We built a forward-looking model focused on weekly/monthly sales forecasting, the following modules were implemented:
- EDA - Understand data to identify the right techniques to build forecasts.
- Benchmarking - Create baseline accuracies that the models have to beat.
- Modeling - Use different time series models.
- Confidence - Use multiple metrics to understand the model's coverage of the data.
- 95% accuracy in forecasting with a 10% variation from the predictive forecast number.
- Ability to find the trend and pattern of sales.
- Capability to plan for production and capacity.