I wanted to share our experience implementing automated demand forecasting with improved data modeling in Epicor SCM’s supply planning module. We were struggling with forecast accuracy and frequent stockouts, especially for seasonal products across multiple distribution centers.
Our previous forecasting approach used simple moving averages and didn’t account for location-specific demand patterns or seasonal variations. We redesigned our demand data model to capture time-series data at the location level, integrated automated forecasting algorithms, and the results have been impressive - stockouts decreased by 40% while reducing excess inventory by 25%.
The key was restructuring how we model and store demand history to enable more sophisticated forecasting techniques. Happy to discuss the technical details of our data model changes and the integration with Epicor’s supply planning engine.
Impressive results. What forecasting algorithms did you integrate? Are you using Epicor’s built-in forecasting engine, or did you integrate external tools? We’ve been evaluating machine learning approaches but haven’t found a good way to integrate them with Epicor’s supply planning workflows.
This is exactly what we need to implement. Can you share more details about how you structured the time-series demand data? We’re currently storing demand history in a flat table with monthly aggregates, but I suspect we need more granularity for effective forecasting. What time periods did you use, and how far back does your history go?
We use a hybrid approach. Epicor’s native forecasting handles standard products with stable demand patterns. For seasonal and promotional items, we integrated Python-based forecasting models (Prophet and ARIMA) that consume our time-series data and push forecasts back into Epicor via API. The location-based modeling was crucial - we can now generate forecasts at the DC level and aggregate up, rather than forecasting at the enterprise level and allocating down.