
The Challenge.
- Demand forecasting is one of the most important problems in retail. An accurate forecast can improve planning and reduce inventory costs.
- Often, companies use high-level demand forecasts based on trends from the previous year.
- These methods do not forecast at the product/store level, and often do not account for important factors such as seasonality, promotions, and weather.

The Solution.
- Built statistical models to forecast sales by product and store up to 24 months in the future.
- Used sales of similar products to forecast sales of new products with insufficient historical data.
- Included features promotions and weather data into the model for added accuracy.

The Results.
- Product-level forecasts lead to a major reduction in out-of-stock.
- Significant reduction of total inventory.
- Improved the buying process by enabling data-driven demand planning and budgeting.