from reputable analyst firms or industry groups
to determine average product life-cycles and determine when it
should plan for end-of-life to reduce product line obsolescence.

In other words, market intelligence is the

true foundation for opportunity identification
and when combined with analytics, it is the
foundation for opportunity valuation.

14
Part III

Trend Analysis: The

Basis for Spend
Forecasting &
Predictive Analytics

15
Part III: Trend Analysis: The Basis for Spend Forecasting & Predictive Analytics

A
s per our first post, analytics is the new hotness (yet
again). And this time, it’s not the old and busted
hotness of the first generation Business Intelligence
tools that, while revolutionary at the time, leave much
to be desired.

The reason analytics is the new hotness is that modern analytics

systems have gone beyond simply presenting a descriptive
overview of historical spend that, by the time it was compiled
was often of limited use, to providing a predictive capability
that allowed an organization to predict future demand and
spend based upon current demand patterns and integrated
market intelligence.

The basis for this predictive capability is trend analysis. While

trend analysis is nothing new in finance, stock markets, and
inventory management, the reality is that it is rather new in the
domain of spend analysis. Why? There are a couple of reasons
for this. First of all, historically, spend “analysis” focused on
compiling descriptive overviews of past spend. Secondly,
traditional trend algorithms were seen as the domain of
inventory planning or price forecasting.

16
 Part III: Trend Analysis: The Basis for Spend Forecasting & Predictive Analytics

But if the whole point of spend analysis is to

understand what you’re spending to identify the
right opportunities, then trend analysis needs
to be part of the domain of spend analysis as
opportunities are not just sourced today, they are
sourced for the next month, the next quarter, the
next year, and sometimes the next three to five
years. In order to select the right opportunity to
source today, one has to understand the impact of
each potential opportunity over time.

This requires the integration of at least three types of trend

analysis and projection:

Demand Forecasting Across Products and

Components

Just because there is a big gap between the price an organization

is currently paying for a product and the current market price
does not mean there is an opportunity. If demand trailed off
between the compilation of the historical data and the current
time, either due to a product entering end of life or an operation
closing down, then the opportunity will be quite limited.

Success requires identifying those categories where demand is

significant, static or increasing, and the price differential (which
can be expected or locked in) is enough per unit to make a
strategic sourcing event worthwhile. For starters.

17
Part III: Trend Analysis: The Basis for Spend Forecasting & Predictive Analytics

It’s also analyzing component and raw material requirements

across product lines to identify those components and raw
materials that are used across products and that are increasing
in both demand and price. By identifying those products that
are likely to increase in price and stay static, or increase,
in volume, the organization can proactively identify good
sourcing opportunities that would otherwise be missed. Even
if the savings is low today, preventing future increases is very
profitable cost avoidance.

And, if the company contracts for a lot of custom manufactured

goods from smaller suppliers with limited buying power, it can
also identify raw materials that an organization could buy on
behalf of its suppliers at a lower cost. For example, if it bought
tons of steel, it could buy at a lower price and then have the
material delivered direct to its suppliers at its cost, taking cost
out of its supply chain one level down.

Cost Forecasting using Market Price,

Commodity, Energy, & Labour Forecasts

Market Intelligence only identifies what you could be paying

today — it doesn’t tell you what you will be paying tomorrow
if you don’t source today. If costs are decreasing, and there is
no urgency to source, then the best thing to do is wait until
sourcing has to be done, and if supply exceeds demand, put
the demand to auction at the last minute. But if costs are
increasing, sharply, sourcing today will avoid cost increases
tomorrow.

But sometimes you can’t forecast based on (market) price

alone, especially for custom manufactured products. That’s
where deep market intelligence on raw material, energy, and
labour costs is required. With this information, an organization
can build not only should cost models for today, but with

18
Part III: Trend Analysis: The Basis for Spend Forecasting & Predictive Analytics

deep cost forecasting at a category level, it can build should

cost models for tomorrow and identify those categories and
products it should be sourcing versus what it should have been
sourcing last year.

Currency Forecasting

Outside of the financial markets, an often overlooked area of

forecasting that is critical to identifying both the right categories
to source and the right time to source is currency forecasting. An
opportunity only remains if the relative value of the currencies in
play remain the same. If a US organization is buying from China,
but the renminbi rises sharply against the US dollar, then the
savings tomorrow will not be the same as the savings today even
if the price is locked in for a year.

The right buy is not only the right product from

the right supplier built at the right location and
shipped using the right method, but the buy that
is also negotiated, and paid, in the right currency.

In this instance, if the organization also sells in China and has

a fair amount of renminbi at its disposal, then the organization
should be cutting the contract in renminbi (and not US dollars)
and paying within China.

This is, of course, assuming that China would still be the best
buy. If the other option was Taiwan, with a slightly higher cost
today, but the projection was that in two months the New Taiwan
dollar was going to fall and remain weak for a year, then the best
contract, taking future expected cost into account, might be with
the Taiwanese supplier that would be higher for two months,
but lower for ten. The reality is that, in global supply chains,
currencies can make, or break, sourcing opportunities.
19
Part III: Trend Analysis: The Basis for Spend Forecasting & Predictive Analytics

Once an organization acquires a modern spend analysis

platform with the ability to project these trends over time,
it can start to more accurately forecast demand and predict
future costs, which allow it to zero in on the true opportunities
of today and tomorrow, not the opportunities of last quarter or
last year. This completely changes the game. No longer will any
time be wasted strategically sourcing categories in decline, or
strategically sourcing categories where the current situation
is almost as good as it gets, and the best thing to do (with
the price increases coming down the pipe) is just extend the
current contracts as long as the suppliers will do so.

Plus, once an organization has trend analytics capability, it can

re-run the projections and analysis on a quarterly or monthly
basis, identify when the forecasted trends change, and when
it might need to take a harder look at a category, supplier,
or geography before a significant price change or disruption
happens. This puts it light years ahead of where it was before
when it only acquired annual insight into a category or supplier,
by which time it might be too late to do anything about an
emerged situation.

And this, our friends, is why analytics is the new hotness yet
again.

20
Part IV

For Deep Insight

Into Cost, Combine
Analytics with (Should-
Cost) Modelling

21
 Part IV: For Deep Insight Into Cost, Combine Analytics with (Should-Cost)

Great sourcing is about minimizing the total cost of ownership

which goes beyond just the unit price or the landed cost, but
also considers other relevant cost factors such as taxes, import/
export duties, and interim storage costs. But just minimizing
total landed cost across a set of bids doesn’t mean you are
making the best buy. It’s only the best buy if the product and
logistics costs are optimized. In fact, even trying to model
and optimize TCO (total cost of ownership) by also taking into
account waste factors, inventory overhead costs, future disposal
costs, etc. does not mean the organization is making the best
buy.

Furthermore, while it might seem easy to just let the market

optimize the total landed cost by going out to bid and inviting
enough suppliers and logistics providers to provide the
requisite competition, it’s not always that easy. Particularly
when the organization is sourcing a manufactured product
and there are only a few suppliers in the mix, who are not
incentivized to get lean and push prices down (because they
have a stranglehold on a significant percentage of the market)
or, even worse, are in a position to collude to keep prices high.

22
 Part IV: For Deep Insight Into Cost, Combine Analytics with (Should-Cost)

But how do you know when the competition in the market is

sufficient to drive an optimized cost or when it’s just not enough?
This is where should-cost modelling comes in. One might think
that basic trend analysis using market pricing or past pricing is
enough to help one determine if the market bids are on track
or not, but there are a few things one needs to remember here.
One, there are generally only market prices for commodities (via
commodity indices) or consumer goods (tied to government
contracts or public catalogs). Two, the organization might not
have enough price/bid history to accurately project what current
pricing should be. Three, the true cost of manufactured goods
depend on a number of variables including, but not limited to,
raw material costs, local labour (market) costs, (local) energy
costs, and other manufacturing overheads (which are often
dependent on the production technique used). The only way to
combine this data and arrive at a likely true cost is with a detailed
should cost model.

But should cost models alone are not enough to

help you understand where the best opportunities
are, or how to go about realizing them. That’s
where analytics comes in. When should cost
modelling is integrated with analytics, the
solution can automatically calculate the current
expected cost of a (manufactured) product,
compare it to the current cost the organization is
paying, multiply it by the expected demand over
a given time period, and estimate the savings
potential of (re)sourcing the category.

23
 Part IV: For Deep Insight Into Cost, Combine Analytics with (Should-Cost)

The organization can quickly see where the biggest

opportunities are using current commodity, labour market, and
energy market costs based upon should-cost models that take
into account current production methods. But this isn’t all
analytics can do.

If the supplier exposes its raw material costs, or if the

organization has enough historical data for both the product
and the component raw materials to determine the impact
commodity market changes have on supplier pricing, it can
also determine when a supplier is likely paying more for a raw
material than it should be. Then, if the organization is buying
a number of (custom) manufactured products from the same
supplier that use the same raw material, it can determine if
it has an opportunity to lower costs further by helping the
supplier achieve a lower raw material cost (by leveraging its
negotiating skills on the supplier’s behalf or by using its buying
power to lock up all of the raw material it needs across its
supply base at a lower cost and providing it to the supply base
at cost).

If the organization also has data on production methodology

for each (custom) manufactured component the organization
buys, then the analytics platform can automatically compute
the average overhead percentage and identify those products
where, if a different production methodology is available, it
might be advantageous to consider an alternate methodology
and create a new should cost model.

24
 Part IV: For Deep Insight Into Cost, Combine Analytics with (Should-Cost)

Moreover, the power of analytics with should cost

modelling doesn’t end there. It can identify those
products where the organization is spending the
most but has the least chance of reducing costs
with a market event and where a lot of that cost
appears to be in overhead. These are prime
situations for the organization to reduce costs by
working with a strategic supplier to implement
lean processes or six sigma to lower costs over
time, or, if raw material costs are rising, keep
costs flat.

In any case, analytics does a great job of allowing an organization

to identify where it is over spending with respect to market costs
or market quotes, but without should cost modelling, especially
where the organization is buying direct (custom manufactured
products), the organization will never understand how low the
costs could be or how much value could be realized.

25
Part V

Demand Projection
and Demand
Management: A Basis
for Cost Avoidance

26
 Part V: Demand Projection and Demand Management: A Basis for Cost

In our last entry in the series, we dove into should cost modelling
and explained how the use of analytics with should cost
modelling could identify cost saving opportunities “under the
covers”. But that’s not the only opportunity for identifying spend
reduction opportunities. Another is better demand management.

The reality is every dollar spent is a hard cost,

and even if the dollar is spent on an item for
resale, that dollar is a lost dollar until such time
as the item is sold for an amount greater than
its total cost and the dollars are collected. And
if it’s an item for use, that dollar is gone, so the
company better realize the value on that dollar
spent (which, generally, involves using the item
to support business operations).

27
 Part V: Demand Projection and Demand Management: A Basis for Cost

Thus, in order for the organization to be practicing good

spend management, it’s paramount that it only buys items it
will actually use or sell, and not too long before those items
are actually used or sold. In the case of inventory, buying 10
widgets needed to maintain the production line “because
you get 50% off for buying ten” is not a great use of cash if,
on average, the widget only has to be replaced twice a year
and the widget happens to cost $10,000 a unit. Sure you save
$25,000 now, but what’s the opportunity cost of not having
that capital? Added to the extra inventory cost of storing some
parts for 5 years, which, even if they’re small, is still pricey when
you consider the inventory has to be secure? Added to the loss
if the production line ends up being modified or replaced in
three years to support the production of new products and the
widgets end up becoming obsolete?

This is an example of where good demand projection, built

on total cost of ownership models, would determine that the
stock level for the widget should not exceed two, with a re-
order of one unit being automatically made when the stock gets
down to one (and only because there is a chance a part could
be defective and wear out quicker than the expected 6 month
utility timeframe and it typically takes a few weeks to get a new
unit in). The organization frees up 40K of cash for investment (in
a resource to do analytics to find more savings opportunity) and
essentially eliminates the obsolescence risk (as the organization
would never have more than one spare on hand when a
production line hit end of life). Security would still be needed,
but if the total inventory size, and value, is small (using a just-in-
time MRO strategy across the board), the overall inventory cost
would be less.

28
 Part V: Demand Projection and Demand Management: A Basis for Cost

As should now be clear, the reality is that spend Forecasting is

more than just cost forecasting, it’s also demand forecasting. But
it’s not just about demand reduction, as in our widget example,
it’s also about identifying demand growth early to gain volume
leverage during negotiations or (e-)Auctions. If you can project a
new product is going to go from 100K units this year to 1M units
next year, going to the market with a tender for 1M units will get
you a lot more interest, and potential savings opportunities, than
going to the market with a tender for 100K or 110K. If the product
was new, and the organization only looked at year-over-year
demand, it might see that demand for the less than one year old
product was 100K, and use its rule of thumb. But if it looked at
month-over-month demand, it might see that demand was 5K,
10K, 20K, and 40K for the last four months of the year and see
that the demand for the product was doubling. It might then
look at market size and similar trends in the past to see that the
demand for products of this type levels out around 100K a month
at peak, and stays at peak for about 9 months. It could then
project a demand of at least a Million units, and go to market
with a much more enticing opportunity.

29
 Part V: Demand Projection and Demand Management: A Basis for Cost

But demand projection and forecasting is not

the only savings lever at your disposal — so
is demand management. Whereas demand
forecasting really centres on identifying the right
time to buy, and the right quantity to buy at a
time (which is EOQ for you inventory managers),
demand management focuses on reducing the
total number of units needed, particularly when
the item is a consumable. (Generally speaking,
the only times you want to reduce the quantity
of goods for resale are when you want to end-
of-life a product or it is just impossible to meet
a greater demand due to raw material or factory
production limits.)

While this is not always as straight-forward, and often requires

situational analytics beyond just the data, it is quite possible
with good analytics tools at your disposal. The first step is
to identify MRO / consumable categories where spend is
increasing. The next step is to figure out why. Sometimes it
will be spend you can’t do anything about. For example, if the
categories are cell phones and laptops, and every employee
gets a cell phone and laptop, and headcount is increasing
30% year over year, cell phone and laptop spend are going to
increase 30% year over year. The best you can do is used the
increased volume to negotiate a better deal.

30
 Part V: Demand Projection and Demand Management: A Basis for Cost

But if the headcount is flat, and the paper and toner cartridge
buy is doubling, and you are in the banking industry, the question
is why? Why are people printing so much? If it’s because they
are printing and sending more and more regulatory compliance
reports and tax filings to authorities, then there might be an
opportunity for reduction as most authorities now accept, or
prefer, electronic filings, if the authorities don’t mandate it. If the
reason is because the online submission processes offered by the
authorities are too onerous, then it might save the company a lot
of money to subscribe to a SaaS service that makes it easy (as
well as prepare the company for the time when online filings will
be mandatory). Forests could be saved. If it’s because managers
are still printing forms to have them filled out by applicants,
maybe it’s time to invest in an easy tablet-based entry system
where they can just hand a tablet to an applicant in a meeting
which can collect all the info and an e-Signature. More forests
could be saved. And if it’s because they are still printing out
reports to compare different breakdowns, maybe it’s time to
invest in a dual-display video card and second monitor for all
employees who need it. Even more forests could be saved. And
while there is an upfront cost to all of this, it won’t take long
before the cost avoidance triples the one-time cost. Paper is
expensive, and toner ink is more expensive than oil or blood.
Think about it. Carefully.

31
 Part V: Demand Projection and Demand Management: A Basis for Cost

Good demand management also identifies MRO or consumable categories

where demand is staying flat but costs are rising sharply relative to
the demand. Maybe the organization is using end-of-life technologies
now supported by only a single supplier who is now capitalizing on
the opportunity and doubling prices year over year due to the captive
market. In this case, the organization would save oodles, and avoid future
cost increases, by switching to a newer technology. Our production line
example above that requires a $10,000 widget to maintain every six months
is one example. Another example is running the IT Centre on 8 year old
servers that require older SATA 2 devices with integrated security options.
Even if the vendor told you they’d last 10 years when you bought them 8
years ago, at this point you should just upgrade … as the total purchase
price would probably be the same as the annual maintenance cost.

But without good analytics, you’d never identify the

categories where demand projection, or demand
management, is key … and huge savings and cost
avoidance opportunities would go undetected.

Hopefully by now you’ll see that when it comes to spend analytics, the top
N opportunities identified by a canned report is only the tip of the iceberg
— and most of the opportunities are below the surface, as with a real
iceberg. All you have to do is dive below the surface.

32
Part VI

The Power of Invoice

Analysis

33
 Part VI: The Power of Invoice Analysis

Many spend “analysis” providers tell people one of the best

places to start is invoice analysis with on-contract suppliers,
as a thorough analysis m-way matched with contracts, POs
(Purchase Orders), and / or goods receipts will often identify
significant overpayments that can be captured as credits or
refunds.

Why is this the case? Consider the following types of

overpayment that are common in an average organization of
considerable size:

Duplicate Invoice Payments

The most common type of overpayment that occurs when a

supplier invoice is not paid on time for whatever reason and
the supplier re-sends the invoice, which gets re-entered and
queued for payment

Overpayments for Defective/Returned Goods

The warehouse rejects goods that are defective, or accepts

goods from an end customer that broke-down under warranty,
and returns them to the supplier that is supposed to credit the
organization, but doesn’t for whatever reason

34
 Part VI: The Power of Invoice Analysis

Fraudulent Payments, Unintentional and

Intentional

The organization gets a seemingly valid invoice from a seemingly

known supplier which is fraudulent either because a valid
supplier(unintentionally) re-sent an invoice with a different invoice
number or a fraudulent party sent an invoice attempting to mimic
a valid supplier (but with different payment instructions that go
unnoticed by AP)

A spend analysis solution that can do an m-way match between

invoices, purchase orders, goods receipts, and payments and
find all the discrepancies can easily find the duplicate payments,
overpayments, and even the (potentially) fraudulent payments
which can be quickly confirmed by human eyes. Then all over-
payments to suppliers under contract can be immediately put in
front of those suppliers and be reclaimed as credit against future
orders.

And that’s good, but this is just a one-time savings opportunity.

Once the reclaimable credits have been captured, what comes
next?

Well, the smarter providers tell you to automate the process and
repeat it monthly and not only detect these overpayments faster,
but identify if unpaid invoices match invoices that have already
been paid or goods already paid for and/or returned. This insures
that less overpayments are made and those overpayments that
are made are generally detected fast enough so they can be
captured as credits before the supplier goes off contract.

35
 Part VI: The Power of Invoice Analysis

And this is great because not only will the organization make
less overpayments over time, but the suppliers will suddenly
get much more accurate because they know even the smallest
mistake will lead to an invoice rejection and/or a delayed
payment. But then what?

What these providers don’t tell you is that invoices are also
great for spend forecasting. And, in particular, those invoices
not tied to POs or those that have POs not tied to contracts.
Why these invoices?

If the invoices are associated with POs, and, in particular, POs

associated with a contract, then the POs are issued according
to a schedule defined by the contract. In this case, they have
no predictive capability beyond the contract as they were
defined by such contract.

But in the other cases, they have great

predictive capability. Consider the case of a
sequence of POs without a contract, where
users or buyers are making spot buys for the
same product or service and the quantity is
increasing month over month. Through the
application of trend analysis and comparisons to
similar categories, the organization can predict
future volume, spend, and identify the best
opportunities for cost savings and avoidance
before spend in these products gets out of
control.

36
 Part VI: The Power of Invoice Analysis

And the invoices without a PO can be an even better indicator of

emerging spend areas. If the organization keeps seeing spend
across suppliers for a new type of service, it can be a leading
indicate of a change coming down the pipe as the organization
transitions to new product lines that need to be supported by
new services. And if it sees new products, that could indicate
changes in production line technology or back office equipment
that will soon be standardized and allow Procurement to get
involved early, even before the organizational departments realize
that they need Procurement’s support.

But beyond that, you can identify commonalities not just

across products or services, but supplier selection (are certain
individuals trying to phase suppliers in discretely), categories
(which might just be emerging and not well defined), and
even geography (especially once one or two time invoices are
filtered out). The power of invoice analysis goes well beyond
overpayment detection. At the end of the day, all business is
about changes in the balance sheet, and invoices are the best,
and most consistent, early indicators of that. So why not use
them to your advantage?

37
About the Author

Michael Lamoureux, aka the doctor, is the Editor-in-Chief of Sourcing

Innovation (.com), a resource for sourcing, procurement, and supply chain
professionals who are interested in improving themselves and the overall
performance of their organizations. A regular contributor to Spend Matters,
he is a Computer Science PhD who has been heavily involved in the Sourcing
and Supply Chain Space since 2000 and the e-Commerce space since 1997.
As a freelance procurement consultant with extensive expertise in sourcing,
procurement, and supply chain processes, he aims to continually push
innovation in and beyond the supply chain space. With particular expertise
in analytics, modeling, and optimization, he is able to dive much deeper into
technology and core issues, striving to help businesses with their internal
knowledge transfer, positioning, and planning problems.

38