The Truth About Forecasting: Part Two--Obviousness

In the first part of this series, I wrote about the errors that make most forecasts meaningless, and gave examples of how I committed most of them in my very first job. Now, I'd like to tackle the one error that I didn't make at that time, the error of obviousness. A forecast that tells you what you already know isn't a forecast, it's redundant. This example comes from when I was working for Toshiba in the late 1980s. I had a conversation with my boss, Hank Yamamoto, about where the design of laptop computers was going. Keep in mind that the standard at this point was VGA (640 x 480) monochrome LCD displays, with Toshiba and Fujitsu also selling portable computers with monochrome plasma displays. Toshiba was already experimenting with pen computers, and was delivering a small number of them to customers.

Hank pointed out that there were several areas in which laptop design would change over time:

• Processors would get faster
• Displays would move from static to active-matrix thin-film LCDs (better for handling graphics), resolution would improve, and color would become affordable
• Hard drives would get bigger and faster
• Memory would also get bigger and faster
• Battery capacity, and thus run-time, would improve
• Everything would get cheaper

That was 1989, and all of that happened. Everything that Hank said was simply an extrapolation of the components and capabilities of existing laptop computers. He didn't try to forecast what prices would be for each of the components, or for the finished computers, at any given time, but Moore's Law could have provided guidelines for timetables. The same extrapolations can still be made today; the only component that we didn't consider back in 1989 was flash memory, which is now being used to replace hard disk drives.

There were plenty of companies in 1989 that were selling research reports and forecasts that stated essentially the same things that I just listed above, albeit with more charts, graphs and tables. These reports sold for thousands of dollars, and would have told us what we already knew. However, these reports usually added prices, dates and even sales quantities, most of which turned out to be wrong. Companies that bought those reports and relied on their forecasts were in worse shape than those that simply used the component breakdown and extrapolation method. Hank knew that he couldn't forecast prices, dates and sales quantities, but he could forecast the direction of development and its eventual payoff.

In the third and final part of this series, I'll revisit the five sources of error, examine what I consider to be the worst ones, and discuss a few ways to be a better forecaster and consumer of forecasts.