MES vs ERP: Where the Boundary Really Lives in a Regulated Plant

ERP — Enterprise Resource Planning — is the business backbone. It owns the chart of accounts, the item master, the bill of materials at financial granularity, customer and supplier records, sales orders, purchase orders, AR and AP, MRP planning runs, and the inventory ledger that reconciles back to the balance sheet. SAP S/4HANA, Oracle Cloud ERP, NetSuite, Microsoft Dynamics 365 F&O, Infor LN, IFS, and the mid-market QuickBooks Enterprise / Sage X3 all live in this layer. ERP knows that work order WO-24-0918 exists, what it consumes (the planned BOM), what it should yield (the planned output), what cost roll-up applies, and when the finished goods need to ship. What ERP does not do well — and was never designed to do — is real-time shop-floor execution at second-by-second granularity, native ALCOA+ data capture for GxP records, kiosk-grade operator UX in a glove-and-gown environment, calibration tracking on the sensor that recorded a critical weight, deviation workflows tied to a specific test step, or the audit-trail depth that a 21 CFR Part 11 inspector expects. Asking ERP to do those jobs is the classic mistake — it's like asking the accounting system to also be the security camera.

MES — Manufacturing Execution System — is the shop-floor system of record. ISA-95 places it at Level 3, sitting between ERP (Level 4, business planning) and the control systems on the line (Levels 1–2, PLCs and SCADA). MES dispatches work orders to specific lines and operators, renders electronic work instructions step-by-step, enforces required inputs (weights from a connected scale, scans of lot numbers, e-signatures at gates), captures every action contemporaneously to the audit trail, manages deviations and out-of-spec events, tracks equipment status and OEE in real time, and produces the signed batch record (eBR / eDHR) that QA releases against. V5 Ultimate, Werum PAS-X, Körber's MES suite, Rockwell PharmaSuite, Honeywell's manufacturing platforms, and Apprentice all live in this layer. What MES does not do — and should not do — is general-ledger accounting, customer master data, sales-order processing, MRP planning, or financial inventory valuation. MES tells ERP the work was done and the materials were consumed; ERP turns that into the financial posting. The MES is right when it knows what happened on the line to the second; the ERP is right when it knows what the company owes and owns to the cent.

Most MES-vs-ERP arguments collapse to seven design decisions: (1) Where does the work order originate — planned in ERP and dispatched to MES, or created standalone in MES for unplanned work? Both are valid; pick one as primary. (2) Who owns the BOM — ERP's financial BOM (cost roll-up) or MES's execution BOM (step-by-step recipe with in-process controls)? Usually both, with MES expanding the ERP parent. (3) Where do lots get created — at goods receipt in ERP, or at first weighing in MES? Pharma typically does both; food typically creates lots only at MES. (4) Where does inventory move — every transaction reflected in ERP immediately, or batched at end-of-shift? Real-time is cleaner; batched is simpler. (5) Where does the device history / batch record live — MES is non-negotiable for GxP. (6) Where do deviations live — MES for execution-time deviations, QMS for post-release deviations and CAPAs. (7) Who is the system of record for finished-goods cost — ERP, always; MES feeds the actual quantities. Get these seven right up front and the integration is straightforward; ambiguity in any of them produces dual entry, reconciliation pain, and audit findings.

The integration pattern matters more than the vendor choice. Avoid file-based nightly batch jobs — they create reconciliation gaps that data-integrity inspectors flag instantly. Prefer event-driven messaging: MES emits domain events (work-order-started, material-consumed, batch-released, finished-goods-produced) over a durable queue or webhook stream; ERP subscribes and posts the corresponding GL/inventory transactions. For the reverse direction — ERP-to-MES — pull the master data (items, BOMs, work orders, customers) over a stable API on a short polling cadence, with ETag / If-Modified-Since headers so MES caches what hasn't changed. Critically, every cross-system message must carry an idempotency key so retries don't double-post. Wrap the integration in a thin anti-corruption layer on each side so an SAP upgrade or a NetSuite reconfig doesn't ripple into the MES schema. Sign payloads with HMAC so an audit can prove what MES sent and what ERP received. This is the pattern that survives the inevitable ERP upgrade and the inevitable MES major version — and it's the one most legacy MES-ERP integrations got wrong, which is why so many manufacturers are mid-re-platform right now.

Several ERP vendors ship a 'shop-floor' module: SAP Digital Manufacturing, Oracle Manufacturing Cloud, NetSuite Advanced Manufacturing, Dynamics 365 Production Control. For low-mix discrete assembly without GxP requirements — packaging non-regulated consumer goods, light industrial assembly, kitting — these modules are often sufficient and avoid a second system. They start to break the moment you need (a) signed electronic batch records under 21 CFR Part 11 or EU GMP Annex 11, (b) granular in-process controls with kiosk capture and bound e-signatures, (c) deviation workflows that block the next step until QA dispositions, (d) real-time OEE per line, per shift, per SKU, (e) calibration-aware sensor capture, or (f) operator UX designed for gloved hands and gown-up environments. Pharma, biotech, medical devices, dietary supplements, regulated food (FSMA 204), and cosmetics under ISO 22716 effectively require a dedicated MES once the plant has more than one line or one product. The honest test: print the last quarter's batch records and ask QA whether they would release them to an FDA inspector. If the answer involves Excel reconciliation, the ERP module is not enough.

Five failure modes recur across every MES-vs-ERP project. (1) Buying MES from your ERP vendor as a 'native module' without verifying it has separate certification under Part 11 / Annex 11 — the vendor's slide does not equal the inspector's checklist. (2) Letting the BOM live in two places without a single source of truth — execution drifts from financial within months. (3) Trying to do real-time inventory in ERP from MES events without an idempotency contract — duplicate postings during a planned outage corrupt the ledger. (4) Underestimating master-data governance: who can change a recipe, who approves a routing change, who retires an obsolete item? Without an MDM discipline, every ERP-to-MES sync becomes a debugging session. (5) Treating the project as a software install rather than an operating-model change — the floor practices, the QA review SOPs, and the IT change-control discipline all have to evolve. The plants that win at MES + ERP run the integration project as a 12-month operating-model programme with software as one of seven workstreams, not a six-month IT implementation with change management bolted on at the end.

V5 Ultimate is a Level-3 MES purpose-built for regulated manufacturers. It does not try to be ERP — there's no general ledger, no AR/AP, no MRP. It integrates with whatever ERP you already run (NetSuite, SAP, Dynamics, QuickBooks Enterprise, Sage) through signed REST + webhooks with idempotency keys built in, and it ships pre-configured for the seven boundary decisions above so the integration scoping conversation takes hours, not months. On the floor it owns work-order dispatch, kiosk-rendered electronic work instructions, bound e-signatures under Part 11, audit-trail hashed and exportable, deviation workflows, OEE in real time, eBR / eDHR generation, calibration-aware capture, and ALCOA+ compliance natively. The result: one system of record on the floor, one system of record for the business, and a documented contract between them that an FDA or notified-body inspector can read end-to-end without an Excel reconciliation in sight.