People think software goes live on a date. In my experience the date is the easy part. What decides whether a system sticks is the thirty days around it, and most of the work happens before anyone logs in. Here's roughly how a rollout goes when it works, from someone who does this for a living.
What happens in the first 30 days of an ERP rollout?
Week one: I mostly watch
I don't bring anything to install in the first week. I bring a notebook. I follow one job from order to dispatch and write down every register, every WhatsApp group, every "go ask Suresh" in the chain. You can't replace a workflow you haven't seen with your own eyes, and the floor's real process is never quite the one drawn on the office wall.
Week two: clean the data nobody wants to clean
If you are the one taking this on rather than the one who built the place, the first two years have their own shape, which we have set out in taking over the family manufacturing business.
This is the unglamorous heart of the job. Item codes that mean three different things, BOMs that stopped matching reality years ago, one customer spelled four ways. We fix this before go-live, never after, because a system fed bad data produces reports nobody believes, and a report nobody believes kills adoption faster than any bug ever could. It's the most common reason a first ERP attempt quietly dies.
Week three: one cell, not the whole floor
We go live on a single line or department, never the entire plant on one Monday morning. The operators there get fifteen minutes on the two or three screens they'll genuinely use, and nothing more. Keeping that first version tiny is exactly what makes it stick. The same one-step-at-a-time approach works whether you're moving off paper job cards or rolling out a full system.
"The trick wasn't the software. It was that they fixed our item master first, so the stock report finally matched the racks."Production supervisor at an auto-components unit, Faridabad
For a real floor working through exactly this curve, with the daily user count published week by week, see how Karm Machine Tools got 58 shop floor workers onto the system.
Week four: I make myself unnecessary
By the last week, my job is to hand it over. I make sure someone on the floor owns the system, that the daily numbers get read in the morning meeting, and that the questions reaching me have slowed to a trickle. If I'm still the only person who can run it after a month, then I've done the job badly.
The mistakes I watch owners make
Buying for a long feature list instead of daily use. Going live everywhere at once. Skipping the data cleanup to save a week, then losing three months to broken trust. Treating go-live as the finish line when it's really the start. Every one of these is avoidable, and every one comes down to how you roll the system out far more than what you bought.
