Friday, 6 p.m. The cutover begins. Everyone’s optimistic, pizza is ordered, and the project manager is quietly terrified. This is the part of ERP Consulting nobody puts in the brochure.
Saturday, 2 a.m.: the first sign of trouble
Customer records loaded, but 4,100 of them had missing credit limits. That meant every order would stop at the credit check on Monday morning. Nobody had flagged it, because the field was optional in the old system and mandatory in the new one.
By 9 a.m. Saturday, they’d found three more of the same problem: two dozen supplier payment terms set to default, a warehouse location code that hadn’t transferred, and opening inventory balances that were off by $180,000.
What actually saved the Monday
They didn’t push the date. They froze two critical decisions instead: no new orders processed without manual credit approval, and only five order-entry staff worked Monday instead of eighteen. Slower, but nothing shipped wrong.
That company survived because they had three things ready before the weekend started:
- A written list of the 12 checks that had to pass before anyone could transact.
- Two people on call who knew the old data well enough to spot nonsense.
- A pre-agreed fallback: manual approval for anything that would otherwise block a customer.
The second go-live, eight months later
For the next site, they did something simple and unglamorous. They ran the entire cutover twice into a test environment, using a full copy of live data, three weeks before the real thing.
- Rehearsal one: load everything, log every failure, fix nothing. Takes about 14 hours.
- Two weeks of cleaning data based on what broke.
- Rehearsal two: same steps, same people, same clock. Target under 10 hours with zero blockers.
- Real cutover: Friday 6 p.m. start, Saturday 1 p.m. finish, Sunday off.
The second go-live wasn’t flawless. Three reports came out wrong and one integration ran two hours late. But Monday morning, 96% of orders processed without anyone touching them — versus a number nobody wants to admit from the first attempt.
The lesson isn’t “test more.” It’s that a rehearsal with real data finds the problems that meetings never will. Two dry runs cost three weeks of effort and saved a month of chaos.
If your go-live is more than six weeks out, book the first rehearsal today. If it’s closer than that, write your 12 checks and your fallback plan this week.