Manufacturing System Integration Checklist

Document which modules are in scope (work orders, BOM, routing, inventory, labor, OEE) and which directions data moves between ERP, MES, CMMS, and PLM. Call out master-of-record for each entity — part master in PLM, inventory in ERP, machine state in MES — to avoid duplicate-source-of-truth problems after go-live.

Name the cutover lead, IT lead, plant manager sponsor, quality engineer, materials lead, and a floor super-user per shift. Steerco approves the go/no-go on cutover weekend; super-users are the hypercare frontline on Monday morning.

Schedule cutover during a low-volume window — typically a Friday-night-to-Sunday weekend or planned shutdown. Build the rollback runbook in parallel: if go/no-go is No-Go at hour T, what gets reverted, by whom, by when, and how production resumes Monday on the legacy stack.

Score severity, occurrence, and detection for each failure mode — duplicate work orders, stale BOM rev synced to MES, lost lot/heat traceability, machine feed disconnect, kanban replenishment broken. Anything with RPN above your threshold needs a documented mitigation before cutover.

Confirm each endpoint version (ERP REST/SOAP, MES connector, MTConnect agent, OPC UA server) against the integration spec. Check authentication method, rate limits, and payload schema. Vendor patch level should match what was tested in sandbox — surprise upgrades on production are a common rollback trigger.

Pull a quality report on active part numbers: missing UoM, blank routing, obsolete revs still flagged active, duplicate part numbers across plants. Engineering owns BOM cleanup; production owns routing cleanup. Stale data migrated as-is becomes everyone's problem on day one.

Map source-to-target field by field for lot, heat, serial, and shelf-life attributes — these are non-negotiable for AS9100, ISO 13485, and FSMA traceability audits. Confirm UoM conversions for any item stocked in one unit and consumed in another (sheet stock, bar stock, bulk resin).

Take full backups of every system in the integration, verify the restore on a non-production host, and document the snapshot timestamp. The backup IS the rollback plan if cutover fails — an untested backup is worth nothing.

Patch sandbox to match the target production state, then re-run integration smoke tests. Do not patch production within 7 days of cutover — surprise vendor regressions discovered on go-live weekend are a known rollback trigger.

Pick one cell or work center — ideally a low-mix, high-volume one — and run a full end-to-end migration: release a work order in ERP, route to MES, capture machine state, complete inspection, close the WO. Reconcile counts, scrap, and lot data against the legacy system.

Triage every pilot defect to root cause — data mapping, API version, business-logic mismatch, or user-procedure gap. Re-run the pilot until pass before scheduling the full cutover. Cutover with a failed pilot is the most expensive way to learn what was wrong.

At cutover-window start, stop new WO releases, complete in-process WOs to a clean stopping point, and lock the legacy system to read-only. Communicate the freeze to production planners and shipping at least 48 hours in advance.

Run the steps in order, check off each in real time, and log timestamps. Do not improvise — a step skipped at 2 AM is the step that causes the 6 AM incident. If the runbook needs to deviate, the cutover lead approves the deviation and logs it.

Walk the floor or use the OEE dashboard to confirm every monitored asset is reporting state, cycle, and alarm data. CNC, PLC, and tester feeds all need verification — silent disconnects are a common post-cutover surprise that distorts OEE for weeks if missed.

Steerco reviews the runbook log, machine-feed verification, and any open incidents. Default is No-Go unless every Sev-1 risk is closed. The decision, the steerco signature, and reviewer notes all get captured here for the audit trail.

Super-users on every shift for the first 5 production days, with a dedicated incident channel and a 15-minute response SLA for blockers. Track every incident in a hypercare log so the team can pattern-match quickly when the same issue surfaces on second and third shift.

Execute the rollback runbook: restore database snapshots, re-enable the legacy MES connectors, unfreeze legacy WO transactions, and notify production planning that Monday morning runs on the prior stack. Capture the failure root cause for the next cutover attempt.

Treat the first work order through the new stack like a setup FAI — quality verifies every critical dimension and confirms the inspection record makes it back to the QMS. A clean first piece on the new system is the most credible signal that the cutover took.

Pick a finished good shipped post-cutover and trace backward: FG lot to WO, WO to component lots, component lots to incoming receipts and supplier COCs. A break here means the next ISO or AS9100 surveillance audit will surface it.

Compare availability, performance, and quality from the integrated MES feed against the prior week's baseline from manual logs or the legacy MachineMetrics dashboard. Significant unexplained drift usually means a state-mapping bug — alarm code categorized as planned downtime, or vice versa.

Short structured interviews with operators on each shift and the production scheduler. Ask about workflow friction, missing data on the traveler, and any workarounds they've invented. Workarounds are leading indicators of the next CAR — feed them back into the backlog.

Quality and IT conduct a layered audit covering data integrity, transaction completeness, traceability, and user adoption. Outputs feed the change-control backlog and update the PFMEA so lessons learned land in the next integration.