Choosing a new CRM is a buying decision. Switching to it is an operations project — and it’s where brokerages actually lose leads. Not in the demo, not in the contract: in the week between “we signed” and “everyone’s on the new system,” when contacts export badly, text threads orphan, and nobody’s quite sure which database is live. Here’s the checklist we’d run, in order, so nothing falls in the gap.
Before you touch anything: inventory what “the CRM” holds
A real estate CRM isn’t a contact list. It usually holds contact records, tags and lead sources, notes, email history, SMS threads, tasks and follow-up schedules, active drip sequences, and sometimes transaction links. Write that list down for your current system, because your export is only as good as the weakest of those categories — a platform that exports names and emails but strips notes has just deleted your institutional memory.
Demand the full export first — before you sign
The most important step happens pre-contract with the old vendor. Ask, in writing: what exactly does your export include, in what format, and does it cover communication history or only contact fields? Vendors differ enormously here, and “we’ll help with migration” can mean anything from a real data move to a CSV of names. If the answer is vague, that vagueness is a cost of switching — price it into the decision.
The migration itself, in seven moves
- Freeze a date. Pick a low-inventory week. Migrations go wrong at the edges; don’t run one while 14 deals are closing.
- Export everything, twice. One working copy, one untouched archive. Keep the archive — it’s also your records-retention copy.
- Clean before you import. Dedupe, fix dead phones, merge duplicate tags. Garbage that migrates is garbage you’ll pay to clean forever, inside a system you now trust.
- Import in waves, not one dump. Agents with active deals first, then the long tail. Active leads are the ones where a dropped follow-up costs money this month.
- Run both systems in parallel for one week. Old system read-only for lookups, new system for all new activity. Nobody gets to “keep using the old one for a bit” — that’s how you end up with two half-databases.
- Rebuild automations deliberately. Don’t clone every old drip. You’re switching because something wasn’t working — rebuild the 20% of sequences that produced conversations and let the rest die.
- Decommission on a date. Cancel the old subscription the day the parallel week ends cleanly. An open old account is an open door to shadow data.
Where leads actually get lost
- Orphaned SMS threads. Text history frequently doesn’t migrate with the phone number. If your team texts from the CRM, ask both vendors how number portability works before scheduling anything.
- Dedupe collisions. An import that merges “Bob Smith” (client, active deal) with “Bob Smith” (vendor contact) by name alone will corrupt records silently. Dedupe on phone and email, not name.
- The dead week. Every day between export and import where inbound leads land in the old system is a day of leads someone has to remember to move. Shorten it to hours.
- Silent task loss. Follow-up reminders that exist only in the old system vanish the day it’s cancelled. Audit open tasks the week before cutover, not after.
One more option: don’t switch — orchestrate
Some brokerages ripping out a CRM don’t actually hate it; they hate re-keying between it and everything else. If that’s you, an orchestration layer that runs on top of your existing CRM — reading and writing through it, adding AI follow-up and the rest of the operation around it — avoids the migration project entirely. That’s a real design choice, not a dodge: it trades a big cutover risk for an ongoing integration dependency. The honest status of every connector we’re building is on our integrations page, and the CRM page covers both paths — native, or orchestrated on top of yours.
Published August 2026. Data-handling and records-retention obligations vary by state, association, and MLS — check yours before decommissioning any system, and treat nothing here as legal advice.