Writing · from a pharma CRM migration
Boring cut-overs, by design
Notes from assessing a CRM migration: the parts of the plan that decide whether go-live is quiet or not.
Jul 2026·2 min read
Writing · from a pharma CRM migration
Notes from assessing a CRM migration: the parts of the plan that decide whether go-live is quiet or not.
A panel watched through a shift where nothing happens. Library of Congress, FSA/OWI Collection.
A cut-over is the moment you switch the old system off and the new one on. The work I was assessing was moving a commercial CRM in Europe off a legacy platform onto Veeva. The sales reps who live in that system every day would, on some agreed morning, find their accounts, contacts, and call history in a new place. The question the assessment had to answer was narrow: what does the plan need so that morning is uneventful?
Most of the answer isn't about the new platform. It's about a few decisions that get made early and quietly, long before anyone logs into Veeva.
The legacy system held years of accumulated history: duplicate accounts, contacts who'd left, fields people had quietly repurposed for something the form never intended. None of that is visible in a demo. So the first real task wasn't configuration, it was profiling what was actually in there and deciding, field by field, what earns a seat on the move and what gets left behind. A migration is only as clean as the data you agree to carry.
The plan I'd back moves one market at a time rather than everything at once. A wave is small enough that if something is off, you can see which market and which data, fix it, and apply the fix to the next wave. Between waves you set a plain checkpoint: did the records land right, and can a rep actually do their day's work in the new system? You don't start the next wave until that answer is yes.
Sequencing isn't caution for its own sake. It's how you keep any one mistake small.
For each wave you keep the ability to fall back to the old system for that market. It's rarely needed. But planning it forces you to answer, in advance, the questions you'd otherwise be answering in a panic: what state was the data in, who decides to reverse, how long do we hold both systems open. Writing those down is most of the value; you seldom have to use the escape hatch.
Put together, that's the shape of the recommendation: profile the data, move in waves, gate each wave, keep a fallback. None of it is clever. The reps don't want a story about the migration. They want their Monday to work. A plan that delivers a forgettable go-live is doing exactly its job.