Assessing the Frequency and Causes of Out-of-Stock Events Through Store Scanner Data

Both retailers and manufacturers see in-store out-of-stock events (OOS) as a major problem, but there is a lack of research about their frequency, the sales losses they generate, and their causes. We provide a twofold contribution: We describe a new sales-based measure of OOS computed on the basis of store-level scanner data, and we identify several of the main determinants of OOS. We also introduce a significant distinction between complete and partial OOS. In both types, the observed sales level is significantly below its expected value. Complete OOS occur when there are no sales at all; partial OOS takes place when sales, though abnormally low, are not zero. Our analysis of seven different data sets reveals that complete OOS are far less frequent than partial OOS. In addition, complete OOS are more frequent in stores with lower category sales and for stockkeeping units (SKUs) with lower market shares. In contrast, partial OOS are more frequent in stores with higher category sales and for SKUs with higher market shares. With regard to the impact of assortment size in the store, we find mixed results. Finally, we find that variables related to the segment to which an SKU belongs, the manufacturer, and the package format all have a significant impact on both partial and complete OOS.