Major retailers have long tapped predictive analytics as a way to stop employee theft, shoplifting, vendor fraud, and administrative errors. Having access to point-of-sale information makes it fairly easy to create models to forecast behavior patterns that help stores isolate existing problems to determine brands and items stolen most often. Today, midsize companies, such as Chase-Pitkin Home & Garden Inc., have begun to use predictive analytics to hold onto some of the revenue lost by theft.

At Chase-Pitkin, a division of Wegmans Food Markets, CIO Christopher Dorsey needed a way to stop the disappearance each year of $2.52 million to $3.24 million in merchandise, or 1.4% to 1.8% of its $180 million in annual revenue. So the company decided to test predictive-analytics software from SPSS Inc. in mid-2003 in one of its most susceptible departments, which sells a variety of tools, including power tools.

Within a year, Chase-Pitkin saw a return in its investment. So far, the application has reduced theft by more than $540,000, Dorsey says. "SPSS gives us access to clean data faster, so we can identify merchandise that is most susceptible to theft, down to the brand and the item level," he says. "Out of the 10,000 items we carry in the power tools and tools department, we now know that 16 items represent half the missing inventory for our stores, which carry about 38,000 active part numbers."

Having the data to identify items at risk helps store clerks remain vigilant. In fact, the application has been so successful that Chase-Pitkin will make the SPSS tool available to all departments and stores throughout 2005, beginning with plumbing in late January.

Department managers at all 15 home-improvement stores throughout upstate New York will gain access to the SPSS predictive-analytics tool through a server at its corporate headquarters. The application interfaces with Chase-Pitkin's enterprise-resource-planning software. Data is fed from point-of-sale systems, inventory replenishment, and warehouse-management tools. Information also is received from applications that identify merchandise from store transfers and defective products. Information is updated daily.

The kind of information unveiled by predictive analytics is invaluable to retailers. A combination of employee theft and shoplifting in 2003, the latest numbers available from an annual study conducted by the University of Florida, cost retailers $15.8 billion, according to Richard C. Hollinger, professor at the university's Department of Criminology, Law, and Society. That number jumps to $34 billion annually when errors and vendor fraud are factored in.

Hollinger wouldn't estimate 2004 financial losses, which are scheduled for release in June at the National Retail Federation conference in San Diego. He did say the percentage of loss from theft has accelerated in the last few years to nearly 50%, of which employee theft accounts for approximately 47%. Compare this with 30% in the mid-1930s, when shoplifting and employee theft each contributed about a third to the losses. "Administrative markup and markdown problems have decreased because of point-of-sale systems and technology, but the employee theft correspondingly has increased," he says. "Retailers are beginning to realize that many external problems are really internal or hybrid, where employees allow shoppers to steal."

SPSS also has helped Chase-Pitkin better understand how products leave the store. Consider the combination set with a drill, bits, and hand sander that retails at Chase-Pitkin for $199. SPSS revealed that inventory counts didn't match up with sales. Upon further investigation, store cameras caught one customer stuffing the tools into a box belonging to a $19.99 lamp, and then tossed the toolkit box behind a gondola rack.

The analytics tool also identified that Chase-Pitkin had been selling at one store what Dorsey calls "a fair amount of a specific gas grill" model during the first three weeks in December. This led management to conduct a quick cycle count that pointed to the store's loading yard where customers drive through and employees load heavy products into their cars. Somehow, customers were getting the $499 stainless gas grill instead of the $199 model purchased.

The SPSS application helps Chase-Pitkin identify the item and creates awareness to monitor how the theft or errors occur. "Predictive analytics is a powerful tool," Dorsey says. "In the power tools and tools department, theft has decreased by 50% based on new business processes in place."

This story was modified on Jan. 13.