Forecasting: A Recurring Nightmare

"Garbage in, garbage out" is a much used phrase in the electronics and information technology market. Oftentimes the axiom is applied by supply chain professionals to the sales, manufacturing, and procurement data generated by economists, researchers, and accounting experts who believe they have a handle on how businesses will perform in the near and immediate future.

Few believe the forecast numbers, yet many cite them. Unfortunately -- despite the distrust -- many critical business decisions are predicated on companies achieving these often elusive figures. Reputable organizations trot them out on a yearly, semi-annual, and quarterly basis, often revising them as new data comes in. The latest forecast may be updated mere weeks before actual sales figures are announced. Over the years, I have come to the conclusion that the only numbers worth accepting, within reason and barring accounting shenanigans, are final, verified shipment and sales figures. The rest is bunk.

Let's take numbers from the Semiconductor Industry Association. In November 2010, the industry group projected sales for 2011 would rise 6 percent to $318.7 million. "The projected compound annual growth rate is estimated to be 13.4 percent for the period 2009 through 2012," the association said in a report. It also said sales for 2010 would be approximately $300.5 billion. Only a few months earlier, though, the SIA had predicted 2010 sales would be $290.5 billion.

The earlier forecast had been issued in June 2010, which meant major changes had occurred within a span of five months to raise projected industry sales by $10 billion. That's a huge swing for the industry, and it's natural to wonder how semiconductor vendors cranked up manufacturing to meet the extra demand.

Somehow, the industry stretched its manufacturing capacity and satisfied most of the foreseen demand. Total sales for the year were $298.3 billion -- slightly below the updated SIA forecast, but still a good batting average for the organization.

The SIA's forecast for 2011 chip sales has since changed slightly -- the latest projection, issued in June, calls for sales to hit $314.4 billion. And that forecast will most likely be altered again before the end of the year.

The moving-target scenario is unavoidable, and the cost to the industry is staggering. In many ways, the electronics supply chain is similar to an orchestra; all the parts must fit snugly without any discordant tunes. If forecasts prove wrong or are unreliable, procurement will be hurt.

Why does the industry keep generating unpredictable numbers, and why has the practice of forecasting found such a major position in the analyst community? The answer is simple: A bad forecast may be better than none. Forecasts are something the industry can use to establish a floor for determining other requirements, including manufacturing, procurement, logistics, and total output. A forecast, even one that keeps shifting, is better than a finger stuck in the wind.

The only antidote for the unreliability of sales forecasts for the electronics industry is continuous monitoring. Once a forecast has been issued, manufacturers need to update it constantly and match it with actual sales results. Eventually, the numbers will make sense -- once they have been adjusted.