The Follies of Holiday Forecasting
With so many corporate fortunes decided in the fourth quarter, smart companies are rethinking the task of seeing Christmas future.
By Laura Rich

(Business 2.0) – Last winter, swarms of shoppers invaded Restoration Hardware clamoring for one of the retailer's hottest holiday items: a sleek red cabinet with 25 tiny drawers that counted down the days to Christmas during Advent. But in the middle of the shopping season, the cabinet sold out. Meanwhile, a less festive white version gathered dust on store shelves.

In January, Restoration Hardware reported that a shortage of seasonal items like the red cabinets helped drag down holiday sales by 3.5 percent compared with the previous year, and as a result, income for the quarter was down 18.6 percent. The problem wasn't shipping or tracking; the company had reportedly installed software to record where products were in its supply chain. Rather, Restoration's buyers failed to accurately predict consumer demand. By the time they realized that certain items were hot sellers, it was too late.

Demand forecasting is never easy, in large part because elusive factors like taste and consumer confidence—even acts of God—drive purchases. Yet with 27 percent of U.S. retail sales occurring in the fourth quarter, the consequences of misguided holiday forecasts can be dire. And it's not just retailers and manufacturers that suffer the effects of flawed projections. Service providers that set expectations too high have to write off wages like so much bad inventory, while those blindsided by excessive demand lose potential sales, not to mention customers who don't easily forget the time a company couldn't deliver.

Nevertheless, business planners routinely misread the market. But now a growing number of companies are turning to new technology and techniques to slay the forecasting ghosts of Christmases past. Of course, there's no such thing as a crystal ball for gauging future demand, but many forecasting foul-ups are eminently preventable in the present. Here are the most common reasons firms are caught by surprise—and what the smart ones are doing to get better prepared.

1. Missing the Small Picture

As Restoration Hardware learned with its cabinets, forecasts can fail if they don't reflect item-specific considerations. Warner Home Video, for instance, found that its sales projections for individual DVDs were off by an average of 40 percent in 2003. According to Thomas Tileston, Warner's executive director of data management, "We notoriously overforecasted teen romances." And that directly hurt profitability. "You do better when you have a good idea of how much you need to manufacture," Tileston says.

This year Tileston has been using enterprise software provider SAS's statistical modeling applications to weigh variables such as genre, stars, box office, and audience demographics when making title-specific forecasts. Before Warner releases the DVD for Troy this winter, for instance, Tileston will have created a forecast based not only on how well previous Brad Pitt movies have sold but also on how similar action-adventure releases—such as The Last Samurai—performed. As a result, he says, Warner has slashed its average forecasting error to about 13 percent in 2004. According to Charles Chase, SAS's forecasting-software expert, it's important for buyers to refine computer-generated forecasts with human knowledge about why a particular color or style—or cast—might be hotter or colder than in the past.

2. Ignoring Mother Nature

Companies love to blame forecasting gaffes on unseasonable weather. Coca-Cola Bottling and Ann Taylor cited the unusually harsh winter of 2002 as a reason for missing revenue expectations, and restaurant chain Bob Evans blamed snowstorms for poor sales last December. "If the season doesn't feel like it's supposed to feel," says Paul Walsh, senior VP for client services at weather data provider Planalytics, "people don't buy."

Companies like Planalytics (see "Can This Weatherman See Your Future?," August 2003), WeatherData, and Japan-based Weathernews help retailers and manufacturers figure out how sales of their products vary based on factors such as temperature and snowfall. Gap, Home Depot, and J.C. Penney rely on Planalytics's weather-driven demand estimates—forecasts of the percentage change from the previous year that retail sales will rise or fall due to weather factors. This year Planalytics is predicting a cold, early winter that will lift profit margins and sales of apparel and have a negative impact on consumer electronics.

3. Politicizing the Forecast

Employees from different functional areas have agendas that can diminish the accuracy of holiday forecasts. Salespeople who aren't achieving targets, for instance, favor especially low numbers so they have a better chance of hitting goals in the future. Finance departments, attuned to the costs of excess inventory, are likewise conservative, while production planners shoot high to avoid being blamed for lack of product. "Fewer headaches if they overproduce," says Chaman Jain, professor of economics and finance at St. John's University in New York and editor of the Journal of Business Forecasting.

Jain and others say accuracy is also affected by where forecasting personnel sit in an organization. According to SAS's Chase, who previously handled forecasting at Heineken USA and Johnson & Johnson, it's best if forecasters report directly to a CEO to avoid departmental bias. Another good place for them is market research, which often has access to customer data and modeling tools. "Companies typically put forecasters in finance or operations, two of the worst places," Chase says. "Too far from the market."

4. Out-of-the-Loop Suppliers

Last December, after redesigning its stores to emphasize a strategic expansion into consumer electronics, Gateway saw strong initial customer response. So strong, in fact, that two of the company's most popular products—plasma televisions and a home-theater-compatible MediaCenter PC—sold out early in the holiday season. Blame fell on Gateway's PC parts supplier as well as Taiwanese TV maker Sampo for underestimating demand. But it was Gateway that posted a $114 million loss for the quarter and shuttered 188 retail outlets. This year an industry group consisting of 120 companies, including Best Buy, IBM, and Procter & Gamble, issued a new "collaborative planning, forecasting, and replenishment" standard (see cpfr.org) with guidelines for sharing forecasting data across the supply chain.

5. Hits Happen

With retailers placing their orders as far as six months in advance in many categories, there is not always much a company can do when one of its offerings turns out to be either a total dud or the next Cabbage Patch Kids. Except, of course, identify which of its assumptions were wrong and make a better forecast next year.