Chipmakers Compound Forecasting Problem

There are still too many wrinkles in the electronics industry supply chain. Many of them came under a harsh spotlight following the March 2011 earthquake and tsunami in Japan and the more recent flooding in Thailand. One fact stood out from both disasters: The supply chain is still operating without enough safeguards or redundancies that would guarantee continuity of supplies.

Natural disasters cannot be accurately predicted, but the industry has the resources to simulate all kinds of scenarios and thereby plan a course of action that would limit the impact of such occurrences. It might be a vain hope. Today the electronics industry supply chain is in the hands of folks who spend a great deal of their time obsessing about short-term gains, especially their own.

The industry urgently needs a different approach — one that lets companies compete hard against rivals but propels the idea of group survival to the forefront of business discussions and decisions. This might be another vain hope, considering the roots of the current list of industry problems. The inconsistencies we see in the electronics supply chain can be traced to a deeply flawed forecasting system that is about to get worse as a result of actions by many of the industry's biggest players.

Recently, {complink 2657|Intel Corp.} and {complink 103|Advanced Micro Devices Inc. (AMD)} announced that they would withdraw from WSTS (World Semiconductor Trade Statistics) Inc. The announcements followed other moves that I believe will severely limit the ability of trade organizations to compile reliable data on industry sales, capacity utilization, and book-to-bill ratios, which match orders against actual sales. In this article, I'll discuss what this means for the industry's ability to function effectively at a supply chain level.

My colleagues at EE Times have discussed this issue from other perspectives, and I recommend their articles for additional information. (See Opinion: Did Intel just break my glasses? and End times for industry cooperation?)

Forecasts are already a joke in this industry. Many executives know general sales predictions are useful only as markers on a barren land. Their utility rests in providing very generic guidance, rather than GSP-caliber instructions. In other words, we all know enough to read and take seriously the disclaimers at the bottom of any forecast page.

Granted, the art or science of foretelling the future is not error-free, but the numbers many electronics businesses rely on require more than a pinch of salt.

Take the annual semiconductor sales forecast from the Semiconductor Industry Association. The actual numbers reported by chipmakers at the end of the year typically bear no resemblance to the forecasts provided in January. The industrywide total can run anywhere from $20 billion to $50 billion above or below the forecast — a staggering difference in an industry generating about $300 billion of sales per year. Yet we continue to base production facility construction decisions and capital equipment purchases on these figures.

In November 2010, the SIA predicted $318.7 billion of sales for 2011 and $329.7 billion for 2012, versus a projected $300.5 billion for 2010. The actual 2011 total reported by the industry association was $298.3 million. It's anybody's guess whether the industry will hit or exceed the initial projection for this year.

The lack of accuracy in industry sales projections is understandable, since various factors, including natural disasters, can disrupt regular business activities. However, these factors are not the only uncertainties injected into the system. Industry players too often dig a hole under themselves by failing to provide accurate and sufficient information to organizations that compile such data. Without this information, the industry is walking around blindfolded. It's a typical case of garbage in, garbage out — the forecast output is only as good as the information on which it is based.

Shouldn't we be pulling back from this by now? I thought the response should be an unqualified yes, but the decisions by Intel and AMD to withdraw from WSTS will only compound the industry's forecasting problems. Intel is by far the industry's biggest chipmaker by revenue, and without its input, WSTS might as well call it quits. There's no doubt other industry players will follow, including {complink 4751|Samsung Electronics Co. Ltd.}, the No. 2 market player. Soon we'll be back to calling up each chipmaker for information on sales projections, which it may or may not provide. Being blindfolded was bad enough.

2 comments on “Chipmakers Compound Forecasting Problem

  1. Barbara Jorgensen
    March 6, 2012

    I think there's a really good term for this called the Goliath Fringe (credit to Charlie Barnhart). Companies get so big (Foxconn, Apple) that they skew results. That is not to imply that abandoning organizations such as WSTS is the right thing. Big companies are also seen as leaders and bellwethers. I agree–exiting the WSTS is bad for the industry as a whole.

  2. Eldredge
    March 7, 2012

    @Bolaji:  I certainly agree that forecasting relies heavily on using accurate historical data, and if key players refuse to provide data, forecasting becomes much more difficult. But it also seems like +/- 20B$ fluctuations on a 300B$ base (roughly +/- 7%) isn't an unrealistic expectation on a year-long forecast. Granted, if the fluctiation goes as high a 50B$ (+/- 17%), that would be painful.  

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.