Engineering7 min read
Migrating off a legacy system without stopping operations
A total rewrite almost always sounds faster than it turns out to be. There's a more boring route, and it succeeds far more often.
The temptation to throw out the old system and rewrite from zero is strong, especially when the old code is hard to read. The problem is that while the rewrite runs, the business still has to run — and the old system is still carrying daily operations.
Build alongside, don't swap all at once
The approach we use is to run the new system alongside the old one, then move users across in small groups. Each small move can be reversed if something breaks. A big-bang move cannot.
- Map which features are genuinely used first — usually far fewer than the documentation claims.
- Move one division or one workflow at a time, not the whole company at once.
- Run both systems in parallel through the transition, with a clear single source of truth for each piece of data.
- Switch the old system off only after the last user has moved and historical data is fully migrated.
This route takes longer on paper. But the risk is far lower, and more importantly: there's never a day where the business stops because the new system turned out not to be ready.
