Date of this Version
Hall, L., Keck, Z., Shukaev, A., Yusuf, A., & Zatorski, G. A. 2020. Inventory Forecasting and Redistribution. Undergraduate Honors Thesis. University of Nebraska-Lincoln.
The goal of this product is to outline the details of this Design Studio project. The client is Buckle, Inc., the popular fashion retail store, and their problem has to do with clothing retail inventory redistribution. As a fast-fashion retailer, Buckle always stays on the cutting edge of fashion. The problem they encounter is how to accurately predict how many units of which item and size to deliver to each store in each shipment. Before this project, their approach was based on the intuition and experience of their analysts. Buckle used the opportunity of Design Studio to move to a move data-driven approach through the implementation of Machine Learning. This project is made up of two main parts: sales forecasting and inventory redistribution. The forecasting feature implemented a machine learning algorithm which was selected by Buckle based on exploration they conducted on the seasonality of their inventory data. The team worked in an initial research and discovery phase to learn more about these new technologies before implementing them. The redistribution algorithm was a more complex feature to build. The problem is extremely expensive to calculate naively, so the team had to adjust the project scope to create a satisfying deliverable. The final product that the team delivered came in four parts: First, allow developers to generate forecasts for a given inventory unit, based on the historic performance of similar inventory units. This eliminated human bias in predictions, which is consequently good and bad as humans can take into account more external factors than an algorithm. Second, allow allocation analysts to access forecasted sales. This was done through a web application interface. Third, integrate forecasted sell-through into their existing tool. Lastly, generate an inventory unit ranking page, so analysts could quickly and easily see the most volatile inventory units.