Food, chemical, cosmetics, and nutraceutical plants don't build products by assembling discrete parts — they mix, react, cook, and blend ingredients where the output quantity is never exactly what the math predicted. ERPNext's native BOM and Work Order model was designed for discrete assembly: a fixed BOM of components produces a fixed quantity of one finished good. It has no first-class concept of a scalable formula, no way to record that one batch also produced a saleable by-product and a waste stream, no theoretical-versus-actual yield reconciliation, and no potency adjustment when an incoming raw material assays at 92% instead of the label 100%. Process teams end up running spreadsheets alongside ERPNext, and the two drift apart.

This build-to-order app closes that gap as a proper Frappe application installed on your bench (bench get-app / bench install-app), namespaced so it survives ERPNext upgrades. We introduce process DocTypes — a Recipe/Formula master with percentage- or quantity-based line items, a Batch Production Order that scales a recipe to a target batch size, and Yield Reconciliation records that capture actual outputs against theoretical. Co-products and by-products are modeled as distinct output lines with configurable cost-allocation methods (by weight, by market value, or fixed ratio), so ERPNext's stock ledger and valuation stay correct instead of dumping everything onto a single finished good. Potency/concentration handling reads an assay value on the incoming Batch (a Batch DocType custom field) and adjusts the charged quantity of that ingredient at Work Order creation so the active-ingredient target is met.

Technically it's wired the way a maintainable Frappe app should be: hooks.py document events (validate, on_submit, on_cancel) enforce the yield and allocation logic server-side rather than in fragile UI scripts; Server Scripts and whitelisted Python methods (@frappe.whitelist()) expose recipe-scaling and reconciliation to both the desk and the Frappe REST API (/api/method/...) for line/plant integrations; Client Scripts drive the batch-sizing and live yield-variance feedback in the Work Order form. Batch genealogy is built on ERPNext's existing Batch and Serial/Batch Bundle plumbing, extended with a traceability report that walks forward (where did this input batch end up?) and backward (what raw batches went into this finished lot?) — the foundation of a recall in a regulated plant. Role Profiles and permission rules keep formula IP visible only to the roles you authorize, and scheduler events can push nightly yield-variance and expiring-batch digests.

Because this is build-to-order, nothing ships until we agree on scope. A short scoping call maps your recipes, allocation rules, and traceability needs; we then build against ERPNext v15/v16 on a staging bench, run UAT with your real formulas, and hand over installable source plus a git repository. Typical delivery is one working week from confirmed scope, and every engagement includes a post-go-live support window and a documented rollback plan.