Go-live weekend goes smoothly. The cutover finishes on time, the dashboards are green, and the project team sends the congratulations email. Then Monday arrives. Orders take longer to release. A routing is wrong on one line. Operators find a workaround, then another. By Thursday, the plant is behind — and nobody can quite say why.
This is the go-live dip: the drop in output that follows most major operational changes — a new ERP or MES, a line relocation, a new layout, a new way of scheduling. It is common enough that many organisations simply plan around it. The more useful question is not whether the dip will happen, but how deep it goes and how long it lasts — because those two numbers are largely within your control.
The dip is normal. Its size isn't.
The evidence points the same way across different kinds of change — from ERP systems to industrial AI:
The last figure is worth a second look, because it comes from a large study of US manufacturers by researchers working with the U.S. Census Bureau. Adopters did eventually outperform — but only after a period where productivity went down first. Economists call this the productivity J-curve. On a plant floor, we just call it go-live.
Why the floor absorbs the hit
It is tempting to treat the dip as a technology problem: the system was slow, the interfaces had bugs, the reports were wrong. Those things happen. But the research keeps pointing somewhere else.
In Panorama's 2026 survey, the most common reason projects ran over schedule was organisational issues — governance, resistance to change, process redesign — showing up as delayed sign-offs and extended stabilisation periods. Fewer than a quarter of organisations reported an intense focus on change management.
The MIT study found something even more specific. In older, established firms, the use of structured management practices — such as monitoring key performance indicators and production targets — actually declined after adopting AI. That decline alone accounted for nearly a third of their productivity losses.
The dip usually isn't caused by the new system. It's caused by the plant quietly letting go of the routines that kept it stable while everyone's attention is on the new system.
That is good news, because routines are something a plant can protect.
See the dip
Two go-lives of the same change on the same plant. One goes in unprepared. One goes in with the floor protected. Enter your own numbers — the shaded area is the output you don't get back.
Five ways to make the dip shallower and shorter
None of these require a bigger project budget. They require deciding, before go-live, that the plant floor is part of the project — not the place where the project lands.
| 1. Rehearse with real work | Walk operators and supervisors through real transactions, with real data, at the stations where they will use them. A demo in a conference room is not a rehearsal. |
| 2. Clean the master data early | Bills of materials, routings and item masters are a common breaking point at go-live. Have the people who actually use them validate them — not just the project team. |
| 3. Build a buffer | A modest build-ahead of finished goods or critical parts buys breathing room in the first weeks, when throughput wobbles and customers shouldn't notice. |
| 4. Keep daily management running | Huddles, KPI boards and escalation paths are exactly what catch go-live problems early. They are often the first thing paused "until things settle" — and the last thing that should be. |
| 5. Define "stable" before you start | Agree the numbers that end the support period — output, OEE, order accuracy back at baseline — so support leaves when the floor is ready, not when the calendar says so. |
Is your floor ready for go-live?
A quick read for anyone with a change coming up. Nothing tracked, nothing saved.
What good looks like
Every major change will cost the plant something while people learn a new way of working. That is not a failure — it is the investment. The difference between a good go-live and a painful one is rarely the software. It is whether the organisation kept its routines, its data and its people ready for the moment the change reached the floor.
Planning an ERP go-live, line move or layout change? Happy to compare notes.
Get in touch →Where these numbers come from
- Panorama Consulting Solutions, 2014 ERP Report press release — 51% material operational disruption at go-live; 60% of those lasting a month or longer (192 respondents, Jan 2013 – Feb 2014)
- Panorama Consulting Group, The 2026 ERP Report — over-budget and over-schedule findings, causes of overruns, change-management focus (170 respondents, Jan 2025 – Jan 2026)
- MIT Sloan, The "productivity paradox" of AI adoption in manufacturing firms, summarising McElheran, Yang, Kroff & Brynjolfsson, The Rise of Industrial AI in America (U.S. Census Bureau, CES-WP-25-27)
- Hero image: Sergey Sergeev via Pexels