Every plant manager who’s lived through an ERP rollout has the same nightmare: go-live weekend, the new system fails on Monday morning, and the line stops. Days of lost production. Customer commitments missed. The CFO walking the floor with a calculator and a very specific look.
It doesn’t have to go that way. Here’s the playbook we’ve used for US manufacturers to ship ERP without halting production — phased, defensible, and reversible.
Why big-bang ERP go-lives fail
Big-bang means switching every site, every module, every user on the same weekend. It looks fast on a Gantt chart and fails in the real world for predictable reasons:
- Untested edge cases. The 50 work orders open on Friday afternoon don’t fit the new system’s data model. Nobody noticed in UAT.
- Operator panic. Day-shift on Monday meets a UI they saw once in training six weeks ago. Throughput crashes.
- No rollback path. Once the old system is shut down, you’re committed. Bugs become production crises.
- Master-data debt. Bad part numbers, missing BOMs, duplicate vendors — multiplied across every module simultaneously.
The phased approach that actually works
Phase 0: Master-data cleanup (4–8 weeks before go-live)
Don’t move bad data into a new system. Run a structured cleanup: deduplicate vendors and customers, retire dormant SKUs, verify BOM accuracy on top-100 active parts. Phase 0 is the cheapest risk reduction in the entire project.
Phase 1: Read-only shadow (2–4 weeks)
Bring up the new ERP in read-only mode, populated with live data synced from the old system. Power users explore the new UI on real data without touching production. Reports, dashboards, and key workflows get validated against known-good baselines.
Phase 2: Pilot site or pilot module (4–8 weeks)
Pick one site, one product line, or one module (often financial first, since accounting cutover is a known pattern). Run dual-entry or one-way sync between old and new for the duration. Operators on the pilot get the new system; everyone else continues as before.
Critical: have a written rollback procedure. If pilot data starts diverging from old-system reality, you can pause and fix without torching the rest of the business.
Phase 3: Site-by-site rollout (2–3 months per site)
After the pilot stabilizes, roll out to the next site. Then the next. Each site go-live looks like a controlled cutover with the previous site as the safety net — old system available for 30 days post-cutover, then decommissioned.
Phase 4: Decommission (after final site stabilizes)
Only after every site is on the new system and 60+ days have passed without rollback do you actually retire the old ERP. Premature decommissioning is the single biggest avoidable disaster.
Hyper-care: the 30 days that make or break go-live
- Embed implementation engineers on-site, not just on-call.
- Daily standups with operations + engineering for the first 14 days. Weekly thereafter.
- Issue triage with strict severity tiers. P0 (line down) gets a response in minutes; P3 (cosmetic) gets a backlog ticket.
- Clear escalation path to the CIO and the implementation partner principal — written, distributed, posted on the floor.
- A “war room” on site for the first 5 business days.
Cutover weekend: the actual hour-by-hour
- Friday end-of-shift: Stop new transactions in old system. Final reports printed and reconciled.
- Friday evening: Final data extract from old system. ETL into new system. Smoke tests on key workflows.
- Saturday: Reconciliation against pre-cutover balances. AP, AR, inventory, GL must tie. Any variance must be investigated and signed off.
- Sunday morning: Sandbox testing by power users on the new live system. Sign-off go/no-go decision by 2pm.
- Sunday evening: Communications to all operators. Quick-reference cards distributed. Helpline numbers posted.
- Monday 6am: Day-shift starts. War room active. On-site support per cell.
What to keep dual-running
For 30–60 days post-cutover:
- Old system in read-only mode for historical lookup
- Daily reconciliation report comparing old-system close balances to new-system equivalents
- Backup ERP environment ready to receive a snapshot if rollback becomes necessary
FAQ
How long does a phased ERP rollout take versus big-bang?
Big-bang sells as 6–9 months. Phased takes 9–15 months for the same scope. The trade-off: phased almost never causes line stoppages, while big-bang frequently does.
Doesn’t dual-running cost more?
Yes, modestly. Plan for 10–15% additional implementation cost for dual-run infrastructure and reconciliation labor. Compare that to the cost of a 3-day production halt and the math is obvious.
Can we do a phased rollout with off-the-shelf ERP, or only with custom?
Both. Most modern ERP vendors support phased site rollouts; you’ll need a competent implementation partner. Custom ERP gives you more control over the cutover sequence, especially when integrating with legacy MES.
Bottom line
ERP rollouts that halt production are almost always preventable. The phased playbook isn’t glamorous, but it works. If you’re facing an imminent ERP cutover and want a second set of eyes on the cutover plan, we run no-obligation reviews for US manufacturers every month.
