June 1997 • Volume 24 • Number 3

Surrounded by Success

Operations research and management science is everywhere; all you have to do is look around. A day in the life of an OR/MS practitioner in a non-OR/MS job

By Douglas Gray

I often hear and read about people bemoaning the death of the operations research and management science profession. To these people, OR/MS seems to suffer the indifference of a world too busy to care about the value that mathematical models can bring to better understand what makes businesses and organizations operate more effectively and more efficiently. Many OR/MS professionals seem to be at a loss as to how to create and declare success.

I think that depends on how you define success. Is success deriving some new elegant algorithm? Is success having the CEO of your company stand up and say your OR/MS work is truly amazing? Is success winning the Edelman Award, the INFORMS Prize or the Lanchester Prize? Although desirable ends, these are pretty narrow measures of success by which most of us are doomed to fail. However, if you define success as using OR/MS effectively in your daily work and contributing to some value-added positive end — regardless of the endeavor — I believe that we can have many more OR/MS successes, albeit on a smaller-scale, in addition to the large-scale ones. There are more of those seemingly "little successes" that ultimately have a larger positive impact than most people realize.

Successful OR/MS applications are all around us, if you know where to look. Plenty of good OR/MS practice is hidden by job titles. Lots of people do OR/MS, but don't carry the title OR/MS analyst. I know professional people, e.g., market researchers, investment analysts and project managers, who effectively apply OR/MS methods in their daily work to achieve significant, positive results despite the fact that their primary educational training wasn't in OR/MS. Product managers, who must balance the cost to create a new product with market share objectives to determine a competitive price, apply fundamental OR/MS principles. I have one friend who is a manager of information technology for a local police department who writes computer programs that do everything from generate and analyze crime statistics to optimize patrol routes and officer allocations. OR/MS is where you choose to find it.

Alan Greenspan, chairman of the Federal Reserve Board, recently spoke to Congress at length on C-SPAN about the econometric models he uses to reveal the mysteries of the economy, specifically the Consumer Price Index (CPI), which attempts to measure inflation by analyzing changes in the prices of basic consumer goods. However, he also talked about his "personal rules-of-thumb" that he uses to validate the models' predictions, and gauge the "true health" of the economy, e.g., calling his friend at the Bureau of Labor Statistics to see how many unemployment claims were filed this week in major U.S. cities. He was candid about the limitations of the sophisticated models to predict economic trends, but at the same time defended them, saying, "It is better to be approximately right than certainly wrong."

OR/MS, in general, shares a similar nature with Greenspan's econometric models. Despite the limitations of those models, I think the world is better off with OR/MS practitioners than without us. The world is so full of practical problems, of all sorts, and with businesses having downsized, in some cases to the point of adversely impacting customer service, the urge to "do more with less" beckons the OR/MS solutions that can address such problems.

What is 'Good OR/MS'?
Sometimes, however, I believe we too stringently limit our definition of what passes as "good OR/MS," and prohibit ourselves from taking credit for successes. Loren Platzman, my Probability Models/Queueing Models professor at Georgia Tech, told us graduate students that an OR/MS practitioner's most powerful tools were common sense, good judgment, intuition about a problem's structure, some data points and a spreadsheet. He encouraged us to collect some data, enter it into a spreadsheet, draw some graphs, and "get a feel" for what is happening with the system at hand, to validate your intuition before you build any models. Then, once you have built and validated your models, go and tell someone, and do something about it.

I have followed that rather straightforward advice throughout my career. By the strict definition of what is "good OR/MS," some of my professional work may look, to some people, like "glorified spreadsheets." However, customers consider them "successes" when the salient model principles, features and functions are implemented, in the form of decision support systems, and ultimately reduce costs or increase revenues by several million dollars.

For more than 10 years, I have had the opportunity to apply many different OR/MS methodologies to a lot of different real-world transportation industry problems, ranging from discrete-event simulation analysis of airports to determine capacity of airspace and airfield structures to mixed integer programming heuristics and scheduling algorithms for scheduling aircraft maintenance activities and resources. I also had the luxury of working in a successful OR/MS organization, so I never really knew what it was like to be in position where doing OR/MS wasn't my primary job function.

OR/MS Mindset
Recently, I transitioned into a new assignment in the travel distribution technology practice area. I am currently responsible for a group of 60 software engineers focused on building Internet/WWW-based consumer travel reservations systems, and travel agency Internet-enabling software applications. Needless to say, I thought my days as an OR/MS professional were over. However, as I sat in meetings with my team and my new clients wearing my "new IT director hat," something interesting happened. Some very real and perplexing business problems began to arise for which no one in the meetings seemed to possess a solution. However, given my OR/MS training and experience, for me each of the problems fit into a "class of problems" for which model forms were readily available. After I had convinced myself that I was out of the "OR/MS business," I was still able to apply my OR/MS "mindset" and accompanying "bag of tricks," and add value by recommending viable solution approaches to some of the pragmatic business problems at hand. Here are just a few examples:

My client wanted to better understand the number of people who were being blocked from entering their consumer travel reservations web site, given limited capacity and waiting space to enter the travel transaction processing system. We could simply count the number of "system busy pages" served, but we wanted to know more about the distribution of how many people were arriving at the system. Of those customers who got into the system, how long were they staying? What was the combined effect of arrivals, system capacity and how long people stayed in the system on the number of people blocked from entering? A seemingly complex problem, but one that fits nicely with a classic queueing theory application — the Erlang blocking formula, with Poisson arrivals and exponential service times.
Electronic commerce and transaction processing systems on the Internet raise many of the same issues faced by telephone engineers (like Erlang) around the turn of the century regarding arrival rates and service times, and the telephone system capacity necessary to balance cost and customer service objectives. Having collected and validated data on arrival and service time distributions, using an off-the-shelf queueing analysis package we were able to quickly estimate the number of customers blocked from the site, or in queue at any given time during the day. This enabled the client to make an informed decision about how many people they would allow into the system to make reservations, and hence generate revenue, versus the cost necessary to provide that system capacity.
A client wanted to offer a service by which a customer would receive an e-mail message notification in response to an airline fare ccircle 3 on reader service card hange reduction of $50 or more on any of their five city-pair airline flight schedule selections, e.g., DFW-LAX. Offering such a service raises several questions regarding IT infrastructure requirements. The number of e-mail messages you will have to send out over a certain time period (e.g., overnight) will drive the need for an e-mail server of a certain capacity. The number of messages sent is based on the number of people who sign up for the service, and how often the fares are likely to change. It turns out that given an estimate of the probability of an airfare change, and some assumptions on the number of people who are likely to sign up for the service, a spreadsheet-based probability model, based on the binomial distribution, can be quickly created to estimate the number of e-mail messages that will need to be sent out (given that a customer will receive a message if one or more their five airfare city-pair selections change).
This simple model, and subsequent analysis, helped the client determine the cost and benefit parameters associated with providing the service, and enabled them to ensure that the appropriate e-mail server capacity was available to send messages to customers in the timeframe promised.
One of the "killer apps" in the travel business is to find the "best fare," where "best" depends on customer preferences with respect to price, airline ticket restrictions, time windows and airline choice. The best fare is also limited, of course, to the availability of airline schedule and fare combinations. Given the extremely large number of possible combinations of schedules, airlines and fares, coupled with the fact that the answer must be delivered in real-time (i.e., a travel agent on the phone with a customer, or a consumer on a travel reservations web-site), this turns out to be a very challenging problem for which explicit enumeration is obviously out of the question.
Using branch-and-bound-based implicit enumeration heuristics, combined with a travel agent's knowledge of customer behavior and a lot of heavy-duty mainframe computing power, OR/MS plays a significant role in solving this practical travel industry problem, which is reminiscent of the classic Traveling Salesman Problem, millions of times per day every day.

Why is OR/MS Struggling?
I discovered in my own career that OR/MS is where you find it. And, more importantly, you don't necessarily have to have an OR/MS job description, or work in an OR/MS department or a consulting company specializing in OR/MS applications to do what I consider "good, valuable OR/MS." The bottom line is that the mathematics to do OR/MS has been there for decades, the computing power is now orders of magnitude better, faster and cheaper than ever before, and certainly the problems and challenges of business abound. So what is missing? Why is OR/MS struggling in a world full of problems that beg to be solved?

My experience tells me that there is no "silver bullet" solution; there is no magic formula cure-all panacea for what ails OR/MS, because there is no substitute for getting your hands dirty with a lot of hard work, combined with a diligent customer focus, excellent communication skills, and most importantly, a measurable, demonstrable value-add. "OR/MS done well" is necessary but not sufficient for OR/MS to be successful. I don't believe that there is anything inherently wrong with OR/MS, other than the way it is applied, or rather, misapplied in many cases. I believe that we, as a community of OR/MS professionals, are struggling with the same challenges as corporations, government, the military, individuals, ad infinitum; that is, in a world of diminishing available resources, what is OR/MS's perceived and demonstrated value-add and return-on-investment? I firmly believe that the customers of OR/MS, whoever and wherever they are, will decide our success or failure for us, based on our performance against that simple metric.

Today, I think of OR/MS less as an academic discipline or a profession in and of itself, and more as a framework for analyzing and solving complex, real-world problems, and making decisions in a structured mathematical way (using a heavy dose of computer and information technology). Like everyone else in the world today, I don't get paid to fulfill a job description, e.g., OR/MS practitioner; I get paid to achieve a bottom line result profitably by solving problems practically as they arise.

Bottom Line Benefits
The way to create an effective role for OR/MS in the world depends on champions who are willing to roll up their sleeves, take the time to collect some data, model the problem and explain to management, in English (without Greek letters), the bottom line benefits associated with their solution. Reducing costs, increasing revenues, enhancing customer service, improving efficiency and, most importantly, sticking around long enough to see their ideas get implemented and make a difference is what it's all about. Edison said it first, and any inventor would agree, that "Genius is 1 percent inspiration and 99 percent perspiration." The OR/MS profession is full of inspired geniuses, but we could use a little more perspiration.

The best examples of individuals and companies that apply OR/MS effectively, on a broad scale, are INFORMS Prize winners. If you look at the companies that have won the INFORMS Prize you will find a lot of people who understand as much about business as they do OR/MS. A little OR/MS goes a long way when applied with common sense, sound business judgment, the 80-20 Pareto principle of 80 percent of the benefits for 20 percent of the cost. At Sabre Decision Technologies, in particular, we continue to work according to that fundamental principle every day. All of our senior executives have education in OR/MS, engineering or computer science; more importantly, they all understand the impact of their respective technologies on their respective customers' businesses.

These groups did not spring up over night. There were not suddenly 50, 100, 500 or 1,000 people doing OR/MS projects. They started small as groups of one, three, five or seven analysts and a manager, and grew project by project, deliverable by deliverable, customer by customer, because they added a value so substantial, and so unique that it could not be replaced by the senior management of their respective corporations. These groups have instantiated themselves as a critically necessary and value-added part of their organization's business process.

The best any of us, OR/MS professionals included, can aim for in today's world is to have customers who, at the end of the day, value our contribution enough to hire us for another assignment, and then another after that one. The future of the OR/MS profession will be bright if that is the only criteria for success that ultimately matters.

Douglas A. Gray is a Senior Director of Applications Development and Consulting at SABRE Decision Technologies (SDT), the software and consulting arm of The SABRE Group. Gray has 10 years experience as an OR practitioner, consultant, manager and director at SDT. He has successfully led several decision support application development and consulting initiatives for transportation industry clients worldwide. Gray currently leads a 60+ member group which is responsible for consumer-direct and travel agency Internet-enabling travel distribution software applications development. He holds a BS degree in Mathematical Sciences from Loyola College, and a master's degree in operations research from the Georgia Institute of Technology's School of Industrial and Systems Engineering. He is a member of the OR/MS Today Committee.

For more information, put the number 3 in the appropriate space on the
Reader Service Form

E-mail to the Editorial Department of OR/MS Today: orms@lionhrtpub.com

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com