Where is the “P” in OR/MS?

Operations research and management science modules are rarely the most exciting or popular modules in the business school curriculum. This seems to be the case in the United States and elsewhere around the world. Why?

We believe that the most significant problem lies in the mismatch in interest between the students and the professors. Most audiences that O.R. professionals talk to are (aspiring) managers. Unfortunately, managers, even those overseeing operations, spend most of their time managing people and not operations. They are more interested in “People,” whereas the O.R. professionals are more interested in the “Solution.”

To get their attention, we have to bring back the “P” in our O.R. models.

While it is well-known to practitioners, scant attention to the role of “People” has been paid in most OR/MS courses, and many of our graduate students often overlook this crucial consideration in their modeling work. By ignoring or dismissing the impact of their decision variables on people, these models can become irrelevant. In fact, people (deployed properly), not sophisticated models, are what really solve the operational problems.

Legend has it that Sun Bin, the great-grandson of Sun-Tze, was once challenged to revive the fortune of the losing team of a tug-of-war contest. The team had lost several rounds against the stronger opposition and had low morale. Sun Bin had to find a way to turn around the performance of the losing team quickly. He chose to whisper a few words into the ears of those in the losing team and rest his hand on their shoulders for a short while. Amazingly, the team won in the next round. What did he do? His reply: “The people idolized me because of my mastery of the art of war. I told them that I have transferred my energy to them by touching their shoulder. They believe they can win now.” By turning despair into a firm belief that they would prevail, Sun Bin was able to reverse the fortune of the losing team immediately.

The Dabbawallah System of Mumbai is another example of how sheer human determination can overcome harsh environmental constraints, in this case, to produce a highly reliable delivery system for the masses. The people behind this famous meal delivery system in Mumbai are predominantly illiterate, and they succeed without any advanced IT platform or modern communication tools, going about doing their job mainly on foot and bicycle. How do they maneuver through the crowded streets of Mumbai to deliver lunch boxes to the right address at the right time consistently?

The secrets to their success: hard work of the people, a tightly knit relay system to move the boxes and clever exploitation of the railway system in Mumbai to create a “rhythm” for the delivery operation [1].

The first author witnessed on the Juhu beach of Mumbai another marvelous example of people prevailing in a harsh environment. The beach was packed with people, and there was a carnival atmosphere in the air. Several enterprising businessmen had set up rides such as Ferris wheels and merry-go-rounds along the beach. However, getting power generators onto the beach to propel these rides appeared to be difficult. They chose to solve this problem using what is readily available on the beach and in India – people.

A few young men were employed to propel the Ferris wheel using their body weight. They took turns jumping off a high platform while hanging on to a sidebar attached to the ride. In the process, they had fun, made some money and provided the needed power for younger kids to have fun on the Ferris wheel, too.

We need to emphasize to our students the amazing ability of people to come up with ingenious solutions to indigenous operational problems. We also need to remind them that our “optimal” solution has an impact on the life of the people in the system.

Several years back, we built an elaborate mathematical programming model and produced a nice schedule for a mass transit company in Singapore, as part of its crew scheduling operation. While the managers were visibly pleased (Not surprisingly, they were more impressed by the GUI and its report generation capability. It made the life of the scheduler much easier), we learned later that they were not interested in the potential manpower cost savings produced by the optimal schedule. Why?

It turned out that the supposed savings in manpower cost in the optimal schedule came from extensive usage of “split-duties.” In this duty structure, the crew would work for a few hours in the morning, take a break to go home and return later in the afternoon to complete the second half of their duty. (There are other rules to ensure that such duties will not be too harsh on the workers.) This is an approach the company had adopted, accepted by the workers, to cope with peak travel demands in the early morning and evening rush hours. While our algorithm exploited this effect relentlessly to achieve its manpower savings, the managers recognized that the savings come from the sacrifices of the workers who were put on these duties. Should the management then exploit this structure to cut costs further, while more of their workers go on split duty? Of course not!

Can we do beautiful mathematics if we consider the “P” in our model? Well, it is difficult, but we can at least enrich our OR/MS models by asking other questions. Can we reap the benefits of a fully integrated (sophisticated) model by making only minimal changes to our existing (simple) system? Can we create the most conducive operational environment for our people, without incurring much additional cost? The first question focuses on profits and tries to minimize the impact on people affected by the changes, whereas the second question focuses on improving the working condition of the people, without incurring much additional cost to the operation.

Mathematically precise statements, elegant arguments, and relevant managerial insights are still possible from these perspectives. In the crew scheduling work, we can seek to minimize the number of split duties (thus making life better for the workers), while sticking to the same level of the workforce. This is in itself a nice optimization model. Recent research on the topic of process flexibility [2] has also yielded surprising insights: Instead of n plants each producing a dedicated product (with random demand), we can simply make each plant a little more flexible (capable of producing only a few products), and with a carefully selected “sparse” plant capability structure, the (simple) system can reap almost the same benefit as a fully flexible system where each plant can produce all the n products [3].

Interestingly, the same insight is also valid in many other operational contexts, from server allocation in queueing systems to skill training in workforce agility. Minimal changes properly incorporated into an existing system can already reap the majority of the benefits obtained from the best possible (but complicated) solution.

The roles of “P” in other contexts are also well-documented: (market-driven) prediction markets are typically fairly accurate and outperform most sophisticated benchmarks [4]; the use of human subjects in laboratory experiments has also changed the field of economics [5].

Maybe it’s time we bring “P” back into our OR/MS modules too.

REFERENCES AND FURTHER READING

1. See POM Chronicle, 2005, Vol. 11, No. 3.

2. See, for instance, Jordan and Graves, 1995, Management Science, and Iravani, Van Oyen and Sim, 2005, Management Science, and the references therein.

3. More formally, a sparse partial flexible system can be within (1-e) optimality of the capability of a fully flexible system, for large n. See Chou, Teo and Zheng, 2005, for a formal proof using a counting argument.

4. See the paper on “Prediction Markets” by Justin Wolfers and Eric Zitzewitz.

5. See Colin Camerer, “Behavioral Game Theory: Experiments on Strategic Interaction,” Princeton University Press, 2002.