Most manufacturers and distributors running ERPNext hit the same wall: the platform plans supply beautifully once you tell it what demand will be, but it has almost no opinion about the demand itself. ERPNext's Production Plan pulls from open Sales Orders and Material Requests, and the reorder engine relies on static reorder_level / reorder_qty values you type into the Item master by hand. There is no moving average, no exponential smoothing, no seasonality, no safety stock derived from demand variability and lead-time risk, and no feedback loop telling you whether last month's plan was any good. Planners end up maintaining forecasts in spreadsheets and re-keying reorder points quarterly, which is exactly where stockouts and dead inventory come from.

This is a proper Frappe app (demand_forecasting_mrp) that ECOSIRE builds for your instance, not a theme or a set of custom fields bolted onto stock ERPNext. It ships a Demand Forecast DocType and per-period Forecast Item child rows, a Forecast Model Config DocType where you choose the method (moving average, single/double exponential smoothing / Holt, or seasonal Holt-Winters) and its parameters per item or item group, and a Reorder Policy DocType that computes safety stock and reorder point from demand standard deviation, lead time, and a target service level (z-score) rather than a fixed number. A scheduler event registered in hooks.py recomputes forecasts on your chosen cadence (nightly or weekly) by reading historical consumption from Stock Ledger Entry and Sales Invoice Item, so the numbers stay current without anyone touching a spreadsheet.

The output is wired back into ERPNext's own planning, not a parallel report. Whitelisted server-side methods generate a draft Production Plan (and, where relevant, Material Requests and draft Purchase Orders) directly from the approved forecast, and a hooks.py doc_events handler keeps each Item's reorder_level and reorder_qty in sync with the latest computed policy so the native reorder engine fires on optimized thresholds. An ABC/XYZ classification job segments items by revenue contribution (ABC) and demand variability (XYZ) so planners spend attention where it matters, and every forecast cycle stores forecast-vs-actual so a dashboard can track MAPE and bias over time. All of it respects Frappe permissions — a dedicated "Demand Planner" Role and Role Profile, DocType-level and (where needed) User Permission scoping so planners see only their responsibility area — and every number is reachable through the standard Frappe REST API for BI tools or an external planning workbench.

Because this is build-to-order, nothing is pre-listed for instant download. We start with a short scoping call, confirm which forecasting methods, service levels, and MRP hand-off behavior fit your operation, and then build against that signed-off scope. Typical delivery is 2-4 weeks from confirmed scope. You get the installable app source for your version of Frappe/ERPNext (v15 or v16), installation and configuration on your instance, UAT on a staging bench with a documented rollback plan, training for your planners, and a post-go-live support window. The git repository is handed to you at the end, so you own the code.