Skip to Content

When Predictions Go Only So Far

Dell’s attempts to master PC pricing offer a case study in the challenges of predictive modeling.
December 28, 2010

Dell learned a tough lesson a year ago. In the quarter that included the 2009 holiday season, its revenue jumped 11 percent, which sounds like good news: the company sold more computers than it had in the same quarter the year before. But its net income dropped 5 percent during the same period. Dell was making less money on each machine it sold. Perhaps most distressing for Dell and its stockholders, one of the biggest reasons for the shortfall was that the company had been caught off guard. Prices for key computer components, especially memory chips, had risen more than Dell had expected.

A better forecast? Dell’s gross profit as a percentage of revenue has been hurt by fluctuations in component prices. After some misfires, Dell’s predictive models may have improved results in its most recent quarter. (Note: Dell’s fiscal 2011 ends January 28, 2011.)

The fact that Dell, which sells $60 billion worth of products annually, was insufficiently prepared for a price spike in its supply chain is a reminder that even some of the world’s most complex businesses have struggled to master predictive modeling, the technology at the heart of this month’s Business Impact report. Since that bad holiday quarter, the company has tried to get more sophisticated in its modeling efforts, but it’s not clear it’s had much of an effect.

To understand Dell’s situation, you have to go back to the start. After being founded in Michael Dell’s dorm room at the University of Texas at Austin in 1984, the company mastered the science of supply-chain efficiency. It was a model that made Dell the top-performing stock in the S&P 500 during the 1990s. Because it curtailed its retail store business early on and sold directly to consumers and businesses, Dell could build computers “just in time,” which meant that it didn’t have to assemble a machine and then let it sit in a warehouse or a retail location until someone bought it. Instead, it generally put together PCs only after customers had already ordered them. That meant Dell could order certain parts for its computers just days before they were needed—and often not pay for them until after the assembled computers were shipped off to customers.

But in the past few years, Dell has tried to expand its market by selling in stores. That has forced Dell to deal with several new challenges, among them that big chains such as Best Buy and Wal-Mart stock their shelves with a fixed lineup of PCs rather than customizing machines for each buyer. “We’ve had to change the entire supply chain to build fixed configurations,” the company’s chief financial officer, Brian Gladden, recently told Technology Review. And retailers order these machines months in advance, not days or weeks.

As a result, Dell must try to figure out over the summer what to charge for PCs that will actually be made and sold during the holiday season. If the price of a major component such as memory chips jumps between July and December, Dell’s profits can get squeezed. That’s what happened in 2009. Even a plunge in prices can be damaging, because the company hedges many of its component purchases to lock in prices within a certain range. If prices fall way below the expected level, it has overspent for the parts.

So in 2010, Gladden says, Dell stepped up its efforts to make sophisticated predictions about every element in its supply chain. The company has analysts who specialize in such things as LCD prices, memory prices, even the efficiency of a particular factory. Data from each analyst is fed into a larger model that higher-level managers can see as they negotiate prices with retailers and other large customers. The model also lets Dell rapidly adjust the computer configurations it highlights for people who are ordering custom-built PCs on its website; it might recommend two gigabytes of RAM in a PC one day and four the next.

How well have the changes worked? It’s too early to know whether Dell avoided another bad holiday quarter, but in the first three quarters of calendar 2010 the results were mixed. Dell’s gross profit margin, which is the percentage of revenue left over after stripping out production costs, slumped again in the first half. Brian Marshall, an analyst with Gleacher, says that Dell gobbled up memory chips earlier this year for fear that prices would go higher still.

Meanwhile, Dell’s main rival, Hewlett-Packard, largely held off on buying the same components for an extended period over the summer when its own Procurement Risk Management system predicted that prices would eventually fall. The system, which has been in place for a decade and has been the subject of case studies at the Stanford Graduate School of Business and other business schools, turned out to be right. Marshall estimates that the decision guided by its more mature model saved the company $15 per computer.

More recently, Dell seems to have bounced back—at least for now. In the fiscal quarter that ended October 29, the company had its highest gross margin since 2008 (see chart). The head of the company’s consumer business, Steve Felice, told analysts in a conference call that Dell has been making “great progress in matching our supply chain with retail buying seasons.” But analysts remain skeptical that the recent results are more than a blip. “I don’t really think we’ve seen any optimization of [Dell’s] supply chain,” Marshall says. “At the end of the day, where is the proof?”

And even if Dell can master modeling, the PC industry doesn’t give it much room to maneuver in response to price predictions. Ashok Kumar, an analyst with Rodman & Renshaw, points out that companies like Dell are often at the mercy of their suppliers. “They can model microprocessor prices all they want,” Kumar says. But “the fact is, Intel can decide to price it and Dell has to take it.” Its only alternative is to decide that advance order prices are too high and wait until the last possible moment to lock in its rate.

In other words, predictive modeling in the computer industry often functions more as a risk management tool than a driver of profits. It won’t make you rich, but you need it to reduce the chance of being waylaid by nasty surprises.

Keep Reading

Most Popular

It’s time to retire the term “user”

The proliferation of AI means we need a new word.

The problem with plug-in hybrids? Their drivers.

Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.

Sam Altman says helpful agents are poised to become AI’s killer function

Open AI’s CEO says we won’t need new hardware or lots more training data to get there.

An AI startup made a hyperrealistic deepfake of me that’s so good it’s scary

Synthesia's new technology is impressive but raises big questions about a world where we increasingly can’t tell what’s real.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.