Replacing working software is a risk decision, not a features decision. The question is never whether the new system is better. It is whether the transition can go wrong in a way you cannot recover from.
- ›The four migration mistakes that cause real damage
- ›How to design a rollback you will never need
- ›Sequencing by blast radius
- ›What to tell your team, and when
Sequence by blast radius
Rank your systems by what breaks if the migration goes badly.
Low blast radius: documents, internal knowledge, task tracking. If these hiccup for a day, work is annoying, not stopped. Move these first.
Medium: CRM, applicant tracking, project delivery. A bad day here costs deals and candidates, which is recoverable but visible.
High: anything touching money or legal obligation. Invoicing, payroll, compliance records. These move last, at a period boundary, with the old system live behind them.
Design the rollback first
Before you migrate anything, write down the answer to: if this goes wrong on day three, what do we do? A real rollback plan has three parts.
The old system stays live and readable for at least one billing period. Not archived, live.
A full export exists from before the migration, stored somewhere neither system controls.
Someone owns the call. Name the person who decides to roll back and the threshold that triggers the decision, in advance. Migrations fail slowly and ambiguously; without a named threshold, teams push through problems they should have stopped for.
Warning: The most common migration failure is not data loss. It is a team quietly maintaining both systems for months because nobody decided the cutover was done. Set the date, announce it, enforce it.
The four mistakes
Big-bang cutover. Moving everything in one weekend maximizes the chance that a problem in one system blocks everything. Move one system at a time.
Importing everything. Bringing 8 years of history into a new system imports 8 years of mess, and slows the migration for data nobody consults. Import live records plus 12 to 24 months of history; archive the rest.
Porting every automation. Most stacks have automations nobody remembers creating. A migration is the rare chance to rebuild only what earns its place.
Skipping the parallel period on the systems that matter. Parallel running is how you find the field you forgot, while it is still cheap to find.
Communicating it
Tell your team three things: what is changing, when the old tool stops being authoritative, and where to go when something is wrong. Most migration resistance is not about the tool; it is about people not knowing which system to trust on Thursday.
Name one person per system as the go-to during cutover week. Questions that go unanswered for a day become workarounds that outlive the migration.
A realistic timeline
For a 30-person company moving five systems:
| Week | Activity |
|---|---|
| 1 | Set up, configure, import people and permissions |
| 2-3 | Migrate task and project management, parallel run |
| 4 | Cut over projects; migrate CRM |
| 5-6 | CRM parallel run, cut over |
| 7-8 | Hiring and HR |
| 9-12 | Documents, then finance at the quarter boundary |
Slower than any vendor will suggest, and it is the pace at which migrations succeed.
FAQ
Can we migrate without downtime?
Yes, if you run parallel. There is no moment where nothing works; there is a period where two things work and one is authoritative. The authority handoff is the cutover.
What if the new system is missing a feature we need?
Find out during the parallel period, which is exactly what it is for. If the gap is real and material, you have not yet committed and the old system is still live.
Who should run the migration?
Someone who uses the systems daily, not only someone technical. The person who knows which fields actually matter is worth more here than the person who can write the import script.
Every paid Brainis plan has a 30-day trial with no card, which covers a proper parallel period. All 11 Operating Systems are on every plan, with no per-seat cost. See pricing.
Sharing insights on business operations, AI, and modern team management.
Run your company on Brainis
All 11 Operating Systems on every plan, from $29 a month. No per-seat pricing — you pay for AI capacity, not headcount.
See pricing