INDUSTRY NEWS

So predictable

The reason that market forecasts for the semiconductor industry are often so humanly wrong is, not surprisingly, they rely on input from humans, argues Moshe Handelsman. Handelsman's firm, Advanced Forecasting Inc. (AFI) in Cupertino, CA, recently crowed about its success in predicting a year-end upturn in equipment sales in 1998. The founder and chief seer says AFI had told its clients in the tool sector "they should expect an increase in bookings in October and November." When SEMI reported the predicted increase, AFI seized the moment to boast that forecasting can be an exact discipline with no need for "retroactive modifications."

And why was AFI's forecast so spot on? Humans tend to overemphasize the impact of sales in the most recent period and ignore the historical pattern of cyclicality in the sales data, AFI argues. This tendency leads to forecasts based on extrapolations that miss key turning points in demand. The company refuses to use past historical IC data as one of its economic factors and instead bases its forecasts solely on published numerical sources of these factors. AFI predicted the industry's "major turning points" for the past 12 years with 90% accuracy, claims Patrick Driscoll, vp of sales and marketing.

The question for AFI then is: Since the limb you're on is so short, what's gonna happen in 1999? Replies the self-assured Driscoll: "We are seeing trends that point to a very strong year ahead."

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit