Software Implementation Checklist

State explicitly which systems are in scope, which plants and work centers are covered, and what's deferred to Phase 2. Scope creep — bolting QMS onto an ERP rollout mid-flight — is the most common reason these projects miss go-live by 3+ months.

Name a project sponsor (typically Plant Manager or COO), a project lead, and core members from IT, manufacturing engineering, quality, maintenance, materials, and finance. Include at least one shop floor operator or shift lead — implementations that skip the floor voice tend to ship workflows operators won't use.

Walk the floor and document how work orders, travelers, kit pulls, FAI, in-process inspection, NCRs, and shift handover happen today. Capture the workarounds — the tribal knowledge that keeps production running but won't be in any SOP. These are exactly what the new system either accommodates or breaks.

Budget includes license, implementation services, internal labor (often the largest line, frequently underestimated), hardware (shop floor terminals, scanners, printers), and a 15–20% contingency. Anchor go-live to a low-volume window — never to month-end or fiscal close.

Get the sponsor's signature on scope, budget, and go-live target before vendor RFPs go out. Without a sponsor on record, scope and budget drift have nobody to push back on.

Limit to 4–6 vendors with proven fit for your segment — Plex, Epicor Kinetic, NetSuite, Tulip, Aegis FactoryLogix, ProShop for discrete; SAP Business One or Acumatica for mid-market; specialized stacks for regulated work (Greenlight Guru for medical, MasterControl for cGMP). Generic horizontal ERPs without manufacturing modules rarely survive the demo.

Include concrete requirements: BOM and routing structure, lot/heat traceability depth, FAI and PPAP support, SPC and gauge R&R hooks, PLC/MTConnect integration, AS9100 or IATF 16949 compliance if applicable. Vague RFPs come back with vague responses.

Send each finalist 2–3 of your real BOMs and routings, plus a representative work order scenario (changeover, in-process scrap, NCR with disposition). Demos that don't use your data are sales theater.

Talk to two references at companies of similar size, similar process complexity, and ideally on the same module set. Ask about implementation overrun (calendar and budget), ongoing support quality, and what they would do differently.

Lock in named-user vs. concurrent licensing, hosting model (on-prem vs. SaaS), data residency, SLA with response and resolution times, and exit/data-extract terms. The exit clause is the one nobody reads and the one that hurts on year five.

Decide multi-level vs. flattened BOM, phantom assemblies, and how routings handle alternate operations and sub-contract steps. Get this wrong and every work order downstream inherits the error.

Define work centers at the right granularity — too fine and scheduling thrashes, too coarse and capacity reporting is useless. Set standard cycle and setup times from time studies, not tribal estimates.

Wire up machine signals (run/idle/down, cycle counts, alarm codes) for OEE calculation. Older PLCs may need an adapter (MachineMetrics, Ignition by Inductive Automation). Validate that idle-vs-down classification matches how your supervisors actually code downtime.

Define roles per shop floor function: operator, setup operator, shift lead, quality inspector, planner, supervisor. Operators should not have access to standard cost or pricing screens. Validate against your IT segregation-of-duties policy.

Document every config decision (BOM structure rules, work center definitions, role permissions) in a change-controlled repository. For ISO 9001, AS9100, or IATF 16949 plants, the auditor will ask for this on the first surveillance audit.

Use real production scenarios: a normal work order, a changeover with FAI, a hot job that jumps the queue, an NCR with rework disposition, a kit short that triggers a buy. Synthetic test cases miss the edges where systems break.

Have actual operators run UAT on the actual terminals on the actual floor — not the IT lab. Glove-friendly screens, barcode scanners, and shop-floor lighting reveal usability issues that polished demos hide.

Reconcile counts between legacy and new system to the part level — not aggregate. A 99.5% match across 20,000 parts is 100 wrong parts, and the wrong ones will be the high-value ones. Spot-check lot/heat traceability chains end-to-end.

Run end-to-end flows: PM work order from CMMS triggers downtime in MES; NCR opened in QMS holds the lot in inventory; shipping pulls the right BOL data and serial numbers. Integrations are where most go-live defects hide.

If the plant produces FDA-regulated product (medical device 21 CFR 820, drug cGMP, food FSMA) or operates under IATF 16949 with customer validation requirements, the system needs formal IQ/OQ/PQ validation. Skipping it because UAT passed is a finding waiting to happen.

Run Installation Qualification (hardware/software per spec), Operational Qualification (functions per requirements), and Performance Qualification (process performs in production conditions). Sign each protocol; archive in the validation master file. Required for FDA and most aerospace/automotive regulated environments.

Both Quality Manager and Operations Manager sign off that test results meet acceptance criteria before go-live is scheduled. No verbal approvals — this is the artifact the post-mortem will look for if go-live fails.

Build a training track per role — operator, setup, quality inspector, planner, supervisor — not a single 4-hour everything-deck. Job-aid laminated cards at each work center beat any LMS module on the floor.

Identify a trainer per shift (1st, 2nd, 3rd, weekend) before training. Train-the-trainer sessions should cover not just system mechanics but how to handle pushback — operators who think the new system is slower than the old one.

Run live production through one cell for 5–10 working days before plant-wide go-live. Track defect rate, transaction completeness, and operator-reported friction. The pilot is your last cheap chance to find problems.

Triage pilot issues by severity: blockers must close before plant-wide go-live; major issues need a workaround documented; minor issues go in the Phase-2 backlog. Do not advance the cutover date with open blockers — that's how go-live becomes a 6-week emergency.

Hold pre-shift meetings on every shift covering cutover timing, who to call when something breaks, and the fallback plan. Operators on 3rd shift hearing about go-live the day-of is a common avoidable failure.

Designate 2–3 super-users per shift with deep system knowledge and a direct line to IT. They are the floor's first line of support during hypercare and the long-term knowledge holders after the implementation team disbands.

Document the cutover sequence (final data load, legacy freeze, new-system enable) hour-by-hour with named owners and rollback triggers. Without a defined fallback, every minor go-live problem becomes an existential debate.

Have the implementation lead, sponsor, IT, and super-users on-site (or on-call) for the first full shift. Watch first work-order release, first FAI, first NCR, first shipment. If any of those break, intervene before the floor builds workarounds.

Hold a 15-minute defect standup every morning for the first 2 weeks post-go-live. Triage by impact on production, ship critical fixes same-day. Hypercare ends when defect inflow drops below a defined threshold — not on a calendar date.

Compare actuals against the project objectives: transaction completeness (work orders closed in system, not on paper), OEE data flowing from PLCs, scrap and NCR entry rates, on-time data from operators vs. supervisors entering after-the-fact. Gaps here predict where the system erodes within 6 months.

Hold a structured retrospective with the implementation team and floor super-users. Capture what worked, what hurt, and which deferred items move into Phase 2 (additional plants, additional modules, deeper integrations). Without a documented Phase-2 roadmap, deferred scope dies in someone's inbox.