There has been a proliferation of e-commerce companies that are digital-first brands in the last few years: firms that solely sell their products online, occasionally opening a few retail storefronts. However, it is difficult for these companies to accurately forecast demand and account for uncertainty in order to appropriately stock inventory, especially for new products that have no past data. Companies lose millions of dollars each year due to under-forecasting and over-forecasting errors that result in missed sales and unnecessary inventory costs.
Our team’s goal was to design a model for e-commerce companies to more accurately forecast demand for both existing and new products, utilizing various methods including smoothing, regression, and TBATS models. We have partnered with an e-commerce makeup company and used their data to create and test our models.
With our final product, an online user interface, a client is be able to upload their historical order data and receive an output that provides graphics and information on expected future demand. With flexible lead time and order inputs, our model and interface can be used by companies in a variety of industries to help them more accurately forecast product orders and needed inventory.