Modeling merchandise returns in direct marketing

Abstract

Returns are a significant problem for many direct marketers. New models to more accurately explain and predict returns, as well as models that will allow accurate scoring of customers and merchandise for return propensity, would be useful in an industry where returns can exceed 20 percent of sales. We offer a split adjusted hazard model as an alternative to simple regression of return times. We explain why the hazard model is robust and offer an example of its estimation using data of actual returns from an apparel direct marketer.