Monday, July 9, 2001

'Lousy' Sales Forecasts Helped Fuel the Telecom Mess
By Dennis K. Berman
Staff Reporter of The Wall Street Journal

How did the telecommunications industry get it so wrong?

As late as last November, Merrill Lynch & Co. predicted that domestic equipment purchases by phone and data carriers would grow by 15%, to over $65 billion, in 2001. Now, the brokerage firm believes such purchases will fall 7% this year.

A year ago, Nortel Networks Corp. said it would spend $1.9 billion to boost production and add 9,600 jobs to meet "explosive customer demand." Last month, Nortel said it would lose $19 billion this quarter and eliminate 10,000 jobs -- on top of 20,000 previous cuts.

What happened? The blame lies, in part, with the calculations used to predict future sales and inventory levels of telecom gear, and in turn feed into Wall Street earnings guidance. Such forecasts are "lousy," says Gregory M. Duncan, head of the telecom practice at consulting firm National Economic Research Associates, and former forecaster at GTE Labs.

Even Nortel Chief Executive John Roth in April chided his own company's demand forecasting methods after he chopped 2001 expectations to a large loss from a healthy profit. "As recently as October of last year, all the conversations I was having with my customers were, 'John, you haven't shipped me enough equipment yet, when are you going to get the volume up?' " Mr. Roth said. "The people we did not talk to in our customers were the treasurers, who found out in January that they were having difficulty in raising the money" to pay for equipment.

Frederick Fromm, president of Oplink Communications Inc., a Nortel supplier, says his company's forecasts over the last few years were based mainly on informal conversations with customers. "Quite openly, our forecast was that demand was exceeding our capacity, [so] let's build capacity as fast as we can," he says.

Oplink, which makes devices that transmit light over fiber communications networks, eventually overextended itself. It reported a fiscal third-quarter net loss of $25 million, including a $19 million charge for excess inventory.

To be fair, at this time last year, detailed analysis of demand in the then- booming telecom industry seemed a bit superfluous. Spending by phone companies on gear nearly doubled from 1996 to 2000, to $47.5 billion, according to the Telecommunications Industry Association, whose method of calculating that figure differs from that of Merrill Lynch.

At the same time, many telecom companies had fewer experts available to forecast demand. David Loomis, an Illinois State University professor and organizer of the International Communications Forecasting Conference, estimates that equipment makers and telecom carriers let go half of their forecasting employees over the last five years, chiefly for cost-cutting reasons.

Yet such forecasting expertise seems necessary in an industry undergoing the turmoil of deregulation and one that lacks many of the government statistics and research methods used in other businesses, such as construction and manufacturing.

For the automobile industry, for example, the U.S. Departments of Commerce, Transportation, and Labor publish volumes of figures showing everything from the average age of the U.S. automotive fleet to the fuel consumption levels of trucks with more than six tires, which help auto makers predict what kinds of trucks will be purchased and how they will be used.

Though the Federal Communications Commission publishes statistics on the volume of voice-call minutes that phone companies transmit, the agency publishes virtually no statistics on data traffic, which constitutes an increasingly large part of telecom-network use. "The paucity of information stems partly from the difficulty of defining 'data traffic' and partly from the difficulty of measuring it no matter which definition is selected," says one FCC official. "There is no real agreement on what data traffic is or how to count it.

Also, in industries such as computers, both makers and customers participate in surveys by firms such as PC Data and International Data Corp., which create third-party snapshots of overall market activity and demand. But there's little such cooperation in the telecom business, in which carriers tightly protect their network usage statistics. With a reliable network-usage number, equipment companies could at least do some basic historical modeling with confidence, says Tim Stronge, research director at TeleGeography Inc., which publishes telecommunications statistics.

Researchers for equipment companies rely mainly on feedback from their sales forces. They are influenced by the dozen or so research houses that regularly publish estimates, and increasingly they take into account overall economic conditions. But as the parade of earnings warnings and restructurings by telecom-equipment makers illustrates, "You're basing guesses on guesses and that multiplies the ability to get it wrong," says Mr. Stronge.

At Finisar Corp., a maker of optical components, Chief Financial Officer Stephen Workman struggles with the information provided by outside researchers. "They're usually six months behind what's really happening. It's almost laughable," says Mr. Workman. To help compensate, he has begun talking to the actual users of telecom networks, not just the carriers.

Recently, some companies scrambling to stop overproduction have started sharing information. "We're seeing sharing industrywide, because when there's a dip, everyone is looking for an answer," says Corning Inc. Chief Financial Officer Jim Flaws. Corning has begun sending detailed weekly purchase forecasts for optical components to Oplink. Working from those numbers, Oplink allows Corning to change its orders up to 45 days before delivery. Mr. Fromm says the new system allows Oplink to plan its production far better than before.

But taking customers at their word can be dangerous. Lucent Technologies Inc., for one, has suffered from overeager buyers such as Winstar Communications Inc., for which it was contractually obligated to help build a wireless data network, even while Winstar hit financial straits. Lucent forecaster Alex Bangash now includes a financial-viability analysis in his overall method, to "look at the complete value chain" of Lucent buyers. "You can't look at just your customers, you have to look at your customer's customers," says Mr. Bangash.

At RHK, a San Francisco-based telecom research firm, chief analyst John Ryan has instructed his 60 researchers to more thoroughly examine vendors' existing inventories and the specific price discounts demanded by procurement managers. But even with his new rigor, Mr. Ryan concedes, "We're all riding this massive storm, and no one really knows where we're going."