Identificação de rupturas de estoque no varejo alimentar através de técnicas de soft computing

Abstract

Out-of-stock (OOS) leads to significant losses for retailers, and despite decades of study, out-of-stock rates have not decreased considerably. The first step to improve this indicator is accurate measurement and detection of OOS. Therefore, the aim of this study is to determine the effectiveness of a method for identifying OOS in the grocery retail sector based solely on transactional data. The research focus was exclusively on identifying cases of total OOS caused by the actual absence of the product at the point of sale, i.e., when the lack of sales in a certain period was solely caused by stock unavailability. The developed algorithm used fuzzy logic techniques to determine the occurrence of OOS based on probabilities derived from zero sales occurrence and its possible sequence, accumulated sales, and balance from the last entry. This algorithm was applied to 2,022 selected items over a 12-month period and achieved a correct identification rate of 70.2% with only 16.4% false alarms. Although some previous studies have achieved superior results, their approaches were restricted to a very small number of items. Furthermore, the application of the model proved promising, identifying OOS on average only 4 days after their occurrence, which may result in an estimated 1.3% increase in the company's total revenue.