3 min read

Better Forecasting In Unpredictable Times

sidebar to main story: In Search Of The Big Picture
Compaq expects to miss its forecast for its most recent quarter's sales by more than $1 billion. It blames the shortfall on a slowing economy, fallout from the Sept. 11 attacks, and a typhoon in Taiwan that disrupted supply chains. What's worse, the company says, predicting where the business is headed won't get any easier. "A senior member of one of our customers said to me he doesn't trust anyone who thinks he knows where the buying patterns are going, because no one really knows," Compaq CEO Michael Capellas told an analyst conference last week. "We're in incredibly uncertain times."

Michael Capellas

Predicting where business is headed isn't easy, Compaq CEO Capellas says.
Businesses will never be able to predict disasters such as terrorist attacks or even when the economy will sharply shift direction. So what can business managers do? Here's the advice of several forecasting and supply-chain specialists:
  • Get forecasts from people, not from a black box. Narenda Mulani, a supply-chain management partner at consulting firm Accenture, notes that even if a forecast could be automated by a black-box computer program, doing so would be a bad idea. A forecast is valuable only if there's consensus around it, so managers actually use it to run their businesses rather than ignoring it as something they don't understand. And then it can become a self-fulfilling prophecy. "Once they use it, it's more likely to be correct," Mulani says.
  • Expect biases that favor your company or division. Fred Collopy, professor of IS at Case Western Reserve University and a director of the International Institute of Forecasters, says research shows that people overestimate their own sales and underestimate competitors' performance. He recommends at least averaging internal forecasts-operations, marketing, sales, and finance.
  • Keep it simple. Simple forecasts tend to outperform complex mathematical models, Collopy says. Forecasters must grade themselves on how well they focus on the right people, drivers, and data, and avoid unnecessary factors.
  • Forecast your customer's product, not just yours. Collopy says too many companies are forecasting the wrong things-looking at the integrated-circuit market narrowly, for example, rather than the market for products that use integrated circuits. "A great number of companies have yet to make that leap" to a broader view, he says.
  • Improve your forecasts, but depend less on them. The goal is to make products "postponable" and to shorten transaction cycles, Mulani says. "Postponable" products mean shortening development time before products are sent to market so they can be shelved if demand dries up, despite an aggressive forecast. "The prudent policy," Mulani says, "is to use every possible tactic to rely less on the forecast."

close this window