Six weeks, three tools, one platform
Acme's operations team ran on three systems: a ticketing tool for intake, a spreadsheet for scheduling, and a separate approvals inbox. Every request crossed all three, and roughly a fifth of them crossed twice.
Their CTO wrote the migration playbook. We shipped the imports and stayed out of the way, which is the arrangement that works best and the one we are least good at sticking to.
Six weeks from kickoff to the third tool being switched off. No parallel-running period beyond two weeks, by their choice.
Where the 40% came from
The headline number is workflow count, not time — 310 distinct workflows before, 186 after. It breaks down roughly like this:
About half were duplicates: the same process existing once per tool.
A quarter were routing steps that only existed to move a record between systems.
The rest were genuine consolidation — four approval paths that turned out to differ only in who got the email.
What we shipped for them
Two importers, one of which we kept and one of which was thrown away after the migration weekend and deserved to be.
A bulk reassignment action, built in week three because their playbook needed it. It shipped to everyone the following month, which is usually how our roadmap actually works.
A read-only archive view of the retired ticketing tool, because compliance needed seven years of history and nobody wanted to keep a licence for it.
The timeline, week by week
Week one was mapping: every process on a wall, one card each, colour-coded by which of the three systems it lived in. This is the week that produced the 40% — before a single record moved.
Weeks two and three were the first import and the build-out of the approval paths. Week four ran both systems in parallel with real traffic, and week five was the first week nobody opened the ticketing tool.
Week six was cleanup and the archive view. Their team took the switch-off decision themselves, on a Thursday, without asking us — which is the outcome we want and rarely get this cleanly.
What we would do differently
We under-planned the approvals cutover. Approvals are the part where a stuck record means a person waiting, and we treated them like any other entity for the first two weeks.
Their CTO's advice to other teams: consolidate the processes before you migrate the data, not after. We would add: the tool you are least attached to is the one to switch off first.
The 40% was not a feature we shipped. It was four duplicated approval paths that nobody could see while they lived in three different systems.