Using predictive modeling for detecting Out-of-Stock. Case: distributor company in retailing industry

During the recent years of slow economic growth, companies have found it increasingly important to discover new and more efficient ways of conducting their business and to find new sources of value. As a result of the current economic situation and the surge in technological innovations, concepts such as Business Intelligence, Big Data and Data Mining have gained considerable attention in today's business world. In this thesis, my objective is to apply these relatively new techniques for solving a problem that retailing companies around the world have long been confronting. This problem is called Out-of-Stock and it has already been addressed by the scientific commune over decades ago. Despite the broad research available on the Out-of-Stock phenomenon, the studies focus on solving the problem from a retailer company's perspective and neglect the other entities that are also affected by the same problem. This thesis aims to fill this gap by conducting the research from a distributor company's point of view using data that is available for them. Similarly to a retailer, a distributor's business suffers when its products are not available for consumers because of an Out-of-Stock situation however, the distributor's options to solve the situation are much more limited. This thesis will examine three main research questions: "How can a distributor company in the retailing markets use data mining techniques in order to predict Out-of-Stock situations?", "How accurate are these predictions?" and "What are the most useful variables in predicting Out-of-Stock?". In order to answer these questions, a case study was conducted with data that was available for the case distributor company. Several predictive models were ran on SAS Enterprise Miner data mining software in order to measure the prediction accuracy that could be obtained with the case company's data. The results of the study show that Out-of-Stock situations can be predicted at certain accuracy even with the limited data obtained from a distributor company. However, the research fails to produce clear universal rules for predicting Out-of-Stock because the data was available for only one company, one retailer chain and over a short period of time. Nonetheless, this study manages to provide a framework on how the Out-of-Stock problem could be approached by a distributor company and bearing in mind the limited amount of data that was available for this research, the results of the predictive models can be seen as very encouraging. For the future research, I suggest similar studies but with a retailer company's perspective. Based on the results of this research, I believe that with the more elaborate data that is available for the retailing companies, the Out-of-Stock situations can be predicted with greater accuracy and therefore these predictions would be more helpful in managing the problem