Case Fill Rate Prediction

Stockouts present significant challenges for Fast-Moving Consumer Goods (FMCG) companies, adversely affecting profitability and customer satisfaction. This Capstone report investigates key drivers causing Case Fill Rate (CFR) to fall below target levels and identifies the best model for predicting future CFR for the sponsor company. By utilizing hypothesis testing and feature permutation techniques, we conclude that forecasting error is the most critical driver influencing CFR. Machine learning techniques including classification, regression, gradient boosted trees (XGBoost), convolutional neural network-long short term memory model (CNN-LSTM), and multistep LSTM techniques were deployed to predict CFR. Advanced machine learning techniques demonstrated potential in predicting short term CFR. To improve longer-term forecasts, a combination of models should be incorporated, along with extended historical data, promotions data, and consideration of exogenous variables. Companies should prioritize forecasting accuracy and optimize inventory policy to improve CFR in the long run.