Do the management students know, what can be achieved
at present computationally?


Prof. Sinimole K.R.
Babasaheb Gawde Institute of Management Studies (BGIMS)

Student, studying whatever branch of optimization, needs a thorough education in the related theory, a good understanding of the major algorithm and some experience in the implementation of such methods. Most curricula take care of that appropriately. But what is about "real life problem solving"? Do we teach the "real thing" here? Do we provide them with the right picture of modeling and solving problems from industry?  What really should an O.R student learn? Do we provide them with the maximum size of linear programming problem that can be solved at present in acceptable running time? Or we provide them with special combinatorial optimization problems such as the min-cost flow, the max-cut or the traveling salesman problem? If a colleague, customer or student asks, what would you answer?

Dear students, do you know the exact scenario of today's corporate world? Today, competition, innovation, misconceptions, data modifications and planning or execution mistakes force companies to reconsider plans made in very short time frames. Instead of the manual trail and error methods, Why not try exact methods in Operations Research?. To the management students, if I ask, "how many of you know about the seven steps to a good O.R. analysis?"  Of course, you may say all of us. However, you can't say yes to the question "Modeling alternatives in deep. "  But now modeling skills based on a deeper understanding and computational experience may help you in corporate world. The O.R. body of knowledge is built upon a set of frameworks that enable O.R. people to first model and then resolve complex decision issues. Knowledge of these frameworks can be highly beneficial to real world decision-makers, even if they don't build the model or run the algorithm.  And this may be the (high tech) competitive advantage a company is looking for. We are moving to a day where managers want to do real-time adjustments to their plans and schedules, as well as interactively perform complex analysis in real-time on a desktop computer. In particular, these managers want to simulate different scenarios in order to understand how the solution to a problem is affected by varying the inputs. How can this be taught and learned?

My answer is: Students should be given problems covering a wide range of applications, including computing the tour of a welding robot, solving Steiner tree problems, optimizing pizza production, computing routing weights for IP networks, modeling telecom network design problems, bus routing, driver scheduling, line planning and repairman tour scheduling problems. Then students can experience that, choosing the right model is often more important (and more effective) than having the best solver implementation. Especially with real-world problems, having the ability to experiment swiftly with different formulations is essential. Since real life problems behave differently explaining this idea is a major issue. The O.R. decision-making framework includes many important concepts. For example, if you are going to make a choice, you need to think about what alternatives you have to select from, what criterion (or criteria) are affected by your choice and what value you attach to these, how risky are the outcomes and how much risk you are prepared to tolerate, and how to handle the ever present trade-off between return and risk. The O.R. decision-making framework also recognizes that not all decisions are simple "choose an action and live with it" situations. Many decision situations involve sequences of choices and uncertainties, where decision-making about future choices conditional on observed outcomes is critical to understanding current choices. In such situations, an understanding of contingency analysis is helpful. What do I do if this happens? When should I change my future conditional decision? It may be important to understanding the timing of sequential decisions and to delay making commitments until you have all the available data. In short we can say that the things that can be achieved by O.R. are

1. How to make good decision.
2. How to Tell a Good Decision from a Bad One
3. How to Cope with an Uncertain Future
4. How to Prosper in Risky Situations
5. Recognizing and Exploiting Simultaneous Decision Situations
6. Revenue Management and finally,
7. How to Link O.R. to Corporate Strategy

Many of our senior "C-level" executives are not quantitatively trained and see management as an art rather than a science. Convincing these executives of the strategic value of O.R. so that they will invest in O.R. work requires a framework that links O.R. to strategy. . As an example for the cost of their inefficiency, we can discuss "The Price of Managerial Neglect" here.

Does it cost a company when a manager neglects to improve a supply-chain or other manufacturing process over a three-year period? As far as we are concerned, such sins of omission are commonplace, but difficult to quantify in rupees. A new method for putting a price tag on the cost of "managerial neglect" has been developed by two industrial engineers, Rakesh Nagi, and Alfred Guiffrida, adjunct professors of industrial and systems engineering, in the University at Buffalo School of Engineering Applied Sciences.

In a recent issue of The Engineering Economist, The method, and how it would be applied to a two-stage supply chain, is described.

"Management theory says to improve a process you have to first improve its variability. Well, we've developed a way to put a price tag on the expected costs of failing to improve variability, for failing to improve a process," Guiffrida says.  Adds Nagi, "In this context, managerial neglect is something that a manager should be doing, but is not doing, and it's costing the company something. It's seldom that managerial neglect is quantified in financial terms."

The cost of managerial neglect is found by calculating the difference between learning-rate (which is the rate by which a process would improve naturally, without intervention, through repetition.) returns and the cost of not making improvements over time. In the example of a hypothetical two-stage supply chain, the engineers showed that managerial neglect over a three year period would double costs incurred from untimely delivery of goods, inventory holding, production stoppage or other inefficiency. Their managerial-neglect model could be used by managers to make the case for capital expenditures needed to improve a company's processes, Guiffrida and Nagi say. More over their model is easily applicable to the real world. Managers could use it to get the attention of upper management, to show them in quantifiable terms the costs you incur unnecessarily for failing to make improvements.

There are many well-documented examples of successful business applications of O.R. Examples include the O.R groups at FedEX and San Miguel corporation, ' Global Analysis' group at Procter and Gamble etc. and from India  One such company is Reliance group. An analyst said in "The Hindu Business Line Friday, Mar 18, 2005", "At RIL, there is always a linear programming that is under way, based on the world crude prices and the product demand pattern. The return on the rather high investment made in infrastructure was paying dividends, especially when the delta between the price of sour and sweet crude was significant - a direct corollary of rising crude prices".

If we go through different Operations Research Magazines or News letters from Web we can see news like

"Operational risk management market expected to reach $1.38 billion by 2010" "According to Chartis Research's latest report on operational risk management systems, the worldwide operational risk management (ORM) market (financial services sector only) will grow at a compound annual growth rate of 4.7 percent to $1.38 billion by 2010. "The growth in the operational risk management market is fuelled by replacement spending and new demand from the insurance and fund management sectors" says Helen Townsley, director of research at Chartis."

An article from "Science daily "- "Techniques For Making Barbie Dolls Can Improve Health Care- By the end of the decade, the health care industry will realize that operations research, IT, and other advanced techniques that manufacturers have been using for 15 years to reduce the cost of making items like toys and computer chips will also improve health care delivery," said operations researcher William P. Pierskalla, the John E. Anderson Professor and former dean of the Anderson School at UCLA. "These techniques offer major cost, quality, and access improvements to the healthcare system."

An article from "The Journal of the American Medical Association" -- "In most of the kidney transplants cases, blood type matching is a problem. A handful  of hospitals now supports paired kidney donations. In these cases, a donor and recipient who are incompatible with each other are matched with their "mirror image" another pair with the opposite incompatibility. The idea is that the first donor gives to the second recipient, while the second donor gives to the first recipient. Sommer Gentry, who holds a Ph.D. in electrical engineering and computer science from MIT, tackled the problem using graph theory a branch of applied mathematics to develop an algorithm for matching. The formula compares standard compatibility information from every possible donor-recipient combination available and determines which pairs should be matched to create the highest number of successful donations."

Smt. Sinimole K.R.

Karla Hoffman (2006) at the INFORMS Practice meeting described about - Real-Time Applications "Today, real-time scheduling and real-time adjustments are very important in a number of industries. As an example, the arrival of concrete trucks at a construction site is very time sensitive due to the fact that the concrete dries rather quickly, and if the concrete dries in the truck before it can be poured, it becomes useless. If multiple trucks are being scheduled, they need to arrive at the appropriate intervals so that they can be unloaded and so that room is made for the next truck. Furthermore, orders at additional sites need to be taken into account in real time". This application was described by Karla Hoffman (2006) at the INFORMS Practice meeting.

Robert Bixby in the article, "Solving Real-World Linear Programs: A Decade And More Of Progress," reported a speedup of a factor of 800 due to hardware improvements that can be extrapolated to a factor of 1,600 using today's hardware. He also reported a speedup of a factor of 2,400 over that same time period due to innovations in linear programming algorithm implementation.

"INFORMS News"  " Camm Takes Prize for the Teaching of OR/MS Practice."--Jeffrey D. Camm received the 2006 INFORMS Prize for the Teaching of OR/MS Practice for "his extraordinary dedication to his students' learning of management science practice and the resulting impact they have had in industry."

All these improvements have made it possible for operations research professionals to rethink what can be accomplished not only in terms of the size and complexity of problems that can be solved, but also in terms of how the results can be effectively managed by the business user.

Prof. Sinimole K.R.
Babasaheb Gawde Institute of Management Studies (BGIMS)

Source: E-mail October 18, 2007


Back to Articles 1-99 / Back to Articles 100-199 / Back to Articles 200-299 / Back to Articles 300-399
Back to Articles 400-499 / Back to Articles 500-599 / Back to Articles 600 Onward
Faculty Column Main Page