AI Menu & Sales Insights for ERPNext POS

ECOSIRE builds a dedicated Frappe app — a proper module with its own DocTypes, not a spreadsheet bolted on the side — that reads your POS history and produces actionable outputs inside ERPNext. We model demand with time-series methods on aggregated POS Invoice Item data (seasonality, day-part, weekday and promo effects), classify every menu or catalogue item into Star / Plough-horse / Puzzle / Dog quadrants using contribution margin against popularity, and run basket analysis (association rules over co-purchased items) to surface upsell and cross-sell prompts. Forecasts, classifications and pricing suggestions are persisted to custom DocTypes such as Demand Forecast, Menu Engineering Result and Pricing Suggestion, each permission-scoped via role profiles so a shift manager and a finance controller see the right slice.

Technically, the app is wired into ERPNext the way Frappe intends. Nightly scheduler_events (cron/daily hooks in hooks.py) recompute forecasts and menu classes off the latest closed POS sessions; doc_events on POS Invoice submit keep rolling aggregates warm so recommendations stay fresh intra-day. Heavy model runs are pushed to the background job queue (frappe.enqueue) so the desk never blocks. Whitelisted server methods (@frappe.whitelist()) expose forecasts and upsell suggestions over the Frappe REST API, and a lightweight client script injects the top upsell prompt and prep-quantity hint into the POS and item screens the cashier already uses. Everything respects your existing permissions, company/outlet dimensions and multi-currency setup — nothing is scraped or exported to a black box.

Because every menu, margin structure and outlet mix is different, this is a build-to-order engagement, not a marketplace download. After a short scoping call we confirm your ERPNext version (v15 or v16), your POS data volume, the outlets and item groups in scope, and which decisions you want automated first. We build the app against a copy of your data, run it on a staging bench for UAT, and only then install on production with a documented rollback. Typical delivery is 2 to 4 weeks from confirmed scope. You receive the full source, a git repository handover, training for your team, and a post-go-live support window — you own the app outright and can extend it yourself or have us maintain it.