Evan Schuman's StorefrontBacktalk

Some of the nation’s largest retailers pay Paul Walsh to think about existential-sounding questions like “How cold is cold?” and “How wet is wet?”

No, Walsh isn’t a philosopher. Indeed, as a trained scientist (a meteorologist, to be precise), his answers are anything but theoretical. But it certainly is about perception.

The temperature where people start feeling cold?and start acting cold?varies sharply across the United States and certainly globally. And yet until recently, retailers made buying decisions based on fixed?and usually incorrect?assumptions about how consumers react to weather.

“If you live in Seattle?where it rains an awful lot?your behavior on a rainy day is very different than if you lived in sunny L.A.,” said Walsh, the senior vice president for client services at Planalytics, in Wayne, Pa.

Another Walsh example is how Americans react to three inches of snow. In some neutral parts of the country, schools are closed and traffic is delayed. In Buffalo, N.Y., that’s considered flurries and no one notices. “You get three inches in Atlanta, and that’s Armageddon,” he said.

Many huge global retailers?including Bloomingdales, 7-Eleven, Kmart, JCPenney, PetSmart and The Home Depot?pay Walsh’s company to analyze weather patterns and customer buying patterns and predict likely buying patterns.

“More and more companies are looking at this information very carefully, trying to get an understanding why their customers do what they do,” Walsh said. “What is the weather in Buffalo, and how warm does it have to get before people start buying shorts versus how cold does it have to get in Orlando before people start buying sweatshirts? There’s no voodoo about this. It’s pure statistics.”

Planalytics’ clients run the gamut, from retailers worried about seasonal clothing to convenience stores that have to know what cold/hot weather items to keep on hand to companies that fill ATMs with cash, who need to project the absolute minimal amount of money they need to keep the machines happy.

Not only do retailers have to understand that 45 degrees Fahrenheit is viewed as quite warm in New Hampshire in February and brutally cold anytime in Los Angeles, but they also have to reject historical retail data when projecting likely sales demand.

“The problem that we’re solving is that, by and large, retailers use last year’s sales to predict this year’s season,” Walsh said. “When you forecast using last year’s sales, that only works if you believe that this year’s weather will be exactly the same as last year.”

These projections go beyond the obvious. In weather perceived as bitter cold, consumers avoid large shopping malls. But those who do make it are committed and are extremely likely to make purchases. When the weather is perceived as nice and comfortable, a lot more consumers go to the mall, but they are more likely to be tire-kickers and browsers, Walsh said. Few people feel like browsing during a blizzard.

Walsh’s company typically delivers a seven-day immediate weather and buying pattern prediction report for each client, focusing on appropriate geographies. But they also deliver an 11-month prediction, which Walsh says is accurate about 75 percent of the time. The reports are delivered on the Web and are usually dumped right into Excel spreadsheets.

A key foundation of Planalytics’ value-add is the argument that actual weather motivates consumer buying habits a lot less than weather predictions broadcast on TV and radio. Therefore, Walsh’s meteorologists are certainly concerned with accurately predicting the weather, but they are even more concerned with accurately predicting the TV predictors.

“Weather forecasters drive demand more than actual weather does,” Walsh said.

Given the fact that TV forecasters generally rely on professional weather agencies such as the National Weather Bureau or AccuWeather, how is Planalytics able to predict accurately and earlier? Walsh’s answer is that his team can’t predict earlier or more accurately, but his team is willing to reveal those predictions (to paying clients) a lot earlier.

When hurricane patterns start to emerge, just about all professional meteorologists can spot them at about the same time.

“Our team and the major weather services are looking at the same things, but they won’t come out and start talking publicly until they’re really comfortable with the prediction, until they can pretty much see the whites of its eyes,” which typically happens about 72 hours before the hurricane will land, Walsh said.

“From a retail client’s perspective, the best storm is the one that threatens, but doesn’t hit. They get the surge of business, but not the destruction,” he said.

But even a hurricane’s prediction impacts different clients differently, and that has to come through in Planalytics’ analysis reports. “When hurricane reports come out, people are standing in line, waiting to buy things,” Walsh said. “But they’re not thinking about buying designer fleece. They’re looking at plywood.”

For 7-Eleven, Planalytics analyzes buying patterns and makes some unusual suggestions. When a hurricane is predicted, it recommends stocking up on magazines and other entertainment items.

Why? Parents fearing evacuations want to be able to quickly keep children occupied and happy during a potentially long and very tense evacuation. “We know what people do when a hurricane threatens,” Walsh said.