A M M I 3 ( ( P P P' : G : ER M B < L F : @ : S B G > ' < H F

=KBOBG@;>MM>K;NLBG>LL=><BLBHGL F :R (C N G >+)*)
G>+)*)

8LI*YXYVISJ
,)%08,'%6)
K^_hkfE^`bleZmbhg3Ikh`ghlbl_hklrlm^flZgZerlml
<Zg\^kMk^Zmf^gm3KZ]bZmbhgiarlb\lZg]himbfbsZmbhg

%073-27-() Lik^Z]la^^m
:gZermb\l3
 8)<8%2%0=8-'7Bglb`am_nebgm^kl^\mbhgh_]bl\biebg^l Lnko^rh_*1)
 %2%0=>)8,-7Bfi^k_^\mob^ph__Zlm^oheobg`phke] ikh]n\mlmh
a^eirhnphkd
 79440=',%-27%+%LZe^lHi^kZmbhglIeZggbg` [^mm^k%_Zlm^k
Zg]lfZkm^k
 Experience how AIMMS
B G L B =>  LMH KR
enables you to apply
:g]ma^lnko^rlZrlÀ OR successfully
We recently completed the first of Analytics: 64 percent use it to keep up
what we expect will be many reader- with breaking analytics news (no small
ship surveys. As any analyst knows, feat for a bimonthly), 66 percent use
it’s nearly impossible to make im- it to learn about analytics (welcome
provements if you don’t know where to the show) and 75 percent use it to
Hassle-free deployment
things stand at the moment and you gain broad exposure to analytics top- alternatives to end-users
don’t know the problem you’re fac- ics (never too old to learn new tricks). ,

#


(#

,


#'$
ing. The online survey, conducted in Perhaps the most telling statistic ,

)

March, not only gave us a good idea was how much time the respondents IZkZ`hg ,


!%
,



of where we are in terms of reader- said they spent reading a typical issue ,

%
ship satisfaction, wants and needs, it of Analytics. More than 55 percent
also provided an invaluable roadmap
regarding where we should be head-
said they spend at least 30 minutes
reading each issue and 22.1 percent
?neeIZ`^:kmphkd Fast GUI builder to visualize
results and to interact with
ing in the months and years ahead. said they spend at least an hour read- your model
We want to thank the 350-plus
readers who participated in the sur-
ing it – incredible numbers compared
to other online publications. The re- G^p:kmphkd
vey. While we’re pleased with the ini- spondents also rated the quality, val- CONOPT
tial response to Analytics and believe ue and timeliness of the content quite
we’re on track to attract a huge audi- high, and 83 percent of respondents ILOG CPLEX and many
more...
IZ`^* Seamless links to
powerful solvers

ence of analysts who are hungry for said they were likely to recommend it
our content, we realize we have plen- to a colleague, supervisor or student.
ty of room for improvement. Best of all, 70 respondents said
OBDC Point-and-click integration
The readership survey provided they were interested in contributing an OLE DB with data sources
...
some particularly gratifying results: article to Analytics, and I’m following
On a scale of 1 to 10 with 1 being “very up with them as we speak. As I say
dissatisfied” and 10 being “very satis- almost every issue, Analytics is an
Please visit our website for application
fied,” 82.7 percent of respondents rat- interactive exercise. We look forward examples, industry case studies and a
Transparent and maintainable
ed Analytics a “7” or higher and 62.9 to your comments, editorial ideas and model formulation free full trial license.
www.aimms.com
percent rated it “8” or higher. Nearly contributed articles. ❙
10 percent gave it a perfect “10.”
More insightful, from our stand- ÇI>M>KAHKG>K%>=BMHK AIMMS is a registered trademark of Paragon Decision Technology B.V.
point, was the reason readers use ahkg^k9 ebhgakmin['\hf Other brands and their products are trademarks of their respective holders.
Europe, Middle East & Africa = Haarlem, The Netherlands = Phone: +31 (23) 5511 512
Americas = Kirkland, WA, USA = Phone: +1 (425) 576 4060
Asia Pacific = Singapore = Phone: +65 9640 4182
*       u      : G : ER M B < L F : @ : S B G > ' < H F www.aimms.com = info@aimms.com
< H G M > G M L 
 



 
 
 


 
DRIVING BETTER BUSINESS DECISIONS F :R ( C N G >  + )*)

*)%896)7
8 K^lhnk\^l_hkLik^Z]la^^m:gZerlml
;rMahfZl:'@khllfZg
Lnko^rh_fhk^maZg*1)lik^Z]la^^mZgZermb\ikh]n\mlihbgmma^
pZrmhpZk]phkdbg`[^mm^k%_Zlm^k%lfZkm^k'

14 M^qm:gZermb\l3MphPhke]l<heeb]^
;rDZmarEZg`^
8 Bgm^kl^\mbhgh_mph]blmbg\m]bl\biebg^l`e^Zglbglb`am_khffZllbo^
Zfhngmlh_nglmkn\mnk^]]ZmZ'

20 :]oZg\^]:gZermb\l_hkLHI
;rC^k^fr?'LaZibkh
>__^\mbo^LZe^lHi^kZmbhglIeZggbg`]^\blbhg&fZdbg`\ZgÌm[^
Z\ab^o^]pbmahnmbglb`aml_khfZgZermb\l'

27 A^Zema<Zk^K^_hkfIZll^l'GhpPaZm8
Easier, More Powerful – and Far More Visual.
;r=hn`eZl:'LZfn^elhg
Ik^]b\m^]hnm\hf^l3@nZk]^]himbfblf_hk`^g^kZein[eb\Zg] From its classic Solver Parameters dialog to its modern
14 Ribbon, modeless Task Pane, and charts to visualize the
iZmb^gml4^q\^ee^gmikh`ghlbl_hka^Zema\Zk^lrlm^flZgZerlml'

and a wide array of charts and graphs. One simple,

31 FZmaFh]^elBfikho^<Zg\^kMk^Zmf^gm consistent user interface makes learning and teaching
;r@bghC'Ebf
;k^Zdmakhn`aikhc^\m\hf[bg^l^qi^kmbl^bgmphl^iZkZm^Zk^ZlÇ Free Trials and Support for Your Eīorts at Solver.com.
kZ]bZmbhgiarlb\lZg]himbfbsZmbhg'
about the many other ways we can support your
teaching and learning eīorts – all at www.solver.com.

20 and produce instant reports, charts and graphs, without

programming. And you can do even more in Excel VBA. Tel 775 831 0300ͻFax 775 831 0314ͻŝnfo@solver.com

+       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      P B G M > K  + ) ) 1

 
 



 

 

DRIVING BETTER BUSINESS DECISIONS

 


 
REGISTER FOR A FREE SUBSCRIPTION:
http://analytics.informs.org

INFORMS BOARD OF DIRECTORS

President Susan L. Albin, Rutgers University
6 President-Elect Rina Schneur, Verizon Network & Tech.
Past President Don M. Kleinmuntz, Strata Dec.Tech.
Secretary Anton J. Kleywegt, Georgia Tech
Treasurer Stephen M. Robinson, Univ. of Wisconsin
Vice President-Meetings Robin Lougee, IBM Research
Vice President-Publications Terry P. Harrison, Penn State University
Vice President-
Sections and Societies Ariela Sofer, George Mason University
Vice President-
Information Technology Warren Lieberman, Veritech Solutions
Vice President-Practice Activities Jack Levis, UPS
Vice President-International Activities Jionghua “Judy” Jin, Univ. of Michigan
Vice President-Membership
and Professional Recognition
Vice President-Education
Vice President-
Pinar Keskinocak, Georgia Tech
Jayashankar M. Swaminathan, UNC ?khgmebg^
Marketing and Outreach Anne G. Robinson, Cisco Systems Inc.
Vice President-Chapters/Fora Stefan Karisch, Jeppesen
38 INFORMS OFFICES
www.informs.org • Tel: 1-800-4INFORMS ?neeIZ`^:kmphkd
Executive Director Mark G. Doherty
()4%681)287 Marketing Director Gary Bennett
Communications Director Barry List

1 Bglb]^Lmhkr Corporate, Member, INFORMS (Maryland)

Publications and 7240 Parkway Drive, Suite 300
G^p:kmphkd
Subdivision Services Hanover, MD 21076 USA
Tel.: 443.757.3500
4 IkhÕm<^gm^k E-mail: informs@informs.org Risk Analysis is Easy as What-If with our Ultra-Fast,

6 :gZers^Mabl
Meetings Services INFORMS (Rhode Island)
12 Breakneck Hill Road, Suite 102
Lincoln, RI 02865 USA
IZ`^,
In true Excel 2007 style, select from galleries, drag and
Tel.: 401.722.2595
E-mail: meetings@informs.org the
35 <hkihkZm^IkhÕe^
ANALYTICS EDITORIAL AND ADVERTISING watch charts update instantly with new results.
consistent user interface makes learning and teaching
38 <hg_^k^g\^Ik^ob^p Lionheart Publishing Inc., 506 Roswell Street, Suite 220, Marietta, GA 30060 USA
Tel.: 770.431.0867 • Fax: 770.432.6969
President & Advertising Sales John Llewellyn Free Trials and Support for Your Eīorts at Solver.com.
39 Inssehk Llewellyn@lionhrtpub.com
Tel.: 770.431.0867, ext.209
Editor Peter R. Horner about the many other ways we can support your
horner@lionhrtpub.com teaching and learning eīorts – all at www.solver.com.
Analytics (ISSN 1938-1697) is published six times a year by Tel.: 770.587.3172
the Institute for Operations Research and the Management Senior Art Director Alan Brubaker
Sciences (INFORMS). For a free subscription, register at http:// albrubaker@lionhrtpub.com
analytics.informs.org. Address other correspondence to the Tel.: 770.431.0867, ext.218
editor, Peter Horner, horner@lionhrtpub.com. The opinions Junior Art Director Kat Wong
expressed in Analytics are those of the authors, and do not katwong@lionhrtpub.com
necessarily reflect the opinions of INFORMS, its officers, Tel.: 770.431.0867, ext.223 produce instant reports plus standard, Overlay, Trend
Lionheart Publishing Inc. or the editorial staff of Analytics. Advertising Sales Sharon Baker and Box-Whisker charts, without programming. Tel 775 831 0300 ‡ Fax 775 831 0314 ‡ info@solver.com
Analytics copyright ©2010 by the Institute for Operations sharonb@lionhrtpub.com
Research and the Management Sciences. All rights reserved. Tel.: 813.852.9942

,       u      : G : ER M B < L F : @ : S B G > ' < H F

IKH?BM<>GM>K
:gZermb\lblZeeZ[hnm
gnf[^klÀhkblbm8
M><AGB<:E<A:EE>G@>L Analytics is about numbers. But it isn’t just
<:GLEHP:IKHC><M about numbers. It’s also about people.
=HPG:G=F:RK>JNBK> An analytically inclined manager once shared
<A:G@>LMHFH=>EL%;NM an anecdote about his troubles getting sales
M><AGB<:E<A:EE>G@>L agents to use their laptop computers. Some of
K:K>ER<:NL>:IKHC><M the agents had been around for decades, and
MH<HEE:IL>'IHHK the idea of using laptops wasn’t just foreign, it
I>HIE>F:G:@>F>GM was unthinkable. In one particular meeting a
sales agent showed up and declared that he’d
<:GE>:=MH?:BENK>:G=
finally broken down and used his laptop to write
=:F:@><:K>>KL'
a memo. He then proceeded to demonstrate.
Laying the computer on his lap, he pulled out a
sheet of paper and wrote a memo with his pen,
using the laptop as a writing surface.
The behavior wasn’t appropriate, but it was
a sign of more than unwillingness to change.
Analytics is second nature to individuals with
training in data analysis, but it can be alien to
those who aren’t. Today, computers are ubiq-
uitous at all levels of education. However, an
abundance of individuals remain who’ve never
encountered computers in school – much less
the analytics they enable.
Further, many people fundamentally aren’t
comfortable with mathematics – the foundation
of analytics. It’s not hard to find people who will
;R>':G=K>P;HR= happily share their anxiety about mathemat-
ics, from the algebra classes they failed in high
school to the statistics classes they dropped in
-       u      : G : ER M B < L F : @ : S B G > ' < H F
college. Mathematics comes naturally to some
and not so naturally to others.
Fear of mathematics is only part of the equa-
tion. People also fear loss of control. An individual
who’s responsible for the placement of contain-
ers on a cargo ship may well resent a computer
telling him how to do his job. He’s done it for years
and has developed considerable expertise. Why
should he listen to a computer any more than he
should listen to a novice?
All analytics projects impact people, and
people are often the key factor in whether an
analytics project is successful or not. This of-
ten comes as a surprise. After all, analytics
projects must deal with computers, data quality
choosing and building an appropriate mathe-
matical model. Aren’t these greater challenges
than those posed by people?
Usually not. Good analytics practitioners
can understand and address technical chal-
lenges. Technical challenges can slow a proj-
ect down and may require changes to models,
but technical challenges rarely cause a proj-
ect to collapse. Poor people management can
lead to failure and damage careers.
So what can be done? People issues aren’t
unique to analytics projects, but analytics proj-
ects have their own special challenges. Some
thoughts to keep in mind:
• Communication. Keep stakeholders
informed. Once isn’t enough. A regular
flow of information is required – in both
directions. Familiarity doesn’t breed
contempt; it breeds familiarity. Stakeholders
will come to accept an analytics project and
the project team over time. If not, problems
will be uncovered early and steps can be
taken to rectify them.
• Keep a list of stakeholders. General
attitudes of different parties can usually be
discerned early in a project. Is an individual
an avid supporter? Needs convincing? Dead
set against the effort? A plan should be
developed for every group of stakeholders,
and revised and acted upon at least weekly.
Individuals who are strong supporters and
opponents should be ascertained and
receive special treatment.Analytics projects
are unique in that the results aren’t always
obviously right or wrong. The people in a
payroll department may like or dislike the
design of new payroll software, but everyone
can agree if the amounts on the checks
that go out are right or wrong. An analytics
project to set prices for a company’s goods
may come up with different numbers than
those a pricing analyst would determine.
Whose prices are better, those of the pricing
analyst or the analytics professional? This
naturally leads into our next issue.
• Education. Practitioners with years of
experience won’t accept numbers that
come from a black box. The black box
needs to be opened, and this can only be
achieved through education. Education can
be a challenge since many of the tools used
by analytics professionals are, by nature,
quite mathematical. Creative education is
often required, where important concepts
are taught without mathematical formalism.
: G : ER M B < L      u      F : R ( C N G >  + )*)
IKH?BM<>GM>K

PA>GBMÌL<E>:K Games and interactive brainteasers can be

MA:MBG<>GMBO>L very effective in this regard.
=BK><MERBFI:<M • Incentives. The best analytics project can
MA>LN<<>LLH? be thwarted by a poor incentive structure.
:IKHC><M%MA>R If a sales force receives bonuses based on
FNLM;>:==K>LL>=' volume, sales agents will work diligently to
keep prices low for their customers, even
if analysis of the numbers suggests raising
prices. When it’s clear that incentives
directly impact the success of a project,
they must be addressed.
• Reassignment. In spite of a project team’s
best efforts, there often remain individuals
who, for one reason or another, fail to embrace
the changes brought about by an analytics
project. This impacts not only their own work,
but also the work of those around them. The
problem can be especially disruptive in an
analytics project since, as discussed earlier,
there’s generally no absolute right or wrong.
Legitimate concerns should be aired, but when
the activity is counterproductive, it’s best to
make personnel changes rather than let the
problem fester.
Above all, it’s important to stay focused on
the ultimate goal: project success. Success is
measured by an organization using better num-
bers – numbers that lead to lower costs or higher
margins. But these numbers will only be used if
they’re accepted by the organization. That’s why
it’s important to address that most challenging
element of most projects – the human element.
Andrew Boyd served as an executive and chief scientist at an
analytics firm for many years. He can be reached at e.a.boyd@
earthlink.net.

.       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      F : R ( C N G >  + )*)

: G : ER S >  M A B L 
;nlbg^llZgZermb\l3
Mhh\hhe_hk!cnlm"l\ahhe
Gregory is an old friend of mine, a sea-
soned sales leader, a start-up turnaround
specialist and an early stage company ex-
ecutive. He has seen an awful lot in his long
career, and whenever I get together with him
I inevitably learn something. When I see his
name flash up on my cell phone, I pick up
with a smile, thinking he’s calling to see when
we might sneak out of work for an afternoon
at the ballpark. Turns out he has something
quite different in mind.
“You know about this analytics stuff, right?
Great. I’ve got a company that I want you to
take a look at,” he says. “They have a software
platform that sits in ‘The Cloud’ and it’s using
some smart predictive analytics to help on-line
retailers do better targeting. You know that ‘The
Cloud’ and predictive analytics are both at the
top of Gartner’s list of strategic technologies for
2010 [1], right? So these guys are trying to bring
both of them together in a pretty interesting way.
Anyway, we’d like to come up to campus and tell
you more about the company and see if you are
interested in getting involved – can you make
;ROBC:RF>AKHMK: time for a meeting sometime next week?”
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE! Just visit: http://analytics.informs.org/
/       u      : G : ER M B < L F : @ : S B G > ' < H F
As I listen to him, I realize that I’ve rarely
heard Gregory sound this excited about any-
thing. “Gregory,” I tell my old friend, “come
on over. I can’t wait to hear what you guys
are up to with analytics out there in the real
world.”
A long time ago I was a grad student
in operations research, a discipline sitting
at an intersection of mathematics, indus-
trial engineering, statistics and computer
science. For the past couple of decades,
much of my work has involved O.R., first
as a consultant, then as an entrepreneur
and most recently as a business school
professor. Some time back in the last cen-
tury, I began writing a column called “Was
It Something I Said?” in OR/MS Today, the
membership magazine for the Institute for
Operations Research and Management
Science (INFORMS). I screamed at the rest
of the world to recognize the power and
glory of operations research, and I howled
at my O.R. brethren to look for real appli-
cations upon which to focus their powerful
intellectual capabilities instead of tirelessly Meanwhile, in a parallel universe, a num-
pushing deeper and deeper into theoretical ber of powerful and (imperfectly) correlated
crevices. And over the years, while I got developments were taking place. The cost of
a moderate amount of e-mail from readers computing power was plummeting. The capa-
who appreciated my thoughts on one topic bility of spreadsheets and more sophisticated
or another and the occasional angry note data analysis platforms was exploding. The
from someone who I managed to offend, business world was growing more and more
the world outside of INFORMS generally competitive, pushing managers and execu-
paid relatively little notice to either my col- tives to search for competitive advantages
umn or to our profession. wherever they could possibly find them. Data
: G : ER M B < L      u      F : R ( C N G >  + )*)
: G : ER S >  M A B L 
BLMA>K>K>:EER: storage was getting cheaper and cheaper and
K>OHENMBHG;K>PBG@ “business intelligence” vendors were busy
HNMMA>K>BG:G:ERMB<L8 creating data structures and reporting capa-
BÌ=EBD>MHMABGDLH% bilities to give businesses a way to look at all
;NMFRLMKHG@L>GL> that raw data that was suddenly laying around.
BLMA:MMA>K>BLLMBEE: The Internet had emerged as an amazing
EHG@P:RMH@H' platform for capturing mountains of data, an
inexpensive laboratory for conducting statisti-
cal experiments, and a superhighway that en-
abled information to be moved, analyzed and
presented at amazing speed. And pioneers in
disparate industries – including finance, trav-
el and hospitality, gaming and supply chain
management – were demonstrating that, with
Seemingly overnight, the world awakened to the potential of quantitative analysis to create
huge business value.
0       u      : G : ER M B < L F : @ : S B G > ' < H F
the clear-eyed application of an appropriate
amount of mathematical elbow grease, the
growing piles of data could be successfully
mined for gold.
And then, seemingly overnight, the world
awakened to the potential of quantitative anal-
ysis to create huge business value. The past
couple of years we have seen a slew of books
such as “Super Crunchers” [2], “The Numerati”
[3] and “Competing on Analytics” [4], a New
York Times article [5] advocating how cool it is
to be a statistician today, and, most recently,
huge public validation by people like Gartner
and Accenture [6] of just how important this
stuff called analytics is going to be.
Both INFORMS and I are scrambling to find
our place in this world. In addition to launching
the online magazine that you are reading right
now, INFORMS has recently engaged a major
consulting partner to help us understand how
we can better serve this bigger, broader do-
main of analytics. Meanwhile, I have recently
been hired as a professor of business analytics
at the University of San Francisco, and in an
effort to keep up with what my students should
know (and what many friends in the business
world already know), I have agreed to write
this new column.
Is there really a revolution brewing out
there in analytics? I’d like to think so, but my
strong sense is that there is still a long way to
go. A recent Accenture study [7] of executives
in the United States and the United Kingdom
confirms a lot of things that many of you no
doubt have experienced: data still too often
living in silos, a shortage of dedicated analyti-
cally skilled resources, and managers who still
feel more comfortable relying on intuition and
gut feel than on structured analysis, especially
when planning for the future. Many of my aca-
demic colleagues also bemoan the mathemati-
cal preparation of today’s students, suggesting
that the pipeline of analytical people needed
to solve this talent gap is probably not as full
as it needs to be. On the other hand, things
are moving so fast – new software tools, new
mathematical techniques, new industry appli-
cations, and new success stories – that I will
no doubt have lots to write about.
Welcome to “Analyze This!,” one man’s im-
perfect view on the rapidly evolving world of
analytics. I’m counting on you all to help keep
that view expanding, so feel free to drop me a
note to share your thoughts, ideas and experi-
ences. See you next time. ❙
Vijay Mehrotra (vmehrotra@usfca.edu) is an associate professor,
Department of Finance, Economics and Business Analytics,
School of Business and Professional Studies, University of
San Francisco. He is also an experienced analytics consultant
and entrepreneur and an angel investor in several successful
analytics companies.
REFERENCES
1. See www.gartner.com/it/page.jsp?id=1210613
for the complete list.
2. www.randomhouse.com/bantamdell/supercrunchers/
3. http://thenumerati.net/index.cfm?catID=18
4. www.amazon.com/Competing-Analytics-New-Science-
Winning/dp/1422103323
5. www.nytimes.com/2009/08/06/technology/06stats.html
6. www.sas.com/news/preleases/
AccentureSASAnalyticsGroup.html
7. http://newsroom.accenture.com/article_display.cfm?
article_id=4935
: G : ER M B < L      u      F : R ( C N G >  + )*)
L H ? M P: K >  L N K O > R
K^lhnk\^l_hk
Lik^Z]la^^m:gZerlml
Fhk^maZg*1)lik^Z]la^^mZgZermb\ikh]n\ml
ihbgmpZrmhpZk]phkdbg`[^mm^k%_Zlm^k%lfZkm^k'
;RMAHF:L:'@KHLLF:G
SPREADSHEETS HAVE BEEN an im-
7 portant platform for analytics
ever since VisiCalc launched
in 1979. Spreadsheets put
the power of mathematical modeling and
computer programming into the hands of
anyone with a PC. Today, spreadsheets
are as much a part of the modern office
as a desk; it’s virtually impossible to find
a knowledge worker without a copy of Ex-
cel on her PC.
To understand the true impact of
the spreadsheet, think about how the
Internet “disintermediated” business
1       u      : G : ER M B < L F : @ : S B G > ' < H F
by getting rid of the middleman. In the
world of analytics, disintermediation
happened a decade before the Internet
arrived. It used to be that a business
analyst had to work through the gate-
keepers of the information systems de-
partment and wait months or years to
get a tool to execute his analysis. The
spreadsheet got rid of the analytic mid-
dleman. With a spreadsheet, an analyst
could do his own analytical program-
ming at a fraction of the cost with cycle
times orders of magnitude faster than
previously. This was truly revolutionary.
Since 1997 I’ve been tracking the grow-
ing ecosystem of “Spreadsheet Analyt-
ics.” By this I mean tools and techniques
related to using spreadsheets for busi-
ness analysis. Back in 1997, there were
a limited number of tools, mostly add-ins
or templates. These were concentrated in
operations research algorithms such as
optimization and simulation. They gave
Joe and Jane Analyst access to analyti-
cal power that previously was available
only to technical experts.
Today, the ecosystem of Spreadsheet
Analytics has expanded dramatically. With
that in mind, a team of researchers at the
University of San Francisco recently sur-
veyed hundreds of spreadsheet analytic
products in order to ascertain the current
state of the field. We uncovered more than
180 commercial products from 150-plus
vendors plus a handful of notable academ-
ic tools that we have compiled as an open,
online resource for any interested analyst.
For more details, see the accompanying
story “Spreadsheet Analytics Web Re-
source” and visit “Spreadsheet Analytics:
Resources for Spreadsheet Analysts” at
www.usfca.edu/bps/spreadsheet-analytics.
This article highlights resources that help
spreadsheet analysts do their jobs bet-
ter, faster and smarter while managing the
complexity and risk caused by a prolifera-
tion of spreadsheet information systems.
3:)6:-);
SPREADSHEET ANALYTICS is a big space.
It’s easier to make sense of it if we divide
it into three parts: 1. doing analytics, 2.
before-and-after analytics, and 3. manag-
ing analytics.
First, we’ll discuss resources for actu-
ally doing analytics – the computations
and algorithms that are the stuff of ana-
lytical work. This includes “model-driven”
analytics such as optimization and simu-
lation that are traditionally taught in the
“quant course” at many North American
business schools; “data-driven” analytics
such as statistics, data mining and busi-
ness intelligence; and functional-area
analytics used in finance, six-sigma and
other areas of business.
Second, we’ll look at the unglamor-
ous before-and-after analytics activities
related to data and reporting. These
activities are the blocking-and-tackling
skills that can derail even the best an-
alytic concept if they’re not executed
well. They include the painstaking work
of moving and cleaning data so that it
gets into a spreadsheet in usable form,
and the often tedious work of populating
the reports and dashboards that bring
analytical results to decision-makers.
These activities can be repeated weekly
or monthly so tools to simplify or auto-
mate them can be very valuable.
: G : ER M B < L      u      F : R ( C N G >  + )*)
L I K > : = L A > > M  : G : ER M B < L
Third, we’ll explore the important
managerial work of making spreadsheet
analytics work in organizations – doing it
well and doing it right. This is the stuff that
gets the attention of senior analysts and
the managers accountable for analytical
cost, quality and regulatory compliance.
This includes spreadsheet development,
spreadsheet management, spreadsheet
tracking, management of spreadsheet
information systems and regulatory com-
pliance. These issues can be critically
important to the analytical capability of
organizations, but many analysts fail to
anticipate them and suffer accordingly. In
my view, these areas deserve the atten-
tion of anyone who considers himself to
be a true analytical professional.
(3-2+%2%0=8-'7
'31498%8-327%2(%0+36-8,17
IN OUR SURVEY we determined that the
tools for doing analytics fall into three
distinct groups: data-driven analytics,
model-driven analytics and business
function analytics.
Data-Driven Analytics. Data-driven
analytics is for when you want to learn
A^eiIkhfhm^:gZermb\l
It’s fast and it’s easy! Visit:
http://analytics.informs.org/button.html
2       u      : G : ER M B < L F : @ : S B G > ' < H F
something from a dataset. Statisticians
rejoice; according to our survey there are
plenty of spreadsheet statistics add-ins at
a variety of price points. Data hounds will
find three commercial tools for spread-
sheet data mining. One provides a vari-
ety of the usual excavating techniques,
one uses something called “self-organiz-
ing modeling technologies,” and the third
uses neural networks.
Although business intelligence (BI)
has many server-based products, our
survey found no true spreadsheet busi-
ness intelligence tools. We did find two
products for tracking and visualizing BI
data. Most of the action is on the server
side, and I am glad to tell you that most
BI publishers provide an add-in for getting
data to and from Excel. For spreadsheet
BI, the big news is old news: native Excel
provides decent BI functionality with its
powerful sort, filter, pivot table and pivot
chart features. They’re not glamorous but
they are effective at slicing, dicing and
summarizing business data.
Spreadsheet time-series forecast-
ing has almost a dozen add-in products.
These generate time-series forecasts
using smoothing, regression and neural
network approaches. If you’ve been wait-
ing for spreadsheet neural networks you
will be pleased to learn our survey dis-
covered five general-purpose products.
L I K > : = L A > > M  : G : ER M B < L
Model-Driven Analytics. Model-driv-
en analytics is for when you have a model,
such as a forward-looking planning mod-
el, and need to extract insight or decision
guidance. This is the playground for the
algorithms beloved by operations research
professionals. We’ll start with the unsung
workhorse of spreadsheet sensitivity analy-
sis. Our survey found several tools to “stress
test” a model by varying inputs to see the
impact on a key output and presenting the
results in a Tornado Chart. There are tools
and techniques for quickly computing and
charting the effect of one or two inputs on
an output, and running scenarios through
your model.
Spreadsheet goal-seeking is growing
up. The clunky Excel goal seek tool is no
longer the only option. Our survey found
new goal seeking tools that handle multi-
ple-goal seeks at once and can remem-
ber and repeat them. Nice!
For the operations research crowd,
spreadsheet solvers that deliver spread-
sheet optimization (aka “linear pro-
gramming”) are thriving, with powerful
replacements/upgrades to the free Excel
Solver. Frontline Systems has a break-
through product called RiskSolver Plat-
form that handles models that made older
solvers go belly-up. Long-time competitor
Lindo Systems continues to upgrade and
improve its solid What’s Best!. Our survey
*)       u      : G : ER M B < L F : @ : S B G > ' < H F
shows several new spreadsheet genetic
algorithms and other spreadsheet meta-
heuristics at a variety of price points.
For spreadsheet Monte Carlo simulation,
the old duopoly of Oracle’s Crystal Ball and
Palisade’s @Risk are being challenged at
the high and low end. Notable entrants are
Vose Softare’s ModelRisk and Frontline Sys-
tems’ RiskSolver with innovative “simulation
optimization” capability. Our survey indicates
that the spreadsheet decision tree space is
growing with new products to compete with
the business school standard TreePlan. Pal-
isade and Frontline have decision tree prod-
ucts integrated with their other tools.
Business Function Analytics. Busi-
ness function analytics is about tools
designed for specific functional areas of
business. The survey found a half-dozen
spreadsheet finance tools for pricing vari-
ous types of derivatives, to support trad-
ing decisions (with integration to brokers)
and even a tool for automatic data feeds.
In spreadsheet marketing there is tool for
spreadsheet marketing engineering, tem-
plates for spreadsheet marketing plans
and a spreadsheet advertising keyword
pricing tool.
If you are fan of doing things right the
first time, you will be interested in spread-
sheet Six Sigma and quality tools. Our
survey discovered products for Six-Sigma
greenbelts to focus on the process not
the math, products for statistical quality
control and even tools intended for use
on the factory floor.
Do you hate waiting in line? To con-
nect resources to customer waiting lines,
a wonderful spreadsheet queueing add-
in offers custom functions to program
“queueing theory” formulas right in your
spreadsheet, plus a couple of other
academic tools. For more traditional
spreadsheet operations management
the survey discovered a set of spread-
sheet project management templates as
well as an add-in for spreadsheet data
envelopment analysis, which is a useful
technique to evaluate the performance
of departments or branches based on
their inputs and outputs.
&)*36)%2(%*8)6%2%0=8-'7
(%8%1%2%+)1)28%2(6)4368-2+
BEFORE YOU CAN do analytics, you need
data. It may come as a surprise to some
readers that getting access to usable data
is often a big task in itself. Unless your
outfit has a data warehouse that actually
works like the salesman promised, you’re
probably going to have to clean your data
to get rid of junk and odd characters and
numbers-as-text, and you may have to
consolidate data from different sources
or multiple spreadsheets into one place.
And don’t forget to remove those pesky
duplicate rows – everybody’s favorite
way to spend Saturday morning! If you
have regular business processes, you
might have to go through this monthly or
even weekly.
If you don’t have an intern to do these
exciting tasks, you can get some help
from software. A half-dozen spreadsheet
data management and cleaning tools
made the cut for our survey, including
a couple that already come with Excel.
A half-dozen spreadsheet consolidation
tools, including Excel’s often-overlooked
consolidate tool, are also available.
After you’ve sorted out the data and
done the analytics, you’ll need to share
what you’ve achieved. If you are suc-
cessful enough that your analytics are in-
tegrated into routine decision-making or
pulse-taking processes, you’re going to
want to regularly transfer analytic outputs
from spreadsheets into reports. If your
organization is up-to-date on its buzz-
words, you’ll want to move spreadsheet
analytic outputs from spreadsheets into
dashboards that graphically summarize
results. Doing this by hand gets old af-
ter a while, so utilize tools that automate
posting outputs to dashboards and Web
sites and that will open workbooks, print
reports and close files at the push of a
button. Your days of cubicle drudgery can
come to an end!
: G : ER M B < L      u      F : R ( C N G >  + )*)
L I K > : = L A > > M  : G : ER M B < L
1%2%+-2+%2%0=8-'7
(3-2+-8;)00%2((3-2+-86-+,8
MANAGERS SHOULD BE interested in the
productivity of their spreadsheet analysts;
this is the important task of “spreadsheet
development.” Managers, especially senior
managers, need to make sure that for cer-
tain spreadsheets adequate controls are in
place to ensure compliance with govern-
ment regulations such as Sarbanes-Oxley,
as well as sound risk management for fi-
nancial decisions.
Spreadsheet Development. Spreadsheet
development is about supporting spreadsheet
Spreadsheet Analytics Web Site: www.usfca.edu/bps/spreadsheet-analytics
Just what are the tools and re-
sources available to the spreadsheet
analyst who wants to be more pro-
ductive? It turns out that it is a tricky
question. There is an overwhelming
amount of material on the Web relat-
ed to spreadsheets, including add-
ins, templates, stand-alone software,
server-based software, checkers,
compilers, Web-based spreadsheet
deployment tools, methodologies,
books, courses, conferences, Web
sites and bagel slicers (well, maybe
not bagel slicers). These resources
address a dizzying array of problems
and opportunities. There is wide
range in quality, from commercial
software with full tech support and
detailed user manuals to academic
**       u      : G : ER M B < L F : @ : S B G > ' < H F
programmers to be productive and to work
well in organizations. This is an important issue
for people who spend a lot of time in spread-
sheets, and it should be an important issue for
the people who manage them.
The most exciting news has to do with
deploying spreadsheet applications. By
“spreadsheet application” I mean software
written by one person for use by others.
When done naively, this inevitably leads to
headaches and confusion. A terrific solu-
tion is to upload your spreadsheet app to a
server and provide a simple interface. Users
anywhere in the world can enter their data
tools that provide basic functional-
ity but little support … and every-
thing in between.
We launched a Web site called
“Spreadsheet Analytics: Resources
for Spreadsheet Analysts” to help
make sense of it all. Because of the
complexity of the spreadsheet ana-
lytic space, we organized the avail-
able resources into eight top-level
categories and 30 detailed catego-
ries. This makes it much easier for
an analyst to find what she is look-
ing for. It can be an eye-opening ex-
perience to scan all 30 categories
and see the amazing diversity of
spreadsheet analytic issues. Here
are the eight top-level categories:
• Model-Driven Analytics
and then, after the spreadsheet runs invis-
ibly, see the outputs from the spreadsheet
computations. This is the way of the future,
and if you’re e-mailing lots of spreadsheets
to lots of people, you need to learn about
this. Our survey found that several vendors
plus Microsoft are providing variations on
this service. In addition, some vendors offer
a “spreadsheet compiler” that locks away
the code inside an EXE or DLL so the users
can use the spreadsheet but can’t see or
change the formulas.
The other big news has to do with spread-
sheet development practices. Business
• Data-Driven Analytics/
Business Intelligence
• Business Function Analytics
• Data Management
• Reporting
• Development
• Control & Compliance
• Information Sources
The Spreadsheet Analytics Web
site focuses on commercial-quali-
ty products that are backed up by a
company that is likely to be around
for a while, with someone to talk to if
you need help. We made judgments
to exclude products that didn’t look
fully developed or weren’t backed
up by a company. We filtered out the
freeware and academic products, with
exceptions for areas where there are
schools do almost nothing in this realm, and
in too many companies the attitude is, “Here’s
your computer and your spreadsheet, you
figure it out.” It doesn’t need to be this way
anymore. Spreadsheet programming is
professionalizing and some companies are
enjoying dramatic improvements in produc-
tivity and maintainability. If you build large
financial models, our survey found detailed
formal development methodologies and you
might want to invest in learning one of them
or at least take a look at Jonathan Swan’s
“Practical Financial Modelling” (Swan,
2008). For the general business modeler,
no commercial products or where the
product has something special go-
ing on. We included an open-source
product if we felt there was a com-
munity standing behind it.
The Spreadsheet Analytics Web
site is non-commercial with no ad-
vertising and no selling commis-
sions. We didn’t test these products
and don’t endorse any of them. Re-
member, your mileage will vary. The
Spreadsheet Analytics Web site is
provided as a public service by the
Business Analytics Program of the
University of San Francisco’s School
of Business and Professional Stud-
ies. Go take a look at www.usfca.
edu/bps/spreadsheet-analytics.
– Thomas A. Grossman
: G : ER M B < L      u      F : R ( C N G >  + )*)
L I K > : = L A > > M  : G : ER M B < L
there are some terrific books about how
to approach the modeling/thinking as well
as the spreadsheet engineering, including
“Spreadsheet Modeling Best Practices”
(Read and Batson, 1999) and “Modeling
for Insight” (Powell and Batt, 2008).
According to the survey there are sev-
eral tools, some very sophisticated, for
spreadsheet checking and auditing. These
tools identify anomalies in the formulas
and flag things that you should be con-
cerned about. Some of the more powerful
tools provide model maps, summary data
and stuff you never knew you needed until
you saw it. One of these should be in the
bat utility belt of every spreadsheet super-
hero. The survey turned up handy “diff”
tools for spreadsheet comparison that tell
you what’s changed in two versions of the
same spreadsheet; like yellow sticky notes,
these are indispensable when you need
one. The survey also found a grab bag of
spreadsheet productivity tools that claim to
help you work smarter and faster.
Control & Compliance. For someone
keen on analytics, control and compli-
ance is not obviously exciting. But there
are important things going on in this area.
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE!
Just visit: http://analytics.informs.org/
*+       u      : G : ER M B < L F : @ : S B G > ' < H F
Courtesy of the Sarbanes-Oxley legisla-
tion, the U.S. government requires that
anything in the financial reporting loop has
to be locked down. This means no spread-
sheets or only spreadsheets with tight
controls. Risk managers want to make
sure that spreadsheets used for high-val-
ue or high-risk decisions (e.g., pricing of
financial derivatives that lead to million- or
billion-dollar bets) are bulletproof.
This requires a certain amount of man-
agerial commitment and process skills,
and no tool is going to help with that. It
also requires getting an inventory of the
spreadsheets in your organization and de-
termining which need managerial attention
or control and which don’t. That can be a
challenge, as some servers have as many
as 15 million (that’s not a typo: 15,000,000)
spreadsheets. This requires a tool that
looks at everything on a server or at every
computer on a network and gets a snap-
shot of all the spreadsheets, as well as the
spreadsheets that sneaky users may have
tried to disguise as something else. Our
survey found powerful tools for spread-
sheet discovery and management that
help with this. They’re not cheap, but they
sure are impressive. They can also provide
things like access control, change logs, ver-
sion control and other capabilities that are
routinely provided for professional software
developers. But then again, if you’re using
L I K > : = L A > > M  : G : ER M B < L

spreadsheets to run your company, shouldn’t
you treat and support (and manage and con-
trol) your spreadsheet analysts as the profes-
sional software developers that they are?
For medical or pharmaceutical re-
search, there’s another angle. The FDA
has strict requirements on data integrity.
warehousing systems. It is an open ques-
tion when CIOs, CFOs and other senior
managers recognize the amount of highly
paid time that is invested in spreadsheet
model assets. If and when they do, expect
a surge of interest in development process-
es and productivity.
the open source OpenOffice.org products
might yet take off.
For more details on the categories,
products, information sources and “beyond
the spreadsheet,” visit www.usfca.edu/bps/
spreadsheet-analytics. ❙
Thomas A. Grossman (tagrossman@usfca.edu) is an
REFERENCES
1. Conway, D. G. and Ragsdale, C. T., 1997, “Modeling
optimization problems in the unstructured world of
spreadsheets,” Omega, Vol. 25. No. 3, pp. 313-322.
2. Frontline Systems, 2009, “Optimization Problem
Types Overview,” www.solver.com/probtype.htm,
accessed April 1, 2010.
3. Powell, S. and Baker, K., 2009, “Management
Science: The Art of Modeling with Spreadsheets,” 2nd
edition, John Wiley & Sons.

Spreadsheets can still be used in medical
research, but they need to be locked up
inside a virtual vault. Products that can
do this for you are available.
We now have good tools for discov-
ering all the spreadsheets on a server. It
should not be too long before there are
good tools for mapping spreadsheet in-
formation systems as a set of interlinked
associate professor of Business Analytics at the University
of San Francisco. He teaches courses in Spreadsheet
Analytics, Business Analytics and Spreadsheet Engineering &
Management. He is developing a new MBA major in Business
Analytics. He does research on how spreadsheets are used
in organizations and how they can be used better. Web site:
http://web.usfca.edu/bps/faculty/Thomas_Grossman/
4. Powell, S. and R. Batt, 2008, “Modeling for Insight: A
Master Class for Business Analysts,” John Wiley & Sons.
5. Read, N. and J. Batson, 1999, “Spreadsheet
Modelling Best Practice,” www.eusprig.org/smbp.pdf,
accessed March 29, 2010.
6. Swan, J., 2008, “Practical Financial Modelling, Second
Edition: A Guide to Current Practice,” CIMA Publishing.

'32'097-327 spreadsheets. By connecting these maps

THE QUALITY OF model-driven spreadsheet
analytic tools is now very high. The new
spreadsheet solvers have extraordinary
capabilities to handle very large models, in-
cluding spreadsheet models programmed
in ways that confuse some traditional algo-
rithms. Furthermore, if the analyst learns to
design his spreadsheet to keep the solver
happy, then very large and complex models
can be successfully optimized. (There are
no satisfactory non-techie guidelines for
this yet, but you can get some insight from
Frontline Systems 2009, Conway & Rags-
dale 1997 or sections 10.3-10.4 of Powell
& Baker 2009.)
I expect to see growing awareness and
sophistication in the management of spread-
sheet information systems, especially after
companies finish getting their data house
in order via business intelligence and data
to source code management tools (which
in principle could be adapted to spread-
sheets) and using Web-based spread-
sheet deployment services, there is clear
potential for professional-quality man-
agement of spreadsheet systems without
having to port them to other languages.
The big open question in the spread-
sheet analytics space is, of course, Google.
Where is Google going with Google Spread-
sheet? Google Spreadsheet is currently an
analytic wimp. Google has not released a
product roadmap, so we don’t know its in-
tentions. They are steadily rolling out new
spreadsheet features. A small eco-system
of add-in providers is emerging. They have
the potential to provide significant analytic
functionality to Google Spreadsheet. It will
be fascinating to see what Google does
and how Microsoft responds. In addition,

*,       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      F : R ( C N G >  + )*)

;NLBG>LL<A:EE>G@>L
M^qm:gZermb\l3
MphPhke]l<heeb]^
Bgm^kl^\mbhgh_mph]blmbg\m]bl\biebg^l`e^Zgl
bglb`am_khffZllbo^Zfhngmlh_nglmkn\mnk^]]ZmZ'
;RD:MARE:G@>
“TEXT ANALYTICS” could be
8 the title of the latest mash-
up for business analysts. It
represents the intersections
of two typically distinct disciplines. In our
high school and college years, the core
graduation requirements were “math and
science” and “English and history.” In
fact, the skills measured on the College
Board exams are categorized as math-
ematics and reading/writing skills. Most
people are stronger in one discipline than
*-       u      : G : ER M B < L F : @ : S B G > ' < H F
the other. To simplify, I’ll use the classic
trope, “There are two kinds of people in
this world.” There are those who favor
words (text) and those who favor math
(analytics). Text analytics brings the two
disciplines together, creating a collision
of two worlds to help businesses thrive
and compete like never before.
The last few years have shown an in-
creasing adoption of analytics-powered
applications for better decision-making as
described in “Competing on Analytics” by
Thomas H. Davenport and Jeanne G. Harris
and “The New Know: Innovation Powered
by Analytics” by Thornton May. Organiza-
tions are utilizing data more strategically
to uncover hidden relationships to predict
what’s likely to happen in the future and take
advantage of that foresight. Until now, the
majority of data examined in these analyses
has been numerical or categorical data that
is often housed in relational databases.
;,%8-78)<8%2%0=8-'7#
AN ESTIMATED 75 percent or more of an
organization’s data is unstructured (im-
ages, content of Web documents, stan-
dard documents, audio, video, emails,
call center or claims notes, etc.). The vol-
ume at which this type of data is growing
is alarming. We are constantly barraged
with digital content in the form of e-mail,
news articles, contracts, privacy notices,
marketing offers, tweets, blog posts and
text messages. What business value
could be gained if analytical-powered ap-
plications could be augmented with infor-
mation gleaned from this freeform text?
In the Alta Plana research study “Text
Analytics 2009,” Seth Grimes defines text
analytics as the use of computer software
to automate:
• Annotation and information extraction
from text – entities, concepts, topics,
facts and attitudes.
• Analysis of annotated and extracted
information.
• Document processing – retrieval,
categorization and classification, and
derivation of business insight from
textual sources.
The most important piece of this
definition is “the derivation of business
insight.” In fact, that is the value of busi-
ness analytics in general, regardless of
the source of the data (textual, numeri-
cal, categorical, etc).
With text analytics, we are able to ex-
tract meaning out of large quantities of
various types of textual information. We
combine it with structured data within an
automated and repeatable process to en-
hance business insight.
8)<8%2%0=8-'7%((6)77)7
&97-2)77',%00)2+)7
TEXT ANALYTICS addresses two major
business challenges. The first is infor-
mation organization and “findability” of
the content within documents. The other
is discovery of trends and patterns to al-
low foresight from textual information.
-2*361%8-3236+%2->%8-32 %2(
%'')77°-1463:-2+7)%6',
WHEN WE THINK about finding informa-
tion, the first thing that comes to mind
is search. As any millennial will tell you,
: G : ER M B < L      u      F : R ( C N G >  + )*)
M > Q M  : G : ER M B < L
there’s nothing you need to remember
– you just Google it!
We’re all suffering from information
overload. We can’t consume the volume
of information that we receive every
day, so if we think it might be valuable,
we try to file it in a safe place for use at
some time in the future. When the time
comes to retrieve bits of digital informa-
tion previously stored by us or by others
in our organization, we search. Gart-
ner estimates that information workers
spend up to 30 percent of their working
day looking for information they need.
The content in our organizations’ doc-
uments represents knowledge from our
business and may include customer or
supplier information, operational informa-
tion and learning over time. It may contain
facts, opinions, discussions and intellec-
tual property. In our search, we may be
able to easily obtain information about
document properties (information about
the author or when the document was
created or modified, for example). But it
INFORMS COMMUNITIES
Search engines do a few basic things:
search parts of the Internet or the enter-
prise based on important words; keep an
index of the words they find and where
they find them; and allow users to query
words and match them to those that have
been put in the index.
:: Societies :: Sections :: Fora :: Chapters :: Student Chapters ::
Most large companies have webmas-
ters who focus much of their time asso-
Analytics readers have told us they have many and varied areas of interest.
ciating words with documents (tagging
At INFORMS, we have over 100 Community interest areas that can provide you with the particular
them or creating metadata) on their Web
information you need.
pages in a way that allows them to be
Here are just some of the Community web sites you can visit to receive newsletters, list serves,
indexed by Internet search engines and
discussion boards, specialized conferences, tools of the trade, selected journals, specialized awards,
ranked so they appear on the first or sec- and access to the sometimes hidden Analytics job market. Some of this information is free to the
ond page of the search results. With doc- public; some is not.
uments inside the enterprise firewalls, the (Note: not all INFORMS Communities offer all of the above benefits.)
same rigor is typically not applied to doc- Please visit today!
ument tagging to make enterprise search · Applied Probability · Marketing Science
engines more efficient, so information · Computing · Military Applications
tends to be more difficult to find. The pri- · Decision Analysis · Optimization
· Information Systems · Simulation
orities of most businesses are directed
· Manufacturing & Service Operations Management · Transportation Science and Logistics
toward generating revenue and custom-
er-facing projects rather than more inter-
nally focused activities. In these days of · Artificial Intelligence · Location Analysis
· Aviation Applications · Organization Science
economic instability, there has been a re-
· Behavioral Process Management · Public Programs3ERVICESAND.EEDS
is much more difficult to determine what newed focus on worker productivity and · CPMS: The College on the Practice of · Quality, Statistics, and Reliability
the document is about without opening it knowledge retention and management. Management Science · Railway Applications
up and reading it. Enterprise search has again become · Data Mining · Revenue Management and Pricing
· eBusiness · Service Science
a priority for many organizations. The
· Energy, Natural Resources, and the Environment · SpORts
A^eiIkhfhm^:gZermb\l goal is to create processes to organize · Financial Services · SPRIG: Spreadsheet Productivity
the abundance of documents, spread- · Group Decision and Negotiation Research Interest Group
It’s fast and it’s easy! Visit: · Health Applications · Technology Management
sheets, presentations and e-mail that is
http://analytics.informs.org/button.html · Telecommunications
being generated, so that the information

*.       u      : G : ER M B < L F : @ : S B G > ' < H F

www.informs.org0ARTICIPATE
)N
A
#OMMUNITY
M > Q M  : G : ER M B < L
contained within them is more easily acces-
sible by those who need it. Taking the lead
from lessons learned on the Internet, the
key is associating words with documents
that provide information about the content
that resides within them. Better search with-
in the enterprise can also be enabled with
better strategies for tagging the content.
Tools for automatically categorizing content
that create and maintain rich metadata as-
sociated with the documents are becoming
a central piece of a successful enterprise
search deployment.
8,)463')773*%2%0=>-2+8)<8
WHILE COLLECTING and searching for in-
formation is very important, it’s only the
first step in gaining business value.
The process of performing analysis on
text to discover insights turns out to be sim-
ilar to analyzing traditional data types. First,
you need to explore the documents. This
might be in the form of simple word counts
in a document collection or manually cre-
ating topic areas to categorize documents
by reading a sample of them. For example,
what are the major types of issues that
have been identified in recent automobile
warranty claims (brake failure, air bag de-
ployment, engine failure, sudden accelera-
tion)? As part of the exploration effort, you
will likely find misspelled or abbreviated
words, acronyms or slang terms.
*/       u      : G : ER M B < L F : @ : S B G > ' < H F
Prior to more advanced analysis or au-
tomated categorization, the text will likely
need to be preprocessed to address data
quality issues and standardization so the
analysis will produce more accurate re-
sults. As in traditional types of analysis,
up to 80 percent of the time can be spent
in the data preparation phase of a proj-
ect. In addition to correcting misspelled
words, much of the data preparation in-
volves standardization or transformation
to a consistent set of terms or identifying
specific ideas or concepts. For instance,
you might want to standardize “LOL” to
“laugh out loud,” or expand abbreviations
in patients’ medical files from “upr” to “up-
dated patient record” or “lvm” to “left voice
mail” for better consistency in analysis.
The contact center may be conducting
a loss-prevention campaign to custom-
ers who are delinquent on payments. It
may be interested in tracking how many
attempts were made to contact each cus-
tomer. The contact center would want to
be able to associate “lvm” and “left voice
mail” both as unsuccessful attempts to
contact. It might also want to classify
“no answer,” “bad number” and “phone
no longer in service” with unsuccess-
ful attempts. In this manner, synonym
lists or business rules would be gener-
ated to match against text files to identify
various results that would be categorized
as “unsuccessful attempts to contact
customer.”
Categorizing the content of docu-
ments effectively relies heavily on the
ability to understand the meaning of the
words in text, individually as well as in
context. Each language has its own nu-
ances, and within those languages jargon
exists that may be unique to companies,
to geographic areas, or among functional
areas across companies. Who among us
has not had to ask co-workers or busi-
ness associates to explain an unknown
acronym or saying? How should one in-
terpret “blue screen of death”? Should it
be associated with a computer “crash” or
“hang”? The human resources depart-
ment might talk about layoffs as “down-
sizing,” “right-sizing,” “employee turnover”
or “planned attrition.” In order to analyze
information about layoffs that resides in
text, HR needs to associate all terms that
might be used to refer to layoffs. Typical-
ly, subject matter experts are required to
identify and interpret the unique terminol-
ogy from a particular domain.
Language can be ambiguous. Words
in and of themselves may have different
meanings when used in the context of a
sentence. Some words can be used as
different parts of speech. In a warranty
claim, you may be interested in defects in
the “left” wheel of the automobile but not
concerned that the driver “left” the win-
dow open. The order of the words in the
sentence may also be significant. For an
insurance claim, it is extremely important
to know if “Driver 1 hit Driver 2” or “Driver
2 hit Driver 1.” Notice that the words are
identical, but the order makes a huge dif-
ference for the insurance company in as-
sessing fault for the accident.
Categorizing documents from informa-
tion contained within them can be achieved
through a combination of statistical models
and business rules. As with traditional mod-
el development, sample documents are
examined to train the models. Additional
documents are then processed to validate
the accuracy and precision of the model,
and finally new documents are evaluated
using the final model (scored). Models can
then be put into production for automated
processing of new documents as they ar-
rive. The models can be monitored for con-
tinued performance.
(-7'3:)6=
SOME BUSINESSES may have a predeter-
mined categorization structure that they
use to organize their documents. For in-
stance, the technical support center might
have a list of the top 10 product defects
identified through customer interactions.
Additional or emerging trends could be
identified through the use of text mining to
: G : ER M B < L      u      F : R ( C N G >  + )*)
M > Q M  : G : ER M B < L
discover unknown issues from patterns
discovered in the technicians’ notes. In
this manner, emerging defects could be
quickly identified before they become
top 10 issues, or previously unnoticed
defects that should have been in the
top 10 could be uncovered.
Associated terms or concepts such as
“unsuccessful attempts to contact cus-
tomer” can be used as input to text mining
models to refine the analysis and produce
better, more interpretable results. New
topics discovered by text mining models
can augment rule-based categorization
models. The processes can work in har-
mony to continually learn.
Often text data such as surveys,
product reviews and call center notes
are accompanied by numeric data (3
stars out of 5), demographic data (city,
state, gender, income level), customer
purchase amount or monthly service
usage data. By integrating the text
into the predictive model, the models
can often produce better results. In
the case of identifying customer churn
(customers likely to go to a competitor
or drop your product or service), strong
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE!
Just visit: http://analytics.informs.org/
*0       u      : G : ER M B < L F : @ : S B G > ' < H F
indicator words or concepts could be
identified as a significant predictor of
behavior for a specific customer seg-
ment. The model might indicate that if
a “platinum level” customer mentions a
competitor’s lower price to the service-
center representative, he is 25 percent
more likely to cancel his service with
your company.
7)28-1)28%2%0=7-7
A GROWING AREA of interest for many
firms is the understanding of what the
market (customers, analysts, or key
opinion leaders) is saying about their
products and services. Many refer to
this area of analysis as sentiment anal-
ysis. They want to understand more
about people’s opinions, attitudes and
emotions when discussing their prod-
ucts, services or overall brand. Textual
information is typically accumulated in
the form of surveys, technical support
notes or reviews on third-party Web
sites (such as Amazon.com). Studies
show that consumers are much more
trusting of other users’ opinions than
of the marketing collateral produced
by the manufacturer.
From the company perspective,
listening and analyzing what people
are saying about your products and
M > Q M  : G : ER M B < L
services is the first step in creating a di-
alogue with that audience by listening,
learning and then engaging with them.
This dialogue can better inform targeted
marketing initiatives to customers and
prospects, enabling the organization
to communicate at a significantly lower
cost with increased speed and effec-
tiveness than traditional marketing. It
can also enable a more rapid response
to perceived customer issues and com-
petitive threats.
Sentiment analysis builds on both
human-defined business rules and
computer-generated statistical models
to automatically identify positive and
negative sentiments expressed in free-
form text. Keywords are identified in
the context of the text to classify posi-
tive and negative concepts. Text might
explicitly contain words that identify the
mood or sentiment, for instance, “mad”
or “angry.” You may also be able to infer
attitudes and emotions based on action
words or phrases such as “screamed”
or “hung up.” Some phrases might be
more subtle, like “not impressed by”
or “answers were unsatisfactory.” And
some words can be ambiguous until as-
sociated with a product feature or ser-
vice. In the context of a computer, “long”
battery life would be a positive attribute,
*1       u      : G : ER M B < L F : @ : S B G > ' < H F
but “long” boot time would be a nega- near future, rather than only the subject collided, and clever decision-makers
tive attribute. of research papers. will begin to see the biggest bang for
Early projects with sentiment analy- Today text analytics provides forward- their buck. ❙
sis may start by categorizing positive thinking organizations with a framework Kathy Lange (Kathy.Lange@sas.com) is a senior director
and negative dimensions of the feed- to maximize the value of information in SAS’ Business Analytics Practice. She has more than
25 years of experience selling and implementing analytics
back and counting how many responses within large quantities of text. This tech- solutions. The SAS Business Analytics practice assists
were positive or negative with regard to nology helps automate the process by customers in defining their business problems and crafting
strategies for solving those problems with integrated
a particular product, feature or service. extracting relevant information and in- SAS solutions that include business intelligence, data
More advanced projects may track the terpreting, mining and structuring infor- integration and advanced analytics.
sentiment over time to track trends and mation to improve findability and reveal Copyright © 2010, SAS Institute Inc. All rights reserved.
patterns or compare the sentiment of a patterns, sentiments and relationships A version of this article appeared in BeyeNetwork.
Reprinted with permission.
company’s product with the sentiment among documents. Two worlds have
of the competitor’s product.
Public relations firms use sentiment
analysis to track positive and negative
mentions of their company, company of-
ficials or products in the media or other
publications (pharmaceutical brands,
for instance). They try to determine if
their marketing messages are being in-
terpreted positively or negatively and
what effect that has on their brand in
the marketplace.
'32'097-32
BUSINESS LEADERS COMMITTED to fact-
based decision-making are recognizing
the power hidden in text to yield insight
into marketing, customer service, public
relations, product innovation and compe-
tition. Techniques for analyzing voice and
video and other unstructured content will
be more commercially available in the
: G : ER M B < L      u      F : R ( C N G >  + )*)
A
  ! 
; 


  

 

          >  .   .   

           *      
      
     
 
         
   /      
      

 
 
        !    * 

     
'+-   
2'-   

            
       
           !       
            
     
   
"  *         
  
  . 0      3  
         1
   
              
 4      )"5 
    
      
       
        6  768     
    
  
 6)"5)9:A;9<=  
      
      " ;    
  
               
        
>  
     9 .
     
?@ 

A A     B!#$%

&'
 '(
 
       . 
   ) *  #&+

    ,-

   


##### 
###
###
#
####
#
#
# # 
# # # #  # # #  # # !# # # 
"# #  
 
#
L N I I ER  < A : B G    = > F : G =  F : G : @ > F > G M
:]oZg\^]:gZermb\l
_hkLZe^l
Hi^kZmbhglIeZggbg`
>__^\mbo^LHI]^\blbhg&fZdbg`\ZgÌm[^
Z\ab^o^]pbmahnmbglb`aml_khfZgZermb\l'
;RC>K>FR?'LA:IBKH
SALES & OPERATIONS PLANNING
7 is a methodology for coordi-
nating supply chain and de-
mand management decisions
[7]. Many companies, including Procter &
Gamble, Motorola and Caterpillar, have
implemented S&OP processes. Thus far,
these processes rely on software that
acquires and integrates data relevant
to S&OP but do not employ advanced
analytics – optimization models integrat-
ed with descriptive models – to support
S&OP decision-making.
+ )       u      : G : ER M B < L F : @ : S B G > ' < H F
Despite the many S&OP system im-
plementations, recent surveys [1, 6] indi-
cate that improvements due to them have
stalled. Although reasons for the slowdown
can be debated, a strong argument is that
effective S&OP decision-making cannot be
achieved without insights from advanced
analytics. Given accurate data inputs, a
properly implemented optimization model
can unravel the complex interactions and
ripple effects across sourcing, production,
distribution, inventory and demand man-
agement decisions that make S&OP difficult
Figure 1: S&OP analytic schema.
and important. Companies also need to
implement new S&OP processes that de-
velop and exploit the insights provided by
advanced analytics.
As shown in Figure 1, advanced ana-
lytics for S&OP are comprised of a supply
chain network optimization (SCNO) mod-
el and a collection of descriptive models.
The SCNO models are linear program-
ming (LP) or mixed integer programming
(MIP) models, which are the only tools
capable of holistically analyzing com-
plex S&OP problems. A SCNO model for
a specific application defines the S&OP
decision database that contains both
data inputs to the optimization model and
the optimal solutions it computes. Typi-
cal planning horizons addressed by the
SCNO model are 12 to 24 months.
Descriptive models are used to deter-
mine demand forecasts, manufacturing
and distribution costs and cost relation-
ships, future raw material costs and avail-
abilities, and a variety of other parameters
and relationships required by the S&OP
decision database. Various aggregations of
: G : ER M B < L      u      F : R ( C N G >  + )*)
L : E > L    H I > K :M B H G L  I E : G G B G @
2010 INFORMS
Annual Meeting
*0 ( ,66
the transactional data, such as product with advanced analytics is the critical
/-.$)6 2-
aggregations or aggregations of custom- technological development needed to
/-.$)6*)0 ).$*)6 ). ,6
6$'.*)6
/-.$)
ers who are geographically close, may facilitate and promote the next genera-
be performed. As shown, the decision tion of S&OP tools and practices.  -- 
-

*# (( %$-" $-,--
database will also contain inputs from
managers expressing their judgments '%7)789(=&))61%29*%'896)6
about performance constraints to be im- FIGURE 2 PROVIDES an overview of the
posed on S&OP solutions; for example, network underlying an SCNO model con-
constraints reflecting acceptable custom- structed for a large beer manufacturer in

$'"-
 ' "$',-(( %$(-
 '
er service levels, minimal manufacturing a country in the southern hemisphere. *).#)6, '($,6*+*0
production levels at selected plants, and The model was implemented to plan for # 6)$0 ,-$.36*!6 2-6.6
/-.$) # 6)$0 ,-$.36*!6 2-6.6
/-.$)
%, ($'/. 2- / '($, ($'/. 2- /
the sole sourcing of markets. peak domestic demands during the sum-
'%'#-
 ' ') -'%'#-
 '
In a recent survey of 180 compa- mer season that includes the Christmas
0$6*,.*)  0$)6/,()6
# 6)$0 ,-$.36*!6 2-6.6
/-.$) 22*)*$'
nies in North America [6], ease of use holiday. As shown, the company’s five (*,.*) ($'/. 2- / & 0$)!/,() 22*)(*$'*(
and integration with ERP systems were plants manufacture and package ap- $+ )-(( %$(-
%.
 '( ''$#$)(-
%.
 '(
judged to be the two most important cri- proximately 100 products that are trans-
)).,(6'&,$-#))  -6,) -6
# 6)$0 ,-$.36*!6 2-6.6
/-.$) # 6)$0 ,-$.36*!6 2-6.6
/-.$)
teria for acquiring S&OP tools. As dis- ported through the distribution network to )). ($'/. 2- / 1,) - ($'/. 2- /

cussed in [4], both these criteria can be the markets. The company’s distribution .,%)6/.( ( -66 3 ,
 2-6

6)$0 ,-$.3 # 6)$0 ,-$.36*!6 2-6.6
/-.$)
met by advanced analytics that are im- centers (DCs) are organized in three lev- "/.( .(/ / %3 , (*(-/. 2- /

bedded in a business intelligence sys- els: plant DCs, regional DCs, and local &%$(%'-(( %$(-
 ' $)') +-(( %$(-
%.
 '(
0$'$)6*+*0 , )6($'*1$.4
tem to: (1) facilitate efficient extraction, DCs. In addition, the company has ar-  2-6.. 6)$0 ,-$.3 *,.#1 -. ,)6)$0 ,-$.3
$+ .2-..  / &-($'*1$.4 )*,.#1 -. ,) /
transformation and loading of inputs rangements with third-party warehouses
$'65 .$)&3
from ERP systems to the S&OP deci- to handle surge inventories for peak de-
%$)' *)-(( %$(-
 '
 2-6

6)$0 ,-$.3
/(,6/.#/,()
-$' .(/ /
sion database; and (2) report metrics mand. These DCs are designated with # 6)$0 ,-$.36*!6 2-6.6
/-.$)
&/(, /-.$)/. 2- /
'-'%'#-  (%$-
 '
and other results from the descriptive an S in Figure 2. 6,$6$& '
*)%' "(-
 '
and SCNO models in easy to use dash- The optimization model was construct- # 6)$0 ,-$.36*!6 2-6.6
/-.$)
*#)6- ) $)
$& ' ($'/. 2- /
# 6)$0 ,-$.36*!6 2-6.6
/-.$)
boards with drill down capabilities. The ed by merging 12 monthly sub-models %#- ($'/. 2- /
*$' ( $-
 '
merger of business intelligence systems based on the network in Figure 2 plus -  (%$
,#)6/.)*"'/
# 6)$0 ,-$.36*!6 2-6.6
/-.$)
inventory balance equations linking con- /'6
6 )- )
,#)& ($'/. 2- /
# 6)$0 ,-$.36*!6 2-6.6
/-.$)
A^eiIkhfhm^:gZermb\l secutive months. Forecasted market de- +% )- ) ($'/. 2- /
mand was backed up to demand profiles
It’s fast and it’s easy! Visit:
for the local DCs; hence the dotted lines
http://analytics.informs.org/button.html
in the figure. This construction assumed

+*       u      : G : ER M B < L F : @ : S B G > ' < H F www.informs.org/Conf/Austin2010 $!(-)%-%*'-&%$(%'

L : E > L    H I > K :M B H G L  I E : G G B G @
Figure 2: S&OP network of a large beer manufacturer.
that the customers and markets to be ef-
ficiently served by each DC were known.
The resulting LP model had approximate-
ly 50,000 rows and 125,000 columns and
could be generated and optimized in just a
few minutes on a high-end PC.
Starting early in each calendar year, the
S&OP model was run every month using
updated data about inventories, forecasted
demands, manufacturing capacities and
other parameters. The objective function
was to minimize the total supply costs of
meeting forecasted demand. It provided
S&OP solutions that significantly lowered
these costs compared to earlier years be-
fore the advanced analytics were used.
+ +       u      : G : ER M B < L F : @ : S B G > ' < H F
'3781-2-1->%8-32:7
463*-81%<-1->%8-32
AT THIS STAGE IN THE DEVELOPMENT of advanced
analytics for S&OP, an unresolved issue when
constructing an SCNO model is: Should a
company use an objective function that
• minimizes the total cost over the
planning horizon of meeting sales
which are treated as fixed and given,
or
• maximizes the net profit over the
planning horizon from sales – some of
which are treated as discretionary and
may incur avoidable marketing, sales,
product customization or non-standard
service costs ?
The total cost minimization approach
is easier because it is simpler and directly
extends current and past modeling prac-
tices regarding supply chain management.
Demand management may be handled by
manipulation of inventory balance equa-
tions for each product (or product family)
in the SCNO model. For each product and
at each location where the product may be
held, this familiar equation is
Ending inventory = beginning invento-
ry + product received or made – product
shipped or sold
Cost minimization analysis may implic-
itly evaluate net profits by flexibility built into
the model regarding if and when certain in-
termediate or finished products are shipped
or by running multiple scenarios with vary-
ing demand profiles for finished products.
The net profit maximization approach
provides more explicit modeling of integrat-
ed supply chain and demand management
decisions and therefore provides more com-
prehensive S&OP analysis. It is more chal-
lenging to construct these models and to
use them once they have been validated.
It is important to recognize that demand
management modeling depends heavily
on the industry. For example, companies
that manufacture commodities such as
chemicals or building materials often face
decisions about requests from customers
for customized versions of their products.
The company must decide how much to
charge for customized products, where the
products will be made and when the prod-
ucts can be delivered. Typically, the com-
pany faces a portfolio of such requests and
should employ integrated decision-making
methods that simultaneously evaluate all
such requests received over a meaningful
period of time. By “discretizing” the pos-
sible selling price range for each custom-
ized order, zero-one decision variables in
an SCNO model can select if, where and
when each order might be manufactured
and at what net profit.
Consumer package goods companies
are candidates for S&OP models that in-
tegrate supply chain plans with marketing
and sales plans for the coming months.
Market response models that predict mar-
ket share and market volume as functions
of product prices, advertising, promotion
and sales force effort can be linked to sup-
ply chain models that determine the costs
of supplying the demand generated by the
marketing plans. The objective of the inte-
grated optimization model is to
maximize net profits = gross profits
from product sales
– marketing and sales costs – supply
chain costs
See [2] for details about market re-
sponse models; see [3] for mathematical
: G : ER M B < L      u      F : R ( C N G >  + )*)
L : E > L    H I > K :M B H G L  I E : G G B G @
programming methods that integrate
market response models with supply
chain optimization models.
8,)7
34()'-7-32(%8%&%7)
ALTHOUGH THE FORM and content of
the S&OP decision database will de-
pend on the nature of the company’s
business and on the software they ac-
quire or implement to support decision-
making, some general principles have
emerged as the result of successful
SCNO applications. The input data
for the SCNO model is typically parti-
tioned into master data files describing
the structure of the company’s supply
chain and scenario data files contain-
ing numerical values for relevant costs,
activities, resource constraints and fi-
nal product demands. The master data
files are permanent or semi-permanent
with changes made infrequently. By
contrast, multiple scenario data files
are created and analyzed as part of
each month’s planning process. Out-
puts from SCNO model optimization of
each scenario are stored in extensions
of the scenario data files.
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE!
Just visit: http://analytics.informs.org/
+ ,       u      : G : ER M B < L F : @ : S B G > ' < H F
Master data files list the names
of the company’s suppliers, facilities
(plants and DCs), and markets and their
geographical coordinates. They list the
names of the raw materials, work-in-
process and finished products that flow
through the network, along with their
units of measure. They also list the
names of: the processes at each facil-
ity; the bills-of-materials used there to
transform products; and resources that
are consumed in carrying out these
transformations. Geographical connec-
tions are provided by lists of the links
in the in-bound (suppliers to facilities),
inter-facility (facilities to facilities), and
out-bound networks (facilities to mar-
kets). Finally, the master data files list
other structural data such as available
transportation modes and the number
of time periods in the SCNO model.
The scenario data files provide nu-
merical parameter values for each pe-
riod in the planning horizon including:
• raw material capacities and costs for
each supplier,
• facility indirect costs,
• process, product transformation,
and resource capacities and costs at
each facility,
• inventory throughput and safety-
stock costs and capacities, and
• transportation costs and capacities.
L : E > L    H I > K :M B H G L  I E : G G B G @
For multi-national networks, the costs
may be specified in multiple currencies
with a currency exchange table for each
period. The scenario data may also include
managerial constraints on the shape of the
S&OP solution such as the number of plants
where a product may be made. Finally, nu-
merical data describing demand forecasts
by product and time period are also speci-
fied. These data may be expanded to incor-
porate net profit maximization options such
as product pricing.
(-71%280-2+
()'-7-321%/-2+7-037
MANAGERIAL DECISION-MAKING in Figure 1
refers to integrated decision-making among
supply chain and marketing and sales man-
agers. Advanced analytics and supporting
processes serve to dismantle decision-
making silos that are still present in many
companies. The SCNO model provides the
inspiration, or the abstract representation,
for analyzing the company’s S&OP prob-
lems in an integrated manner. The perspi-
ration, or hard work, for implementing the
advanced analytics is creating, validating
and updating the decision database that
feeds the SCNO model.
Additional effort is needed to devel-
op and maintain S&OP processes that
exploit insights provided by advanced
analytics about integrated decisions.
+-       u      : G : ER M B < L F : @ : S B G > ' < H F
For this purpose, the company needs to impose decisions if and when the dis- ,-)6%6',-'%040%22-2+
to create an S&OP Coordination Com- cussion becomes prolonged and counter- ;-8,%(:%2')(%2%0=8-'7
mittee and an S&OP Analysis Team. A productive. One can expect, or at least HIERARCHICAL PLANNING REFERS TO the
typical S&OP Coordination Committee hope, that contentious S&OP problem identification and application of consis-
will be comprised of the VP for Sup- solving will occur infrequently as the re- tent and effective strategic, tactical and
ply Chain Management, key managers sult of monthly planning exercises that operational plans. The discussion thus
from procurement, manufacturing, dis- look out many months in anticipating far has addressed the use of advanced
tribution, marketing and sales, plus the problems and finding effective resolu- analytics in a company to support tactical
head of the S&OP Analysis Team. The tions. Conflicts in team decision-making S&OP over the coming months. These
S&OP Analysis Team will be comprised can also be reduced by creating and ap- analytics can be transformed into and/
of knowledge engineers who know the plying managerial incentive plans that or linked to similar analytics that address
company’s business and, depending on promote integrated S&OP. strategic and operational planning.
the nature of that business, have rele-
vant, in-depth knowledge about analyt-
ics for demand forecasting, data mining,
supply chain management, managerial
accounting, mathematical programming
and marketing science.
Each month, the S&OP Analysis Team
updates the S&OP decision database
and, working with the S&OP Coordination
Committee, optimizes multiple scenarios
describing supply and demand manage-
ment plans for the coming months. These
scenarios are created in response to
feedback from S&OP managers to the
Coordination Committee regarding the
desirability and feasibility of plans sug-
gested by earlier scenarios. Although
discussion about each month’s revised
S&OP solutions for the coming months
should be encouraged, the Coordina-
tion Committee must be given authority
: G : ER M B < L      u      F : R ( C N G >  + )*)
L : E > L    H I > K :M B H G L  I E : G G B G @
Tactical S&OP analysis will often
reveal the need to strategically re-de-
sign the company’s supply chain net-
work over the coming years. Related
decisions regarding the company’s
portfolio of products or the acquisition
of a complementary company may
also require such analysis. A tacti-
cal SCNO model can be expanded to
address strategic design options by
re-defining the time periods and asso-
ciated data in the model and expand-
ing the decision database to describe
the new strategic options. Converse-
ly, a strategic network design model
can be readily converted to a tactical
S&OP model by fixing the design de-
cisions and re-defining the time peri-
ods and associated data.
When moving from tactical to op-
erational analysis, the challenge is
to translate aggregate and holistic
tactical plans to detailed operational
ones. Detailed models for inventory
control, production scheduling, per-
sonnel scheduling, vehicle routing,
etc. are available for examining oper-
ational plans. An S&OP optimization
A^eiIkhfhm^:gZermb\l
It’s fast and it’s easy! Visit:
http://analytics.informs.org/button.html
+.       u      : G : ER M B < L F : @ : S B G > ' < H F
model can set goals and constraints
for these models over the coming
days and weeks that reflect efficient
tactical plans for the coming months.
'%7)789(=4)6(-+%3
PERDIGAO, A BRAZILIAN FOOD manufac-
turer with more than $6 billion in yearly
revenue, produces poultry, pork, beef
and processed dairy products. Its prod-
ucts are sold in Brazil and in 110 coun-
tries around the world.
As discussed in [5], starting in 2003,
Perdigao began developing S&OP pro-
cesses to deal with many supply chain
challenges including:
• long lead times up to 18 months for
acquiring raw materials,
• tight constraints on slaughtering and
freezing capacity, To find an expert to help you,
• shelf-life constraints on finished
products,
log onto INFORMS O.R. Professional Database at
• no formal sales forecasting process http://www.scienceofbetter.org/find.
or analytics, and
• a push rather than a demand-driven
system for replenishing distribution
centers. INFORMS is the foremost association
Perdigao’s installed SAP system of O.R. and analytics experts. Our members
contained the transactional data needed
to generate and apply advanced ana-
literally wrote the book on how analytics and
lytics. The first step was the design and the principles of operations research are used
implementation of an S&OP process to improve corporate or governmental
spanning sales, marketing, logistics,
decision making.
http://www.informs.org
L : E > L    H I > K :M B H G L  I E : G G B G @
production, finance and procurement for
both international and domestic markets.
The goal of the new process was to ex-
tract insights from the sales organization
that could later be used in developing
analytics for demand forecasting. It also
promoted communication and a greater
sense of coordination between the sales
organization and those in production, lo-
gistics and purchasing.
The next step was to leverage Perdi-
gao’s ERP system to develop demand
forecasts and create a decision data-
base with timely and accurate data de-
scribing sales orders, distribution center
inventories, manufacturing and logistics
REFERENCES
1. Barrett, J. and S. Steutemann, 2010, “Conquering
the Seven Deadly Challenges of Sales and Operations
Planning,” AMR Research, January. Document available
at www.amrresearch.com.
2. Lillen, G. and A. Rangaswamy, 2004, “Marketing
Engineering,” Trafford Publishing.
3. Shapiro, J., 2007, “Modeling the Supply Chain,” 2nd
edition, Duxbury Press.
4. Shapiro, J., 2009, “Business Intelligence & Modeling
Systems Synergy,” Analytics, fall, pp. 18-23.
5. Tohamy, N., 2009, “Perdigao Builds a
Demand-Driven Supply Chain with Axia Value
Chain Consulting and SAP’s APO,” AMR
Research, November. Document available at
www.amrresearch.com.
6. Tohamy, N. and K. Carter, 2009, “S&OP
Technology Market Update: Picture Still Murkey,”
AMR Research, November. Document available at
www.amrresearch.com.
7. Wallace, T. and R. Stahl, 2008, “Sales and Operations
Planning: The How-to Handbook,”
3rd edition, T. F. Wallace & Co.
+ /       u      : G : ER M B < L F : @ : S B G > ' < H F
constraints appropriate to short-term and
tactical planning. Forecasting models
were also used to predict market prices
for the company’s products. The S&OP
transformation team then focused on dis-
tribution planning processes that would
change supply chain planning to one that
was demand-driven from the previous
one that was push-driven.
After three years transforming their
S&OP process, Perdigao took a large next
step in implementing a Supply Network
Planning Optimizer to address both short-
term (three months) and long-term (18
months) issues involving supply chain and
demand management decision-making
across Brazil. The Optimizer was a large-
scale LP model that addressed three sup-
pliers groups, 16 plants, 20 DCs, seven
markets, and two exporting ports. The ob-
jective function of this model was to maxi-
mize profit with constraints on product mix
to ensure long-term stability of total product
offerings. Constraints capturing shelf-life
and inventory balances were also incorpo-
rated in the optimizer model.
The company SKUs were modeled
directly (not aggregated) and numbered
600 to 1,000. As a result, some versions
of the model involved more than one
million constraints and several million
variables and required several hours to
be optimized. Modeling products at the
SKU level facilitated the computation of participants in Perdigao’s new S&OP
implementable short-term operational process employing advanced analytics
plans as well as analysis of intermedi- gained widespread user adoption. Repre-
ate tactical plans. However, the result- sentative results [5] include a 40 percent
ing model size inhibited the running of increase in inventory turns in the com-
multiple scenarios each month. Plan- pany’s domestic markets, a 4 percent
ning continuity was reflected in the main decrease in stockouts, and a 31 percent
scenario run each month by consensus decrease in shelf-life losses. ❙
among the S&OP managers in meetings Jeremy F. Shapiro (jfsinc@attglobal.net) is a professor
held prior to the run. emeritus in the Sloan School of Management at MIT, a
former co-director of MIT’s Operations Research Center
Following successful data validation and author of the textbook, “Modeling the Supply Chain.”
exercises and extensive user training,
WHERE IS YOUR CAREER LEADING YOU?
http://www.informs.org/continuinged
Introducing INFORMS Conti
Continuing Professional Education Database
Your one
one-stop
stop shop for O
O.R.
R semina
seminars, short courses, and workshops
offered throughout the world.
Have a course you would like to add? We want to hear from you!
INFORMS is committed to supporting the lifelong learning of operations research and analytics professionals.
: G : ER M B < L      u      F : R ( C N G >  + )*)
E > @ B L E : M B H G  : G : ER L B L
A^Zema<Zk^K^_hkf
AZlIZll^]'GhpPaZm8
Ik^]b\m^]hnm\hf^l3@nZk]^]himbfblf_hk`^g^kZein[eb\Zg]iZmb^gml4^q\^ee^gmikh`ghlbl_hka^Zema
\Zk^lrlm^flZgZerlml'
;R=HN@E:L:'L:FN>ELHG
CONGRESS’ LANDMARK health
' care reform legislation is
just the beginning of the
restructuring process. Now
health care systems analysts will see
huge opportunities, as everyone poten-
tially affected will need help to figure
out what the act actually provides and
how to respond.
The act is more than 2,000 pages
long, and it is overwhelmingly likely that
there is not one person in the entire
+ 0       u      : G : ER M B < L F : @ : S B G > ' < H F
country who fully comprehends it. This is
not an unusual situation for major legis-
lation. Senators and representatives rely
on analyses by the Congressional Bud-
get Office, legislative counsel, committee
staffs and, yes, trusted lobbyists to guide
their votes. Even a highly experienced
lawyer would have great difficulty reading
such a law and interpreting it, as literally
every word and clause has been or will
be construed by multiple courts, often in
conflicting ways.
Executive branch agencies will now
write implementing regulations that will
probably be several times as volumi-
nous as the act, and these regulations,
too, will be interpreted by numerous
and varied court cases and adminis-
trative proceedings. Some provisions
phase in over several years. In short,
you should dismiss as a raving babbler
or a liar anyone who assures you that
he or she knows exactly what this act
will do, good or bad.
Based on the Kaiser Family Foundation
summary of the act, Alain Enthoven’s prior
work, Tom Daschle’s book that many con-
sider a blueprint for the health care legis-
lation, a number of other sources and this
reporter/analyst’s own considerable experi-
ence in the subject over the past 20 years
(see references for all work cited above),
some predictions seem reasonably solid:
Many but not all of the current un-
insured will now be covered. The
: G : ER M B < L      u      F : R ( C N G >  + )*)
E > @ B L E : M B H G  : G : ER L B L
mandated coverage phases in over
several years, and one major enforce-
ment mechanism is that people will be
compelled to enroll when they seek
care. There are some exceptions to the
mandate, and several states have filed
lawsuits contesting the requirement.
People excluded from coverage be-
cause of pre-existing conditions will
now have more and better choices. This
provision takes effect immediately and is
one of the clearest consequences of the
act. Insurance companies will be inten-
sively analyzing how this may affect them
and what they can do about it.
People with good health insurance
will see their premiums rise. Health care
costs have been rising and will continue to
do so; premiums reflect this. Some of the
people who will now get coverage because
of the new law are high-risk and will be put
into a temporary high-risk pool, similar to
the assigned risk pool for automobile insur-
ance, cross-subsidized by other people,
and insurance companies will want to in-
crease their revenues accordingly.
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE!
Just visit: http://analytics.informs.org/
+1       u      : G : ER M B < L F : @ : S B G > ' < H F
Over time, premiums may rise more
slowly than they would have without
the new law. This is because, along with
some high-risk people, a large group of
young, relatively healthy people will also be
pushed into the insured pool, lowering the
average risk and cost. Economics predicts
that insurance companies will compete for
these desirable customers by offering more
attractive rates. The creation of health in-
surance exchanges, basically purchasing
pools that negotiate better deals for benefi-
ciaries, will push the market toward lower
rates – at least that’s what economic theory
predicts. Some states, in particular Califor-
nia as a result of operations research ana-
lyst Enthoven’s efforts there, have reported
success with this approach. Still, what will
happen on a national scale remains to be
seen.
Some taxes will rise, including some
not yet specified. Again, this would also
have happened without this legislation, and
it will be difficult if not impossible to deter-
mine whether the new legislation decreas-
es or increases the effect. As the economy
improves, deficit reduction will become in-
creasingly important and politically attrac-
tive, and Congress will respond.
There won’t be “death panels” mak-
ing life-or-death decisions case by
case, but there will be some rationing
of care. Actually, this happens now,
and it will continue to happen the same
way: some expensive procedures sim-
ply won’t be covered unless and until
their life-saving necessity can be clearly
demonstrated. What will change is that
some of these exclusions will become
more visible and, therefore, potential-
ly more widely debated and reviewed.
People with more money will continue
to have more choices – such as buying
the care they want, perhaps even trav-
eling outside the United States to do so
– than other people.
With or without the public option,
the nation will move toward a single set
of rules and procedures for payment.
The current hodge-podge of different re-
quirements for claims processing is one of
the big reasons that the American Medical
Association and the American Hospital As-
sociation, among other providers’ groups,
supported this legislation. Some hospitals
have to pay claims specialists to file their
claims with as many as 40 insurance carri-
ers, all with different rules and procedures.
Standardizing will save serious administra-
tive costs. As with Medicare now, standard-
izing rules and procedures can occur while
private insurance carriers continue to ad-
minister the claims.
Standardization of claims rules may
ease the long-delayed transition to
standard electronic patient records.
This improvement in information support
for medical decisions could, in turn, sub-
stantially reduce preventable deaths from
system errors, principally bad hand-overs
of information among providers. In 2000,
the Institute of Medicine, part of the Na-
tional Academy of Sciences, published a
study estimating that medical errors ac-
counted for nearly 100,000 preventable
deaths per year in the United States,
mostly attributable to system problems
despite reasonable care by all the indi-
viduals providing care and related func-
tions (such as labs and radiology). They
followed up with a second publication in
2001, recommending system improve-
ments to address these problems.
Another aspect of uncoordinated care
is polypharmacy, the use of multiple pre-
scription medications in combination,
with too little attention to possible in-
teractions. Polypharmacy is what killed
Anna Nicole Smith and Michael Jackson,
both of whom apparently took deliberate
steps (multiple doctors, multiple pharma-
cies) to sidestep warnings. Polypharma-
cy also most likely killed Brittany Murphy,
who does not appear to have received
adequate warnings or to have ignored or
evaded them. Safety testing is usually
: G : ER M B < L      u      F : R ( C N G >  + )*)
E > @ B L E : M B H G  : G : ER L B L
done one medication at a time, so inter-
actions can take quite some time to be-
come identified and publicized. This is a
growing problem, and information tech-
nology offers a promising answer.
There will be much more research
and assessment about comparative ef-
fectiveness of treatments. This is the
big payoff for analysts who have taken
the time to learn about medical out-
comes evaluation, assessment of ef-
fectiveness and risk of new treatments,
preventive care and other such topics.
The act specifies that the federal gov-
ernment will fund more research in this
area. Providers, regulators and insurers
will want to know what really works. Liti-
gation over “rationing” will hinge on this
question. Employers will want to know
how to tailor the health insurance pack-
ages they offer to get good perceived
value for restrained cost.
National security will improve
somewhat. The best defense against a
pandemic is prompt detection, which re-
quires the first people affected to seek
A^eiIkhfhm^:gZermb\l
It’s fast and it’s easy! Visit:
http://analytics.informs.org/button.html
+ 2       u      : G : ER M B < L F : @ : S B G > ' < H F
treatment quickly from providers who
are part of the reporting network. A de-
lay of as little as a week in detecting an
incipient epidemic can make a huge dif-
ference in the extent of its spread. The
magnitude of the effect of having most
people seek medical attention quickly in
the event of infection is difficult to mea-
sure – another analytical challenge. By
way of illustration, however, the Centers
for Disease Control (CDC) estimates
that, through Feb. 13, 2010, the 2009-
2010, H1N1 “swine” flu had infected 42
million to 86 million Americans, mostly
before an effective vaccine could be
manufactured and distributed in suffi-
cient quantity. This flu, fortunately, had
unusually low lethality, killing less than
.05 percent (8,520 to 17,620) of those
infected. More serious pandemic influ-
enzas within the last century have had
lethality of 15 to 20 percent or more;
according to the World Health Orga-
nization (WHO), the H5N1 “bird flu” a
couple of years ago killed at least 292 of
the small number (493) of people con-
firmed infected, mostly people handling
birds in Asia – that is, 60 percent lethal-
ity. From these data, it is evident that a
flu as infectious as the recent H1N1 but
with 20 percent lethality is a real possi-
bility, albeit improbable, and such a flu
could kill 8 to 17 million people in the
United States alone. No conceivable at-
tack in the United States with a single
thermonuclear weapon could kill nearly
as many people. Earlier detection and
response is the best defense against
such a catastrophe.
There will probably be secondary
and indirect economic benefits. Un-
der the current situation, many people are
forced to base job and business decisions
heavily on availability and quality of health
insurance for themselves and their family
members. Removing this constraint should
stimulate productivity. Again, measuring the
actual effect will be a challenge.
Improved access to mental health
care and in the effectiveness of such
care will produce substantial benefits.
The presidential commission that stud-
ied mental health care during the last
administration was emphatic in conclud-
ing that parity between mental health
claims and those for other ailments
would improve health and lower costs.
Among the probable benefits are low-
er costs in the criminal justice system:
most serious mental health diagnoses
are now made there, as the police and
courts have to deal with large numbers
of non-violent nuisance offenders who
should be treated, not incarcerated.
One thing widespread early screen-
ing will not do is reduce violent crime,
as the false positive rate of screening is
higher than the incidence of the relevant
disorders. On the other hand, more ef-
fective use of information to keep peo-
ple with a known propensity to violence
from acquiring weapons would make a
big difference. The 2008 Virginia Tech
shootings, for example, could have
been prevented if the intent of the exist-
ing state law had been carried out: the
shooter was omitted from the “no gun
purchases permitted” database after a
court finding that he was disposed to
violence because of an ambiguity in the
law specifying who had to be entered
into the list.
Many of the commission’s recommen-
dations were formally adopted but not
funded to the recommended levels. The
new legislation will direct more money to
mental health, and removing the risk of
losing insurance coverage may prompt
some more people to seek treatment.
Again, however, measuring actual effects
and assessing policy alternatives will re-
quire quite a bit of analysis.
The Democrats will suffer a back-
lash in the 2010 elections, but the lon-
ger-term political consequences are
yet to be determined. Many people in
: G : ER M B < L      u      F : R ( C N G >  + )*)
E > @ B L E : M B H G  : G : ER L B L
the country are palpably uneasy about
this legislation. Recent polls have dis-
approval leading approval of the new
health plan by margins ranging as high
as 20 percent while overall disapproval
of Congress is as high as 75 percent.
However, even if the Democrats lose
control of at least one house of Con-
gress in 2010, the news for President
Obama’s re-election campaign in 2012
is not necessarily bad. The same polls
cited above show his approval and dis-
approval ratings about equal.
President Clinton was re-elected com-
fortably in 1996 despite failing to enact
his health plan and losing both houses
of Congress in 1994. According to one
especially successful predictive model –
political historian Allan Lichtman’s “Thir-
teen Keys” – having made a major policy
change helps the incumbent, regardless
of how popular the change is; just having
had the clout and leadership skill to get
it through is a favorable indicator. (Pro-
fessor Lichtman’s model is admittedly
controversial. It does have, however, the
significant selling point of having predict-
ed correctly, well in advance, every presi-
dential election since 1980.)
If, by 2012, the health care legislation
doesn’t look so bad, there is no serious
challenge from within his own party or
from a third party, the economy is out of
, )       u      : G : ER M B < L F : @ : S B G > ' < H F
recession, and there have been no ma-
jor scandals, social upheavals, or major
military or foreign policy setbacks, the
Thirteen Keys model predicts President
Obama’s re-election, and one can eas-
ily imagine him asking on the campaign
trail, “So what was all that fuss about?”
and gaining accordingly. By then, some
astute congressmen and senators will
have gained credit for good constituent
service by helping people deal with the
new system and will therefore campaign
in support of the new law.
There will be much additional legis-
lation “adjusting” and “correcting” fea-
tures of this act, as its effects become
clearer. This is what Congress does to all
major legislation. This is most likely the
safest prediction we can make.
Health care systems analysts will
prosper. We all agree that the new law is
complicated. Therefore, there’s a market for
figuring out what it will do and how to deal
with it – enough of a market to keep many of
us employed for a long time. ❙
Douglas A. Samuelson (samuelsondoug@yahoo.com) is
a senior consulting analyst for IBM, focusing on national
security and defense issues, and president of InfoLogix, a
small consulting and R&D company in Annandale, Va. He
is a frequent contributor to OR/MS Today and Analytics.
He has a doctorate in operations research from George
Washington University.
REFERENCES
1. John M. Barry, “The Great Influenza: The Story of the Deadliest Pandemic in History,” Viking Penguin, 2004;
Penguin paperback, 2005.
2. Donald Berwick, E. Blanton Godfrey and Jane Roessner, “Curing Health Care,” Jossey-Bass, 1990.
3. Centers for Disease Control, U. S. Department of Health and Human Services, www.cdc.gov/flu/h1n1flu/
estimates_2009_h1n1.htm , retrieved April 12, 2010.
4. Committee on Quality of Health Care, Institute of Medicine, “To Err Is Human: Building a Safer Health Care System,”
Washington, D.C.: National Academy Press, 2000.
5. Committee on Quality of Health Care, Institute of Medicine, “Crossing the Quality Chasm: A New Health System for
the 21st Century,” Washington, D.C.: National Academy Press, 2001.
6. Tom Daschle, with Scott Greenberger and Jeanne Lambrew, “Critical: What We Can Do About the Health-Care Crisis,”
St. Martin’s Press, 2008.
7. Alain C. Enthoven and Laura A. Tollen, “Competition in Health Care: It Takes Systems to Pursue Quality and
Efficiency,” Health Affairs, Sept. 7, 2005
8. Alain C. Enthoven, “Choice in Health Care: Commentary,” Health Affairs, Vol. 25 no. 2, page(s) 566-67,
March/April 2006.
9. Laurie Garrett, “The Coming Plague: Newly Emerging Diseases in a World Out of Balance,” Farrar, Straus, and
Giroux, 1994; Penguin paperback, 1995
10. Kaiser Family Foundation, “Summary of New Health Reform Law,” http://healthreform.kff.org , retrieved
March 26, 2010.
11. Allan J. Lichtman, “The Keys to the White House: A Surefire Guide to Predicting the Next President,” 2008
Edition, Rowman and Littlefield Publishers, 2008.
12. New Freedom Commission on Mental Health, “Achieving the Promise: Transforming Mental Health Care in
America,” Final Report, DHHS Pub. No. SMA-03-3832, Rockville, Md., 2003. Also available online from
www.mentalhealth.samhsa.gov.
13. RealClearPolitics, various polling reports, especially http://www.realclearpolitics.com/epolls/other/obama_
and_democrats_health_care_plan-1130.html , retrieved April 13, 2010.
14. Douglas A. Samuelson, “A Dose of O.R.: Analysts Explore Ways to Improve Quality and Productivity in Health
Care,” OR/MS Today, December 1991.
15. Douglas A. Samuelson, “Diagnosing the Real Health Care Villain,” OR/MS Today, February 1995.
16. Douglas A. Samuelson, “A New Frontier? Health Services Research and Medical Informatics,” OR/MS Today,
February 2000.
17. Douglas A. Samuelson, “Can Early Screening for Mental Disorders Reduce Criminal Justice Costs?” George
Mason University Civil Rights Law Journal, 2001.
18. Douglas A. Samuelson, “Can O.R. Help Stop ‘The Invisible Plague’?” OR/MS Today, June 2004.
19. Douglas A. Samuelson, “Can We Detect ‘The Coming Plague’?: How Emerging Health Threats Are Sneaking
Up on Us,” OR/MS Today, June 2008.
20. Jon Stewart, “Commentary: Alain Enthoven: An Outspoken Champion for the Prepaid Group Practice,” The
Permanente Journal, Vol. 8, No. 3, Summer 2004.
21. E. Fuller Torrey, “The Invisible Plague: The Rise of Mental Illness from 1750 to the Present,” Free Press, 2002.
22. World Health Organization, www.who.int/csr/disease/avian_influenza/country/cases_table_2010_04_09/en/index.
html , retrieved April 13, 2010.
: G : ER M B < L      u      F : R ( C N G >  + )*)
A > : EM A  < : K >  : I I E B < :M B H G
AhpFZma^fZmb\ZeFh]^el<Zg
A^eiBfikho^<Zg\^kMk^Zmf^gmIeZgl
;k^Zdmakhn`aikhc^\m\hf[bg^l^qi^kmbl^bgmphl^iZkZm^Zk^ZlÇkZ]bZmbhgiarlb\lZg]himbfbsZmbhg'
;R@BGHC'EBF
CANCER IS THE second leading
' cause of death in the Unit-
ed States. According to the
American Cancer Society’s
Cancer Facts and Figures, 1,479,350
Americans are estimated to have been
diagnosed with cancer in 2009. The Na-
tional Cancer Institute (NCI) estimates
that more than 1,540 people are expect-
ed to die of cancer every day. We are los-
ing one innocent life every minute due to
cancer in this country alone.
,*       u      : G : ER M B < L F : @ : S B G > ' < H F
A NCI report recently announced
that rates of new diagnoses and rates
of death from all cancers combined
has significantly declined in recent
years. NCI Director John E. Nieder-
huber, M.D, says that the reasons
for this decline are thanks to “the ag-
gressive efforts to reduce risk in large
populations, to provide for early de-
tection, and to develop new therapies
that have been successfully applied in
this past decade.” He continues, “Yet
we cannot be content with this steady
reduction in incidence and mortality.
We must, in fact, accelerate our ef-
forts to get individualized diagnoses
and treatments to all Americans, and
our belief is that our research efforts
and our vision are moving us rapidly in
that direction.”
When a patient is diagnosed with
cancer, treatment methods are de-
termined based on the cancer type,
its location and its stage. The most
common cancer treatments are sur-
gery, radiation therapy and chemo-
therapy. Other types include immuno,
targeted and photodynamic therapy.
Physicians often use a combination
of these treatments to obtain the best
results. Ideally, cancer treatments
should be designed to kill all cancer-
ous cells without damaging a patient’s
healthy organs. However, it is practi-
cally impossible to completely spare
healthy organs during the treatments
: G : ER M B < L      u      F : R ( C N G >  + )*)
K : = B :M B H G  M A > K : IR  I E : G G B G @
because the organs are often located
in very close proximity to the tumor.
Designing a cancer treatment plan of-
ten involves a series of optimization
problems with different kinds of objec-
tives and constraints. Researchers in
the optimization community have made
significant contributions in improving
the quality of various treatment plans
for cancer patients such as surgery,
radiation therapy and chemotherapy.
As the oldest form of cancer treat-
ment, surgery is often used to remove
localized tumors. In cases where the
cancer has not metastasized, surgery
may be performed with curative intent.
Radiation therapy uses different types
of radiation particles such as photons,
electrons and protons to sterilize the
tumor or prevent cancerous cells from
proliferating. Two types of radiation
delivery methods are often used for
radiation therapy. The most common
radiation delivery machines belong to
the external radiation therapy or tele-
therapy. In teletherapy, the radiation
source is positioned at some distance
from the patient. Popular three-
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE!
Just visit: http://analytics.informs.org/
,+       u      : G : ER M B < L F : @ : S B G > ' < H F
dimensional conformal radiation ther-
apy methods belong to this category,
and they include Gamma Knife radio-
surgery, intensity modulated radiation
therapy (IMRT), traditional three-di-
mensional conformal radiation therapy
(3DCRT), tomotherapy, intensity mod-
ulated arc therapy (IMAT) and inten-
sity modulated proton therapy (IMPT).
The other radiation delivery method is
called brachytherapy. Brachy means
“short distance” in Greek. Therefore,
brachytherapy is a term of internal ra-
diation therapy where radioactive sub-
stances are placed inside or close to
the tumor.
Brachytherapy is often used for
treating patients with prostate, cervi-
cal or breast cancer. Effective plans of
radiation treatments require delivery of
high dose of radiation to the tumor while
minimizing the impact on normal cells
as much as possible.
Drug therapies such as chemotherapy
use medicines to control cancer systemi-
cally. Such systemic therapy may be ad-
ministered alone or in combination with
other treatments.
Finding a cure for cancer has been
an exciting goal in research commu-
nities. Researchers now have a much
better understanding of cancer. Many
innovative drugs are being developed
from all over the world to block cancer
cell growth by inhibiting specific mole-
cule pathways needed for cells to pro-
liferate. Radiation therapy techniques
have become very precise in achieving
the treatment goals. Engineers work
tirelessly to develop cutting-edge de-
vices to aid cancer treatments. Power-
ful and sophisticated cancer scanning
devices have helped improve the detec-
tion of cancer. Robotic arms are being
used in surgeries. Government entities
and private foundations have increased
awareness of cancer, while public health
initiatives have helped induce life style
changes that reduce the risk of cancer.
6)7)%6',-26%(-%8-32
8,)6%4=40%22-2+
I HAVE BEEN fortunate to work in this
exciting and rewarding area of research.
As a graduate student in optimization at
the University of Wisconsin-Madison,
my thesis advisor Michael Ferris intro-
duced me to a radiation therapy plan-
ning problem for brain tumor patients.
The problem was originally presented
to him by David Shepard, a graduate
student in Professor Rockwell Mackie’s
group in the Radiation Physics depart-
ment. Shepard’s initial modeling ef-
forts gave us the groundwork for our
research.
The project proved to be an excel-
lent example of how expertise in two
separate areas – radiation physics and
optimization – can lead to results that
could not be achieved with expertise
in one area alone. From the optimiza-
tion perspective, the primary goal of
a treatment plan is to deliver the pre-
scribed dose of radiation to the tumor
while minimizing the dose to non-tumor
cells. The total amount of radiation for
a complete radiation treatment varies
depending on the tumor type and the
radiation delivery device being used.
For a typical prostate cancer treat-
ment using IMRT, two gray (Gy) of ra-
diation are administered each day and
the treatment continues up to 40 days.
The reason for this fractionated radia-
tion delivery is based on cell growth
in the body. Cancer cells reproduce
much faster than normal cells. Radia-
tion therapy specifically targets them
by impairing or delaying this cell re-
production process. Normal cells can
quickly recover from radiation damage
overnight when they are exposed to a
low radiation dose. But, if they are ex-
posed to a high radiation dose, it may
cause long-term damage. Cancer cells,
however, are very sensitive to radiation
exposure and two Gy per day can ef-
fectively stop them from replicating.
: G : ER M B < L      u      F : R ( C N G >  + )*)
K : = B :M B H G  M A > K : IR  I E : G G B G @

portion belongs to the located in a patient’s head. It can de- different sizes in a traditional treatment
gross tumor volume (GTV). liver a single beam of high dose radia- planning. Therefore, treatment param-
But it is often difficult to tion by 201 Cobalt-60 unit sources. All eters include a set of radiation shot
draw a clear line between 201 beams simultaneously intersect at center locations, a discrete set of col-
the tumor and the normal the same location to form an approxi- limator sizes and radiation exposure
cells. Therefore, a clinical mately spherical region that is typical- time for each radiation shot.
target volume (CTV) is de- ly termed a shot of radiation. Multiple We formulated this problem as
fined, and it includes GTV shots are often used in a treatment us- a mixed integer nonlinear program-
and its surrounding suspi- ing a Gamma Knife due to the irreg- ming model and solved it as a variant
Figure 1. (Left) Multiple angles are used to achieve a high radiation dose on cious area. Finally, plan- ularity and size of tumor shapes and of the sphere-packing problem where
the tumor while minimizing radiation exposure to surrounding normal cells. ning target volume (PTV) the fact that the focusing helmets (or the PTV can be viewed as an empty
is constructed for treatment collimators) are only available in four container. Next, our job was to find
Figure 2. (Right) Beam’s-eye-view is constructed for each angle.
planning. PTV adds mar-
Radiation treatment planning begins gins to CTV to account for
with a set of images of the organs of movement of organs during treatment
interest. Such images can be obtained and the movement of the patient be-
by diagnostic imaging tools such as tween treatments.
computerized tomography (CT), mag- The images are further divided into
netic resonance imaging (MRI) or posi- a collection of three-dimensional cubes
tron emission tomography (PET) scans. (or voxels) for optimizing treatment pa-
Your one-stop shop to view top presentations from key INFORMS meetings
Those diagnostic images then help phy- rameters. This is where optimization
sicians determine the three-dimension- models can play a key role in treatment
al shape, location planning. The type of radiation delivery 2009 Annual Meeting
and size of the tu- device determines what parameters Plenary
mor and surround- need to be optimized. My colleagues Keynote
Wagner Prize Presentations
ing normal tissues. and I have developed optimization
Delineating treat- models for Gamma Knife radio-surgery, 2010 Practice Conference
ment volume on the IMRT and IMPT. Edelman Award Presentations (coming soon)
images is often a
difficult task to per- +%11% /2-*)6%(-3796+)6=
Your latest member benefit lets you learn from the best on your schedule.
form. A well-trained T HE G AMMA K NIFE IS a stereotac- http://livewebcast.net/INFORMS_Video_Learning_Center
Figure 3. Varian Medical radiation oncologist tic radiation delivery device for treat-
System’s multileaf collimator. can identify which ing tumors or vascular malformations

,,       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      F : R ( C N G >  + )*)

K : = B :M B H G  M A > K : IR  I E : G G B G @
a mix of spheres to fill the container
while minimizing the total space that
was not covered. Another goal: mini-
mize the number of spheres to do the
job. We have successfully developed a
computer program, tested it and used
it in radiation planning for patients at
the University of Maryland hospital in
Baltimore. The reliability of this treat-
ment-planning tool has been proven by
clinical trials at the hospital. In eight out
of 10 cases, the tool produced plans
that are better than the ones made by a
physician. Both the speed of the treat-
ment plan generation and the treat-
ment quality are the main contributions
of this decision-making tool.
-28)27-8=13(90%8)(
6%(-%8-328,)6%4=
IMRT IS A radiation delivery method
under the umbrella of three-dimen-
sional conformal radiation therapy.
3DCRT utilizes 3-D imaging data of
the tumor and surrounding anatomy to
achieve a high conformal dose of ra-
diation delivered directly to the tumor.
This treatment can be administered
A^eiIkhfhm^:gZermb\l
It’s fast and it’s easy! Visit:
http://analytics.informs.org/button.html
,-       u      : G : ER M B < L F : @ : S B G > ' < H F
for a variety of cancers. In 3DCRT,
multiple external radiation beams are
directed from different angles into the
tumor (see Figure 1). Each beam can
be shaped by a computerized multi-
leaf collimator mounted on a gantry to
conform the beam’s-eye-view of the
tumor (see Figure 2). Hence, a lethal
dose of radiation can be precisely de-
livered to a tumor. Photon beams are
used for 3DCRT.
As an advanced form of 3DCRT,
IMRT utilizes an advanced multileaf
collimator that can move its leaves
back and forth to block the portion of
each beam’s radiation dose (Figure
3). This enables a high degree of flex-
ibility in delivering radiation from each
gantry angle. Typically, between five
to nine angles around the 360-degree
circumference are selected for the
treatment. Radiation is delivered at a
fixed angle, and it continues until all
the selected angles are covered. This
treatment-planning problem is often
formulated as a mixed integer linear
programming (MIP) model, in which a
set of gantry angles and their radia-
tion intensities are optimized. In our
IMRT treatment-planning tool, the MIP
model is solved using the Branch-and-
Bound (B&B) technique, and the linear
programming relaxation models are
solved using an interior point method.
Since the MIP model contains more
than one million continuous variables
and several thousand binary variables
and more than one million constraints,
solving the problem optimally using
B&B can take several weeks using a
fast workstation PC. Therefore, heu-
ristic approaches such as local neigh-
borhood search, simulated annealing
and genetic algorithms are often de-
veloped to expedite the treatment plan-
ning process.
-28)27-8=13(90%8)(
4638328,)6%4=
PROTONS WERE first proposed for radiation
therapy by Robert Wilson at U.C.-Berkeley
in 1946. Since then, thanks to more than
five decades of continuous research and
development, we now have a potentially
very powerful radiation delivery machine
that uses proton beams. It has been report-
ed that proton therapy may be more pre-
cise than other forms of radiation treatment
in certain types of cancer. This is because
the dose distribution from a mono-energet-
ic beam of protons increases slowly, rising
to a sharp peak, called a Bragg Peak, at
the target. By superimposing beams of dif-
ferent energies, IMPT does only moderate
damage to tissues on its way to a tumor
and almost no damage thereafter.
The potential of IMPT is only now be-
ing fully understood, and optimization
is an important part of realizing IMPTs
full potential. I am presently working
in collaboration with the University of
Texas MD Anderson Cancer Center in
Houston on optimizing the delivery of
proton therapy. MD Anderson Cancer
Center started treating certain types of
cancer patients at their Proton Center
in 2006.
'32'097-327
C ANCER IS AN abhorrent disease, but
it’s a disease we continue to make
strides against every day. One way is
through cross-disciplinary efforts, such
as bringing the powerful tools of op-
timization to bear on problems in ra-
diation therapy. But the application of
optimization, and analytics in general,
isn’t limited to cancer treatment. It can
be applied to the design of new drugs,
imaging and genetic sequencing, not
to mention hospital planning, schedul-
ing and administration. Medical prac-
titioners from all walks can benefit by
adopting the many powerful tools ana-
lytics has to offer. ❙
Gino J. Lim (ginolim@uh.edu) is an assistant professor
of Industrial Engineering at the University of Houston.
The author acknowledges Andy Boyd for his valuable
comments and suggestions in preparing this article.
: G : ER M B < L      u      F : R ( C N G >  + )*)
< H K I H K :M >  I K H ? B E >
Ikbg\^mhg
<hglnemZgmlBg\'
?bkfZbflmh\k^Zm^\hfi^mbmbo^Z]oZgmZ`^makhn`abfikho^]]^\blbhg&fZdbg`'
COMPANY NAME: PRINCETON CONSULTANTS INC.
Headquar ter s: Princeton, N.J. (with additional office in New York)
Business: Management consulting firm specializing in optimization,
with focus on four key areas: business strategy, process engineering,
software development and project management
Web site: www.princeton.com
Princeton Consultants (www.princeton.
com) is a consulting firm with offices in Princ-
eton, N.J., and New York City. The firm spe-
cializes in optimization, combining information
technology and management consulting in ev-
ery assignment. Its goal: help clients to create
operational and strategic advantage by using
custom-produced software and business-pro-
cess changes to improve decision-making. Its
mission: educate senior executives about how
optimization enables companies to achieve
competitive advantage, drive up the value of
assets under management and provide a high-
priority strategic investment.
8,),-7836=
PRINCETON CONSULTANTS was founded in 1981
;RI>M>KF'MH;B: by Steve Sashihara and Jon Crumiller. CEO
Sashihara majored in philosophy at Prince-
ton University, and COO Crumiller majored in
,.       u      : G : ER M B < L F : @ : S B G > ' < H F
computer science at the University of Delaware.
Their early interest in solving business problems
by creating algorithms and using computers to
apply them led the two men to begin working
with early copies of optimization and statistical
software from nearby Bell Labs and research
professors from Princeton University
Today, Sashihara and Crumiller lead a firm
of 80 full-time consultants, all of whom have
worked their way up the through ranks by dem-
onstrating success in client engagements. None
are “lateral hires” from other consulting firms,
assuring that clients are served by proven indi-
viduals who possess deep, practical knowledge
of the field. The majority of the consultants have
graduate degrees in science, engineering or ap-
plied mathematics. Approximately one-third have
Ph.D.s. The firm remains privately held and is
run by senior staff – directors, senior specialists
and senior consultants.
8,)2))(
ACCORDING TO SASHIHARA, companies spend
massive amounts of money on technology to
supply them with data. However, even with all
this data, most are still using the “three Hs” to
make decisions:
• History – It’s right because it’s the way
we’ve always done it.
• Hunches – It “feels right” to me.
• Hierarchy – Because I say so, and I’m the boss.
There are better ways to make decisions.
For many years, IT consultants/specialists
have had rigorous decision-making technology,
which, like optimization, makes superior deci-
sions. What these “techies” have been unable
to do is analyze organizations to identify where
and how their software can add the most val-
ue. For that, you need management consul-
tants who are both business and technology
savvy – in a word, optimizers.
Using their management consulting savvy,
optimizers analyze a business to identify the
most important decisions the company makes,
how they are made and who makes them.
Then, they look for ways to improve those de-
cisions, using the powerful tools of IT; not tools
that just gather data for reports and dash-
boards, but software that can actually make
recommendations. The result: the ability to
make hundreds of decisions a little better than
they are currently being made – and to save a
lot of money as a result.
92-59)%4463%',
PRINCETON CONSULTANTS does not engage in
research for research’s sake. Its work is geared
toward real-life business applications. While
some optimization firms offer a menu of off-the-
shelf products and “tweak” them for various ap-
plications, Princeton Consultants creates new
software for each client to meet that company’s
special needs. For example, Quad/Graphics, a
Princeton Consultants client, is the largest pri-
vate printer in North America, with 11,000 em-
ployees at 11 large facilities. There are thousands
of small printers, and many of them use schedul-
ing software in their operations. But there is no
way to take an application created for a small
: G : ER M B < L      u      F : R ( C N G >  + )*)
< H K I H K :M >  I K H ? B E >

IKBG<>MHG<HGLNEM:GML printer and scale it up for a company the size of changing their plans at the last minute. Weather Allocating space: Space is another asset
<K>:M>=:GHIMBFBSBG@ Quad/Graphics. Optimizing for Quad/Graphics and air-traffic patterns are another source of con- that can be as scarce as equipment and no
LH?MP:K>IKH@K:FMA:M required building optimization software from the stant uncertainty and change. So choosing when less contentious. Take ad space, for example.
MHHDBGMH:<<HNGM ground up. and where to fly each jet and crew at the moment It is a precious and limited asset, both for busi-
:EEMA>KNE>L:G= of truth – when a customer calls and wants a jet nesses needing to reach a target audience and
8,)6)79087 now – is an energy-draining challenge. for those who own the media outlet, making ad
<HGLMK:BGMLNG=>K
PAB<AMA>:BKEBG> IN ITS 29-YEAR HISTORY, Princeton Consul- Princeton Consultants’s optimization team space a prime candidate for optimization. The
tants has helped clients reap large benefits by successfully created an optimizing software Wall Street Journal provides a case in point.
HI>K:M>L:G=IK>L<KB;>=
applying the principles of optimization to vari- program that took into account all the rules Given the complexity of its numerous edi-
MA>HIMBF:ELHENMBHG'
ous areas. Following are examples of its work and constraints under which the airline oper- tions and the number of rules governing where
with clients in three major ones: scheduling, ates and prescribed the optimal solution. The ads can be placed, laying out the paper’s ads
allocating space and pricing. bottom-line results? NetJets successfully uses is tricky business. For many years, it was done
Princeton Consultants’ models in production to manually each night by two employees. The
Scheduling: For many high-profile person- schedule all worldwide pilots and jets. ads that didn’t fit were shelved, and advertisers
alities who shun commercial flights, the airline
of choice is NetJets. NetJets flies approximate-
ly 400,000 flights annually on 775 aircraft to
approximately 175 countries around the globe.
Each of NetJets’ well-heeled passengers is a
part-time “owner,” for whom a plane is avail-
able whenever and wherever it is needed. The
resulting complexity makes efficient schedul-
ing a major challenge.
Knowing that improvements in fleet utilization
would give its owners tremendous advantages,
NetJets called on Princeton Consultants to create
a scheduling system that would maximize utiliza-
tion while minimizing customer dissatisfaction.
Many rules govern the decision as to who
should get which jet, when. Some are related
to the size and speed of the planes. Others are
mandated by the FAA. There are also union
regulations to consider and company policies.
Another complicating factor: The high net-worth
individuals who use NetJets often think nothing of

, /       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      F : R ( C N G >  + )*)

< H K I H K :M >  I K H ? B E >
:L:K>LNEMH?MA> didn’t learn that they had been left out until the
IKHC><M%MA>P:EELMK>>M next morning. The result: lost revenue and lost
CHNKG:E=B=GHMA:O> advertisers, who headed for more reliable ad-
MH;NRFHK>IK>LL>L% vertising venues.
ABK>FHK>I>HIE>HK Realizing that it was time to replace this anti-
=BL<HNGMBMLK:M>LMH quated system, the paper’s management asked
BG<K>:L>BMLK>O>GN>' Princeton Consultants to develop a program that
would optimize WSJ’s use of advertising space.
The firm’s team worked side-by-side with the
employees who had created the playbooks and
about 100 support staff at the paper to design
software that rationalized the ad-placement pro-
cess and drove up value.
As a result of the project, the Wall Street
Journal did not have to buy more presses, hire
more people or discount its rates to increase
its revenue. Just getting more ads into the pa-
per each day did it. Placing ads with the Wall
Street Journal also became much more cus-
tomer friendly. Now, an advertiser can call and
ask for an ad and find out almost immediately
if it will appear in the next day’s paper, elimi-
nating surprises. The bottom-line results? Dow
Jones & Company successfully uses Princeton
Consultants’ models in production to layout all
worldwide editions of the Wall Street Journal.
Pricing. Agile price setting can be a com-
petitive advantage is many industries, espe-
cially in large, B-to-B transactions. How would
your customers and prospects react if you
Ln[l\kb[^mh:gZermb\l
It’s fast, it’s easy and it’s FREE! Just visit: http://analytics.informs.org/
, 0       u      : G : ER M B < L F : @ : S B G > ' < H F
could give them a price quote in seconds in-
stead of days? What would such agility do for
your company’s reputation? How many more
prospects would call for a quote?
One of Princeton Consultants’ clients is
CSX Railroad. Each day, CSX moves more
than 20,000 railroad cars of essential items
throughout North America. Some shipments
are scheduled months in advance; other re-
quests come in at the spur of the moment.
The employees who schedule the freight ship-
ments need to know, at all times, exactly how
much space is available, on which trains and
how much it can be sold for. If a CSX train from
Jacksonville, Fla., to Chicago is going through
Nashville, Tenn., with several empty cars, and
a business in Nashville can fill a couple of
them, CSX can probably offer the shipper a
favorable price. The optimization software cre-
ated and installed by Princeton Consultants
can calculate the price point that will make it
a win-win for both CSX and the customer, and
it can do so instantly. The bottom-line results?
CSX successfully uses Princeton Consultants
models in production to provide instant Web
pricing and order booking, making what CSX
Intermodal President James Hertwig called “a
prime example of our determination to be the
most progressive intermodal service provider
in North America.”
8,)*9896)3*348-1->%8-32
SOME INDUSTRIES, such as transportation,
have been using optimization for decades with
excellent results. Despite their demonstrated
successes, other industries have been slow to
follow their lead. For example, the healthcare
industry has been slow to adopt optimization
techniques. Much of the discussion concern-
ing optimization in healthcare represents aca-
demic research, not field applications used on
a daily basis to reduce inefficiencies and im-
prove service. Paul O’Neill, former secretary of
the Treasury and veteran optimizer, summed
up the extent of the missed opportunities when
he said, “For me, there is no bigger smorgas-
bord of potential opportunity for the application
of your [optimizers’] talents and insights than
in the practice of American medical care.”
It is not just the medical profession that has
been slow to adopt optimization. Throughout the
business world, companies continue to manage
scarce and expensive assets in traditional ways,
shunning proven optimization techniques.
Sashihara maintains that this is about to
change. “I am very excited about the future of
optimization for two reasons,” he says. “The
plummeting cost of hardware and globaliza-
tion. You can’t just cut costs, centralize and
mass produce anymore. You need to be able
to ‘mass customize,’ which is what optimiza-
tion enables you to do.”
Sashihara, who also serves as co-chairman
of the Association of Management Consulting
Firms (AMCF), adds that optimization is about
to emerge as one of the hottest growth areas
in consulting. ❙
Peter M. Tobia, Ph.D. (pmtma@aol.com) is president of Market
Access, a marketing firm and literary agency based in Wyndmoor,
Pa. He has written for many major business publications.
: G : ER M B < L      u      F : R ( C N G >  + )*)
<HG?>K>G<>IK>OB>P

:nlmbgmhahlm The conference will offer the full range of

networking opportunities, with the INFORMS

+)*)BG?HKFL Membership Meeting on Saturday, the Welcome

:ggnZeF^^mbg`
GHO'0&*)>O>GM Austin, Texas, continually ranked as one of the
H??>KLIE>G:KR most livable cities in the United States, and where
L>LLBHGL%MNMHKB:EL% culture and music meet technology and entrepre-
I:G>E=BL<NLLBHGL% neurship, will play host to the 2010 INFORMS
G>MPHKDBG@ (Institute for Operations Research and Manage-
HIIHKMNGBMB>L ment Sciences) Annual Meeting, Nov. 7-10. Major
:G=FHK>' activities will take place in the newly modernized
Austin Convention Center (offering free WiFi) and
at the recently opened Hilton, both a short walk-
ing distance from Lady Bird Lake – an oasis of
beauty and recreation. Be advised that Austin is
a place that makes everybody feel at home. Aus-
tin has more than its share of offbeat characters,
Mexican free-tailed bats and a spirit of creativity
that has given rise to such national brands as Dell
;RCHG:MA:G;:K= Computers, Whole Foods and Willie Nelson.
The Texas State History
Museum (right) and the Texas
State Capitol Building (upper
right) are just two of the many
must-see attractions in Austin.
Reception on Sunday, the General Reception on
Tuesday and subdivision meetings on Monday
and Tuesday – all in the early evening. Due to
the convention center’s close proximity to the
bustling downtown entertainment district, the
General Reception will be held at the Conven-
tion Center, giving attendees a chance to spend
some time with their colleagues and then explore
the area. However, conference organizers have
lined up an array of novel and intriguing activities
designed to keep attendees fully engaged, so
you might never feel like leaving the premises.
The conference will run five days with work-
shops on Saturday followed by plenary and key-
note sessions, tutorials, panel discussions and
research seminars Sunday through Wednesday.
A partial list of principal speakers includes Chip
Groat, former director of the U.S. Geological Sur-
vey and current head of the Energy Institute in
Austin; Edward Kaplan of Yale University, the Mc-
Cord Morse Lecturer; John Birge of the University
of Chicago, the Omega Rho Distinguished Lectur-
er; Chung-Yee Lee from the Hong Kong Univer-
sity of Science and Technology; and Jerry Brown
from the Naval Postgraduate School. In addition,
the Wagner Prize winner, to be determined at the
conference, will give a talk, as will the Edelman
Award finalists from the 2010 INFORMS Practice
Conference. Finally, there will be 19 tutorials, 14 of
which will be accessible from the INFORMS Web
site, and a series of invited clusters in line with the
meeting’s theme of “Energizing the Future.”
For those interested in a peek at local indus-
try, conference organizers have arranged several
plant tours, which include visits to AMD’s research
center and Freescale’s semiconductor manufac-
turing facilities. Guests will have the opportunity
to take two tours. The first, the Tour of Austin Art,
stops at the Harry Ransom Humanities Research
Center, several museums and the Umlauf Sculp-
ture Garden, all with world-class collections. The
second, the Legends and Lore Tour, focuses on
those outsized Texans that you’ve heard so much
about such as the Alamo defenders, LBJ and a
few not-so-reputable oilmen and, among other
places, stops at Bob Bullock State History Muse-
um where guests can catch an IMAX film.
IBM and AIMMS (Paragon Decision Technol-
ogy) head the list of sponsors who have stepped
up to help make this a memorable conference.
About 4,000 people are expected to attend the
conference. For the most up-to-date information,
go to http://meetings2.informs.org/austin2010/;
for more information about Austin and area attrac-
tions, visit http://www.austin360.com/. ❙
Jonathan Bard (jbard@mail.utexas.edu) is the general chair of
the 2010 INFORMS Annual Meeting.

, 1       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      F : R ( C N G >  + )*)

 M A >  I N S S EH K

IZmb^gmGh'+*
Making a correct diagnosis on a patient is
a very challenging task for a doctor. A complex
disease can manifest through any number of
symptoms that often leave “trial and error” as
the only method for arriving at the correct di-
agnosis. To aid doctors in this daunting task,
mathematical modeling can be used to help
guide a diagnosis.
Table 1 shows 20 patients with varying symp-
toms and corresponding diagnoses. For exam-
ple, Patient No. 1 reported symptoms of sneezing
and a sore throat and was found to have a cold.
Patient No. 20 exhibited symptoms of fatigue
and sneezing that was found to be allergies. All
diagnoses were confirmed through lab tests.

Question: Patient No. 21 has not been di-

Table 1
agnosed yet is exhibiting symptoms of stuffy
nose, sneezing and sore throat. Using only the
data in Table 1, rank the three diagnoses (cold,
flu and allergies) in order of how likely Patient
No. 21 has each.

Send your answer to puzzlor@gmail.com

by June 15. The winner, chosen randomly from
correct answers, will receive an “O.R. The Sci-
ence of Better” T-shirt. Past questions can be
found at puzzlor.com. ❙
John Toczek (toczek@gmail.com) is the senior decision
support analyst for ARAMARK Corporation in the Global Risk
;RCHAGMH<S>D Management group. He earned his BSc. in chemical engineering
at Drexel University (1996) and his MSc. in operations research
from Virginia Commonwealth University (2005).

, 2       u      : G : ER M B < L F : @ : S B G > ' < H F : G : ER M B < L      u      F : R ( C N G >  + )*)