One of the hottest buzzword in today’s businesses is “Big Data”. Based on the new era of Facebook, Google and others, the main idea is to make as much use as possible of a company’s internal and external data, to gain knowledge and create a more informed decision making process. As it turns out, deriving useful information from this data is a tedious task and nowhere near as easy as it sounds.

Junior Professor Dr. Anna-Lena Sachs (University of Cologne) and Professor Dr. Stefan Minner (TUM School of Management), in a recent publication, bring usefulness to the huge amount of company data for inventory decisions in practice. Based on the case of a large European retail chain, they propose an optimization technique for deriving order quantities for perishable products in retail stores. A relatively accurate estimation of future sales provides the basis for determining how much to stock and thus keep available during the day in order to maximize profit. The key is to make an estimation based on those historical sales that actually did not take place because the desired product was out of stock. Sachs and Minner use the company’s Big Data of historical sales and derive from that data what would have been sold given what has been sold on other days at the same time. Through this data-driven approach, retailer decisions on how much to keep in stock is optimized, resulting in a better availability of products while at the same time keeping inventories low.

Another inventory-based problem setting in retail arises when we look at the delivery of products from a central warehouse to the retail stores. Retail demand is highly volatile and prone to seasonal effects during and between weeks. For instance, one can imagine that fresh meat for a barbeque on a hot Saturday afternoon induces a different sales behavior than on a rainy Tuesday morning. Although these fluctuations would make flexibility in the delivery of products desirable, such flexibility is actually quite impractical. Managers need to keep operations at warehouses and stores stable to reduce the operational hassle of constantly changing workload for employees. Again using Big Data from a retailer’s product sales, patterns can be derived that indicate when to deliver a certain product to a certain store (e.g. every Monday and Thursday or only on Fridays). Such patterns remain stable over the planning horizon of several weeks. In comparison to traditional approaches using estimated sales values, rather than the true data, costs can be cut drastically, though the approach remains easily usable and interpretable for the company. The two examples from retail practice show how the buzz of Big Data can be put to good use to actually save money in practice, and this is obviously not limited to the cases presented.

So hop on board and let your data drive your decisions!

Author:

Prof. Dr. Stefan Minner
Logistics and Supply Chain Management

E-Mail: logistics.log@mgt.tum.de

Chair: Logistics and Supply Chain Management

SOURCE

Sachs, A.-L., Minner, S. (2014) The data-driven newsvendor with censored demand observations. International Journal of Production Economics, 149 (March 2014), 28–36. Contact: Prof. Dr. Stefan Minner, stefan.minner@tum.de