GET /mes

The plant made millions of pounds of product a week — and the books found out later. Contemporaneous production tracking sat around 20%: in-plant production often wasn't recorded in the ERP as it happened, so inventory, costing, and lot traceability ran on end-of-day catch-up and educated guesswork. The people who could have typed it all in were busy doing something more important: running a plant.

The goal wasn't a better data-entry screen. It was no data-entry screen — records that create themselves as the physical work happens.

Stage 1 — listen to what the plant already says. The equipment was already talking: sensors, flows, and weights streaming into a supervisory control and data acquisition (SCADA) historian (AVEVA). The MES service reads that real-time process data directly — no new hardware on the line, no operator input, no parallel record-keeping. The physical signals are the source of truth, so the system starts from them.

Stage 2 — turn signals into batches. This is the hard part. A historian hands you time-series soup; it does not hand you a batch. The service reconstructs production batches at every stage of the line from those signals — what ran, when it started and stopped, what it consumed, and what it produced — and ties every pound to a lot. Real production is messy, and this layer is built to absorb that mess rather than pass it downstream.

Stage 3 — post it like an accountant would. Each reconstructed event becomes the matching NetSuite transaction, automatically, over secure OAuth 1.0a REST integrations: item receipts, assembly unbuilds, work order create/issue/complete, waste adjustments, and automated backflush consumption. And because no integration is trusted to be perfect, reconciliation sweeps continuously recheck the record and self-heal drift — the ERP converges on physical truth instead of slowly wandering away from it.