The holidays are not just the busiest time for shopping--they're also the busiest time for shoplifting. For retailers, so-called shrinkage--which includes employee theft and return fraud, along with folks' taking the five-finger discount--is a $34 billion-a-year problem. Technology like closed-circuit cameras and electronic merchandise tags have helped to reduce it, but now stores are turning to a new tool to fight back: computer models that help them better predict which items are prone to being swiped.

The software, marketed by firms like SPSS and Security Source, uses historical data on sales, returns, voided transactions and inventory to identify suspicious patterns. The software helped Chase-Pitkin, a home-improvement chain in upstate New York, realize that outdoor merchandise like gas grills and Christmas trees was disappearing--something it ordinarily wouldn't have caught until doing post-holiday inventory. Now motion detectors and extra staffers keep those items from walking. The software also helped specialty retailer Brookstone figure out that a particular piece of home audio equipment was likely to vanish. "We were surprised that was occurring because they're so big," says Wayne McBrian, director of loss prevention. Now front-of-the-store displays contain just empty boxes; full boxes feature a security tape over the handle that can be removed only by cashiers, making it hard for crooks to carry the bulky item out the door. As the data miners highlight more vulnerabilities, they're hoping the problem of shrinkage grows ever smaller.