Welcome to the MIT SCM Research Symposium 2024 and thank you for joining us! We hope you find the day exciting and informative. To enhance your experience, we are utilizing this platform in order to help identify projects of top interest to you. Below you will find not only the schedule, but also the abstracts and attached executive summaries of each project. Additionally, you can better acquaint yourself with the student presenters and advisors through their profiles. Thank you again and we look forward to hosting you.
FMCG companies are facing the dilemma of improving customer service while keeping out-of-stock and reducing supply chain costs. One way to overcome this is to improve demand forecast accuracy. Whereas achieving this improvement is critical for business success, and using state-of-the-art methods should be highly spread among companies, most FMCG companies still rely on more traditional approaches. AB InBev is trying to upgrade its forecasting methods. The company has reached a maturity level in demand forecasting and uses a mix of time series and machine-learning models. Nevertheless, this hybrid approach has obtained only 60% of demand forecast accuracy in Colombia's market leaving room for improvement. Moreover, AB InBev bought “BEES,” an e-commerce platform for its business-to-business channel, and expects to use BEES’ data as the main input in its demand forecasting process. In this capstone project, we made a proof of concept for two aspects of interest for AB InBev: (i) Is BEES data mature enough to be useful in a demand forecasting context, and (ii) what is the expected forecast accuracy lift obtained using neural networks as a demand forecasting technique. For the latter, we applied a global Sequence-to-sequence recurrent neural network model to forecast the demand of 289 different SKUs-DC combinations. Based on our analysis, we provide several recommendations to improve demand forecasting accuracy successfully.
