August 1996 € Volume 23 € Number 4

Point-and-Click OR

Procter & Gamble use GIS to optimize its supply chain

By David Greenfield

It's no news to anyone in the field that operations research and the management sciences are being taken more and more seriously in the corporate arena. But what is news is the inventive ways in which these sciences are being applied. Procter & Gamble, Cincinnati, Ohio, is one example of a company that has embraced OR for well over 20 years and is now stretching the bounds of the field's applicability.

In 1993, Procter & Gamble decided to reevaluate its supply chain structure. Numerous changes were occurring in business operations at the time and the company realized it would have to be able not only to work within these changes, but anticipate them as well. Starting in the 1980s, these changes began to involve the deregulation of the trucking industry and the company's introduction of more compact or concentrated products, such as diapers and detergents.

Franz Dill, leader of P&G's Analytics Center of Expertise (a part of the company's Management Systems), says these changes allowed more product to be packed in a truck, permitting the company to take advantage of better truckload rates.

At the same time, other changes were taking place at Procter & Gamble involving OR. "We were taking better advantage of the fact that we could run our individual production lines more efficiently using reliability theory and simulation methodologies," Dill says. "If your lines run more efficiently, you can get more capacity out of any individual manufacturing plant, therefore you need fewer plants."

Putting the project together
When reevaluating a supply chain as large and diverse as Procter & Gamble's, the process has to take into account how many people are involved and how greatly they will be affected by any changes made to the system. Dill said that in approaching the problem, the OR team was going to have to deal with a large group of people, in excess of 500.

Thanks to the company's "rich history" of OR use, however, Dill says that Procter & Gamble was not unfamiliar with logistics modeling of this type. But never before had the modeling been put directly into the hands of managers -- the very approach that was proposed for the most efficient use the system to be developed.

"With some 500 people involved, a lot of egos and different thoughts about how to do things were always at the forefront," Dill says. "What you wanted was something that could show people the direction to go in and very rapidly allow them to evaluate different options. So we thought: Can we provide a spreadsheet kind of approach where we can all sit in a room -- all the transportation experts, all the management experts, and upper management in some cases -- and have people throw out potential ideas such as, 'What if we consolidated here?' 'What if we changed the way we do transportation in the northeast?' And we could very rapidly respond with: 'That will cost you or save you X millions of dollars based upon what you've said.'

"So the whole notion was not just OR, but to have something that could be looked at so that the user could visualize the results to try and understand what they meant to the system."

Knowing what was desired of the system to be produced, Dill assembled a team of modelers from within the company, including Glenn Wegryn of MS Global Logistics and Tom Chorman of Finance, who were actively involved in classical logistics modeling. The team then looked to outside sources, such as Wharton, MIT and the Center for Productivity Improvement (CPI) at the University of Cincinnati.

Dill decided to align with CPI because he had worked and taught with Dennis Sweeney (CPI director), and he knew they had experience in this area, especially in the logistics arena. Proximity helped also. "It worked out to be a good relationship because we could go over and talk to them in person," Dill says.

CPI wound up working on the regional distribution center problem -- a central portion of the supply chain reorganization project. Having CPI onboard allowed Procter & Gamble to segment the problem into two pieces: the regional distribution center location problem and the visual laptop optimization problem.

"They (CPI) solved the distribution center part of the problem and prestored a number of solutions into our system, and then when we actually worked with the managers using CPI's solutions to drive the problem, it segmented nicely," Dill says.

Speaking of CPI's contribution to the Procter & Gamble project, Sweeney says, "The people in our center specialize in soft technology. We have OR people, statistical people, and a fair amount of experience in modeling. We brought to the table an ability to deal with this kind of problem.

"To this particular project we presented ideas on how to decompose the issues across product lines and across echelons of the supply chain. We provided some good ideas on how the problem could be attacked and supplied models that would support the organization structure that they had, and at the same time have the models tied together in such a fashion that suboptimalities were not developed."

Visualizing the problem
Now that the team had been provided with models and a distribution center location from which to work, the basic issue remaining was how to visualize the kinds of supply chain problems the managers for whom this system was being designed would have to deal with.

"I'd seen in the past where we had done logistics models and then translated them onto maps either by drawing or floating the data into some mapping package and showing the results," Dill says. "I was always struck by the notion that people got very good insights out of the drawing. But it was always very clunky. If a change had to be made, you had to redraw it and go through another cycle of going back and reworking the problem.

"When we started this project, GIS (geographical information system) packages were coming out for the desktop and I thought: Wouldn't this be nice if we could build a GIS front-end where all the data manipulation would occur from the GIS perspective so you could click on a line that represented shipping from A to B and change the shipping costs right there? And then I thought, what if we could change all kinds of facets about the supply chain visually, as opposed to changing some arcane number in the middle of a spreadsheet that only a few people in the room understood? I knew everyone involved would understand the map because it was a clear metaphor. So the going-in philosophy was to have the map be the entire interface -- no tables of numbers, no big databases that some assistant had to manipulate in the background. The notion was you could look at the map, click on a line, click on a city, or click on a plant and change the number right then and there, and then very quickly get the results from the optimization. Everyone would be able understand that instantaneously."

When Dill began looking at GIS packages, he knew he wanted one that he could hook into an optimization model. He chose MapInfo from MapInfo Corp., Troy, N.Y., because it had a programming language that would allow him to call on external modules. He also wanted the optimization to work as fast as possible.

"We didn't want to go away for an hour while it solved the problem on the laptop," Dill says. "So we simplified the transshipment model to a large degree. We cut out all the detail that we sometimes include on some of the more complex models simply so that we could get a one-minute turn around."

According to Dill, integrating the two packages -- optimization and GIS -- did prove to be somewhat problematic at times because some of the "hooks" in the GIS were not built as expected, or problems arose when calling one program from another. But the Procter & Gamble team had an advantage in that they were using a C code module to do the actual optimization, giving the team a lot of flexibility on that end of the problem. Having the module configured in this way allowed the team to modify it in any way they wanted, Dill says.

Harvest the solutions
One of the key benefits reaped from the display of information on the screen was that users could very quickly determine where there were errors in the data. For example, there were situations where lines on the map crossed each other and it looked as if something very unusual was taking place, like shipping from the East Coast to the West Coast. When these sorts of errors were noticed, the team could drill down into the data and see if there was some rational reason for such shipping, e.g., some sort of special rate to go from Baltimore to the West Coast. Sometimes, however, strange pictures on the map did translate into actual problems, such as data entry error. Then at other times, strange graphics were actually representative of a correct number and this information would come back to the logistics team, leading them to ask: "Why are we doing this, why is this happening?"

The mapping of the supply chain proved very beneficial in highlighting the pluses and minuses of Procter & Gamble's supply chain, making it easier to perform a true optimization of the system with little or no information remaining hidden from view.

Aside from a more efficient and productive alignment of the company's supply chain, the success of the project led to the establishment of the Analytics Center of Expertise (ACOE) at Procter & Gamble, headed by Dill.

"We (ACOE) are a worldwide virtual organization dealing with analytics problems, which is basically OR," Dill says. "Analytics at Procter & Gamble has been declared a competency within the corporation, meaning it's an area that we feel we can internally excel in and do better and more efficiently than anyone else. We are a group of 20 people worldwide that apply these kinds of technologies, and logistics is just one part of that."

Dill says that the ACOE is now involved in dozens of projects, most of which are involved with project supply or logistics engineering. Some, however, are outside that area and involve marketing or research and development.

Although a workable GIS supply chain optimization model is in place, the ACOE is working on improving the system.

"I think we're doing much better on that," Dill says, "but we're still a long way from where we have systems that are completely easy to use and can be put directly on a manager's desk and let them work with it."

Even though this system is still being perfected, according to Tom Chorman, finance manager for new corporate ventures at Procter & Gamble, its use has resulted in quantifiable cost savings in excess of $200 million a year.

David Greenfield is the managing editor of OR/MS Today.
For more information on this article, input the number 1 in the appropriate place 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