A carve-out with a deadline. And four months already gone.

Four months into an eighteen-month deadline, the work carrying the larger share of the acquired revenue had nobody leading it in the US. I took it on.

What I'd been told was extremely optimistic. Four months after close, roughly $1M spent, nothing moved toward the exit. No discovery on either business. Nobody given a role, a goal, or a deliverable. Nobody could say what these plants were doing — so nobody could say what they should do next.

I asked for a US team, and said plainly why. A single effort run from another continent could not deliver the side holding most of the revenue. I built that team and had the business fund it with its own people — twelve full-time from operations, sales, and marketing, a named owner for every function, and a direct line from this program to the group CIO, the group CFO, and the US CEO. By the end, seventy-seven people were working this program: around forty subject-matter experts (SMEs) carrying it alongside their day jobs, six business owners (BOs), thirty dedicated to it full time — twelve process owners (BPOs) from the business and outside expertise for the rest, with me leading the program.

I set the sequence, and resourcing came before design. Approval in hand, the team came first — internal people named, outside expertise hired — and held governance and the work-stream structure back until they were in the room. That wasn't caution. Governance and the work-stream structure had to be built by the people who would live inside them, which meant those people had to be in place first. No decision on this program was made in a vacuum — though plenty were still made by one person.

Then the scope changed, five days after the plan was finished. The plan had been a lift-and-shift: move the applications, change nothing, exit clean. Group leadership decided instead to put every entity on one ERP, with the same deadline. Overnight it became a Tier-1 ERP build from scratch — a new Infor M3 instance and seven surrounding systems, five plants, four of them on a system first built in 1978. The five had been separate companies once, acquired one at a time, and still ran as their own business lines — different products, different processes. Minimum viable product was the guiding principle and we held it: every change request had to prove the standard product could not do the job before it reached the review board.

Change management was fused into the structure, not bolted on at the end. It was a guiding principle for every work stream, it sat on every agenda as a standing requirement rather than an occasional topic, and its actions were tracked and closed like any other deliverable. Nothing on the program got watched more closely.

Every subject-matter expert (SME) was named at the start, and the business did the build — its own people deciding on their own processes. Who we picked mattered as much as how many: people already respected in their departments, because an opinion that carries weight travels further than an announcement. Every business process owner (BPO) then ran a meeting with their own team every week — to bring them into a decision, or at minimum to tell them what had been decided. Nothing at go-live was a surprise to anyone.

The SMEs ran user acceptance testing (UAT), not just training. Testing one process against scenario after scenario builds muscle memory before anyone calls it training. A training PM turned what they were doing into standard operating procedures and work instructions as they did it, so anything undocumented got documented during testing rather than after. At go-live everyone else was trained by an SME, handed that documentation, and knew exactly who their lifeline was.

Automating the reporting was a structural decision. Programs this size answer sponsor reporting by adding people — the parallel program overseas carried a lead program management office (PMO) and five analysts for it. We automated 95% of that reporting into one source of truth, and action items escalated themselves. One person instead of five. The saving was not the point. Nobody spent their week writing up what had already happened, the data was live, and we acted on it live.

And the two jobs were not the same job. The overseas plants had moved to a modern platform two years earlier — clean data, a solved mapping problem, and a process estate already reshaped to fit mainstream software. Ours came off a system first built in 1978, patched for over thirty years with no documentation, where the processes had grown around what the software happened to allow. Every one of those processes had to be examined before the system could be configured around them. Same deadline, same MVP principle, a materially larger scope.

We spent the saving on the part that de-risks a go-live: twice the time on testing and training. Every project manager was hired as an expert in the function they owned, not a coordinator, and briefed to challenge the business on what it did and why. UAT ran purpose-built scenarios end to end — never the happy path — with a documented find-fix-retest loop. Every stage had a gate and none could be skipped. This gate required 95% of the issues blocking the program to be resolved. We cleared it at 98%. Training used UAT as its vehicle, so each paid for the other.

The data was the hard part. We extracted a static set and cleansed it across six or seven iterations while a parallel work stream mapped every field and built the conversion mechanism from nothing.

It did not go perfectly. People miscommunicated, misunderstood, and misread each other — inevitable when you're carrying an entire organization along instead of a project team. What mattered was that we were close enough to hear about it, and dealt with each one the week it happened rather than letting it grow into something that takes weeks to unwind.

Cutover in a single week, done two months early. Build and technical testing closed on schedule. Cutover ran in a single blackout week — all five plants at once. A big bang, not a phased rollout. Straight into hypercare, to the same discipline: 95% of the issues blocking the business closed — and by week two we were at 96%. We held hypercare open to month end anyway. The deadline was still two months away.

We inherited it four months in with nothing done, and went from a standing start to go-live in six months — no penalties, no missed date, no second attempt at the cutover, and five plants still run on the system today.

Written by Dan Harrington.