FLIGHT SIMULATORS FOR MANAGEMENT
Computer models may give execs previews of how decisions pan out
Thor Sigvaldason is late. As part of a novel consulting cluster at PricewaterhouseCoopers, he is supposed to be at an important client meeting at 10 a.m. But he has overslept for the session with a top executive of Macy's East.
The other team members run through their paces, and when Sigvaldason finally arrives, 20 minutes late, his brown hair is still wet from a shower. He eases quietly into the room, watched only by his boss, K. Winslow Farrell Jr., 45, whose face doesn't hide his displeasure.SCI-FI? Soon, though, Thor brings up on a computer screen what has taken the team 1,800 hours to construct. It shows awkward blocklike figures roaming about a crude layout of a department store. The idea is to simulate how real shoppers actually operate in a store. But like bumper cars at a carnival, the little figures bounce off the blue walls. Explains Farrell: "We haven't built certain rules into their brains yet, and they have primitive vision."
The client, Macy's group vice-president William M. Connell, is amused and oddly pleased, if only because he can see that the team is making some progress toward an incredibly ambitious goal: to harness the emerging science of complexity, the notion that complicated behavior emerges from the interaction of many components, and create a powerful new tool for top executives. Farrell's team is creating artificial worlds of evolving, reactive software creatures. If all goes well, the actions of those "adaptive agents" will so closely mimic human behavior that managers for the first time will be able to use them to test the impact of their decisions before implementing them in the real world.
The stuff of science fiction? Well, it may seem a bit far out and abstract. Some of Farrell's own colleagues within PricewaterhouseCoopers think his Emergent Solutions Group is overreaching, trying to invent something out of a pop-science movement. But Farrell sees it as the ultimate "flight simulator" for management--and his firm appears to be the furthest along of several consultants vying to take the concept out of the laboratory and turn it into a commercial business. A handful of corporations are each betting hundreds of thousands of dollars that Farrell can succeed. If he does, PricewaterhouseCoopers expects the tool could be used in a broad range of consulting assignments, generating millions in revenues over the next three to five years.
Farrell is trying to create a virtual world where executives could safely test hunches, run scenarios, and preview the impact of big and small decisions--all without major investments, public embarrassments, and competitive backfires. Macy's, for example, is hoping the computer model will help it figure out such things as where to place cash registers and service desks to improve sales."A SYNTHETIC PUBLIC." The retailer expects a working product by the end of the year. "If we can successfully do it, the implications are enormous, in terms of time compression, cost, and reputation," says Connell. "You can run a day in minutes and a month in hours. You could see whether investments are likely to pay off or not. And if you make a wrong decision in the model, you're only dealing with a synthetic public, not the real world."
Much is riding on the team and the project. Before merging with Price Waterhouse two months ago, Coopers & Lybrand invested millions of dollars on this cutting-edge dream in hopes that the investment would help differentiate its consulting practice from a vast array of competitors. Strong in information technology, it wants to land more high-level strategic work. "It's a way to gain respect and attention in the chief executive offices and the boardrooms where we haven't played before," says John M. Jacobs, global deputy leader for management consulting services. "If we begin getting good predictions out of it, we're going to have a lot of new clients knocking on our doors."
For years, management has sought to harness the power of econometrics and technology. Operations research and linear programming, however, rarely won the attention of the CEO because they were so esoteric and difficult to grasp. Farrell's group is hoping that the ability to see decisions acted out on a computer screen will greatly help sell the idea in the executive suite. Others agree. "American management is numerically illiterate," insists Michael Schrage, a technology consultant and author. "Instead of selling equations, they are selling pretty pictures. Visualization is the marriage of mathematics and marketing."MOVIEGOING CLONES. The potential is great. Among others, Citicorp, Coca-Cola, Shell International Petroleum, and Texas Instruments have been sponsoring research into the field of complexity at the Santa Fe Institute in New Mexico, the foremost think tank on this nascent science. Already, Farrell's work has attracted more than half a dozen clients who are convinced of the commercial appeal of using artificial life and complexity theory to build these simulations.
US West, for example, is hoping Farrell can create a model to predict how its competitors and customers would react to different pricing plans and promotions. For a major entertainment company, his group has created 40,000 moviegoing agents--cloned from a survey of actual moviegoers--to help executives determine the best way to market and distribute films. Farrell says that the model is up and working, forecasting first-week box-office receipts at accuracy rates of more than 30% better than traditional forecasting methods that rely on historical records.
But Farrell, a former scientist at Caltech's Jet Propulsion Laboratory, has a long way to go before he and his team can broadly transform the way companies plot strategies, plan products, and market goods to consumers. "We're trying to bend metal here," says Farrell, who joined Coopers & Lybrand 12 years ago after earning an MBA from Columbia University. "We're trying to apply theory to the real world. It requires some heavy lifting."
The skeptics say, moreover, that any computer model is only as good as the assumptions plugged into it. "As you get more complex systems, your ability to untangle the core assumptions becomes more complicated," says Eileen C. Shapiro, author of Fad Surfing in the Boardroom. "This could be useful, but only if it is used an an aid to thinking and not a replacement for thinking."
To pull it off, Farrell has put together an eclectic bunch of brainy misfits and rebels who quietly labor in the button-down culture of a highly conservative accounting firm. It's a group of 22 people you'd more likely find in Silicon Valley than on the third floor of a New York skyscraper on Sixth Avenue.
The core group of thirtysomething brainpower includes Robert Bernard, 31, whose resume shows degrees from Princeton University in psychology and from the University of Michigan in urban planning; Radoslaw "Radek" Zapert, 31, an intense Polish mathematician; Wei Lin, 31, a former staff sergeant in the Taiwanese military police who has a PhD in electrical engineering; and George M. Janes, 32, a self-described "classic data geek," called "the stable boy" by his colleagues because he maintains the agent population in several simulation models.ARROGANT EGGHEAD? Like Sigvaldason, 31, whose forte is visualization, many are unaccustomed to working in a corporate setting where they have to attend meetings on time, wear suits, and keep their hair neatly trimmed. Until several women joined the group, a couple of team members would show up for work in flannel shirts, jeans, and hiking boots, then strip in the lab to slip into suits for meetings.
In the windowless lab room, it's not unusual for profanity to be heard in an array of languages. The work is often long and tedious. Sometimes, team members toil until the wee hours of the morning to debug some software or meet a client deadline. Wei says his wife often scolds him for coming home late, and he wonders whether his nearly 2-year-old son recognizes him. But they're a passionate group, driven by their desire to do pioneering work. "This is not just a toy world in academia," says Bernard, who came aboard as Farrell's first employee in 1995 wearing a ponytail and an undersized charcoal-gray suit. "We're pushing the envelope."
Farrell knows how it feels to be an outsider. When he was up for partner at Coopers & Lybrand in 1993 after leading multimillion-dollar reengineering assignments at both AT&T and Nynex, he was rejected by some as an arrogant egghead. Several of the firm's old-line accounting partners could hardly understand what Farrell was talking about when he spoke of "adaptive agents" and "cyber-bioengineering," a term Farrell uses to describe the creation of his computer agents. With the help of consulting chief Jacobs, Farrell overcame their skepticism to win partnership status.
It was while managing the Nynex project that Farrell stumbled onto the then-emerging field of complexity. Seeking new ways to better organize the telephone company, he flew out to the Santa Fe Institute and met with some of the field's leading thinkers. Farrell returned mesmerized by the possibility of applying their theories to business. With others, he developed in 1994 a software simulation called TeleSim, a primitive, stripped-down precursor to the computer models he is now building. A year later, he gained funding from Coopers to launch his group.
Modeling complex, adaptive systems that mimic the marketplace is tough work. The team spends thousands of hours writing computer code to embed humanlike properties in the intelligent agents they create. The agents are modeled on consumer research surveys and databases of information. The virtual customers in Macy's model will be given characteristics such as sex, age, how much they buy, and when, from credit records and thousands of pages of cash register data from eight Macy's stores.
Farrell and others believe that by making these agents react to each other, they will provide more predictive results. "The value of this interaction is that you can build in much more real-world behavior," says W. Brian Arthur, a preeminent guru in the field who recently joined Farrell's team from the Santa Fe Institute. "It's like chemicals reacting in a test tube. You can have atoms tumbling around, meeting each other, interacting, and interleaping, and all that kind of thing."GLITCHES. Still, it's easy to overlook things humans take for granted--and that may be the biggest hurdle Farrell faces in creating a practical management tool. In an early iteration of the model built for the entertainment company, almost all the agents rushed out to see a movie in its first week. The reason: The consultants failed to account for whether the agents had enough leisure time to go to the movies.
Or consider the complications of creating a make-believe Macy's. At preliminary design meetings, Farrell and a team assigned to the Macy's project wonder what impact a sales associate's age, gender, length of service, and salary might have on consumer sales. Or, how to factor in the different demographics for Macy's Herald Square store in New York and a mall store in White Plains, N.Y.
If Farrell's team can deliver on the promise of these new concepts, they will likely give birth to a new and powerful management tool that could change the way executives manage their companies. But they have many daunting hurdles to overcome--starting with the need to show up on time for those client presentations.By John A. Byrne in New YorkReturn to top