The whole plant couldn't go live at once, so we started with the hardest warehouse

We couldn’t hold the date for taking the whole plant live at once.

So we split the plant by warehouse area and went live in batches. The first batch was the biggest, most complex automated high-bay warehouse. That one warehouse handles about 60% of the material used in daily production.

The usual move is to test the water with a small warehouse first. We did the opposite.

I was the project manager on that project. This post covers three things: why the first batch went to the hardest warehouse, how the small high-bay warehouses got through 2 months on manual paperwork, and how a phased go-live should be written into the project plan.

1. Taking the whole plant live at once was too big a leap

The original plan was for every warehouse area to go live together. The closer we got to the date, the clearer it became: the system’s features weren’t complete, the people weren’t in place, and the equipment couldn’t be commissioned in one pass.

In the end, the system had never been broken in and proven on a real floor. People, features, equipment: we couldn’t be sure of any of them. And the floor could not afford an accident that stopped production.

More important still, the old plant had to keep hitting its output every day. We couldn’t bet all of production on one new automated line.

Going from manual work to automation meant several new systems had to work together at once. The operators had no experience with these systems or this equipment. If the whole plant ran into trouble together, output targets would be missed and the losses would be large.

So when we switched to batches, nobody objected. It was the safest option, and the most realistic one.

2. The hardest high-bay warehouse went first, because it looked most like real production

The big high-bay warehouse was the most representative. It matched day-to-day production needs most closely, and it was also the most complex.

The small warehouses handle low volumes, so we could stock extra material there as a buffer. The big one handles high volumes and can’t afford mistakes. One error there spreads a long way.

However smoothly a small warehouse goes live, it proves nothing about whether the big one can run. Once the big one runs smoothly, the hardest stretch is behind you.

We didn’t release the full order load at once, either. First 2 units, then 4, then 6, with no pause in between, adding production orders batch by batch.

The hardest part was material that went into the high-bay racks and wouldn’t come out.

The cause, again, was bad data. At the time, the only option was to switch the equipment to manual mode and get the material out by hand. Meanwhile we fixed both ends at once: improving the data coming from upstream, and correcting the data already in the system.

3. The small warehouses ran on paper for 2 months, and the records were cleaned up round by round

There were also 3 small high-bay warehouses. While the big one was going live, they stayed manual: every receipt and every issue was a handwritten document, for about 2 months.

One plant, half in the system and half on paper. Manual paperwork always brings some delay, and people make mistakes. Some records were bound not to match. We got them clean only by going over them again, round after round.

To judge whether the big warehouse was stable, we watched three things: the equipment never stopped; receipts and issues were both accurate; and we no longer had to go back and reconcile the system data with the pallets on the floor. The observation period stretched from a week to two weeks, then to several months.

Only then did the small warehouses switch over to the system. That went much more smoothly, though some things still needed custom tuning, depending on the properties of the materials.

This time, we split into batches only because the date couldn’t hold.

If I did it again, I’d write it into the project plan from the start: go live by warehouse area, which one first, how far it has to get before the next one starts, and how the small warehouses keep their records in between. Then push to that plan from day one.

Would the client accept a plan written that way? I believe so. You just have to give them a final completion date for the whole plant, and the pace to get there.

What clients fear is often not the phasing. It’s not knowing when everything will be live.

Some go-live dates should never have been set as “the whole plant at once.”

This post goes deeper into the acceptance post from the previous series: that one said to start with one line or one warehouse; this one is about which one to start with, and how to get through the stretch in between.


When your project went live in phases, did the first batch go to the easiest area, or the most representative one?

Related notes

Published when written, never an ad: subscribe via RSS