Demand Forecasting for Inventory Optimization

Our Solution

• In the discovery phase, it was found that the data provided by the client didn’t have any seasonality or trends.
• BizAcuity identified variables with maximum data points that were already available
• After identifying necessary variables, a time series-based forecasting model was developed by training the model with the past data. Once trained, the model was tested for accuracy.
• The initial run of the trained model gave an accuracy of 60% which was due to the lack of seasonality and trends in the data.
• We tried to identify other external data points which could impact the inventory pipeline. So, to improve the model accuracy further, we identified a new variable which was rolling pipeline data (90 days/60 days/30 days/same month).
• A new statistical model was developed to accommodate the new predictor variable (rolling pipeline data) along with the past data