Skip to content
Menu
Home / Article

Migrate a sales CRM: checks before cutover

Before cutting over, export every property, test the fields you will map, and filter deals by pipeline so won deals aren't counted twice.

Our testSep 2026 · full HubSpot export, checked field by fieldAll our tests ↗
The short answer

A CRM migration carries bad data along with the good unless you check it first. Export every property, test the fields you plan to map against a sample, and filter deals by pipeline before counting anything.

Reconcile records and post-cutover writes before retiring the old CRM.

Leave with: a migration field map and reconciliation sheet.

Download the editable worksheet (CSV) →

Bring: source ids, complete exports and a cutover window.

This is the checklist we would run before moving off HubSpot or any sales CRM. It comes from a full export of our HubSpot portal in September 2026 and the problems that export turned up. Most of those problems were in our own setup, and most migrations would copy them across without anyone noticing.

1. Export every property

Our sync had been reading fewer than a dozen of a few hundred company properties, and the rest held data we would have left behind. Export all of them. HubSpot's search API stops at 10,000 results, so page through the list endpoint or split the export by create date.

2. Filter deals by pipeline

We had set up a second pipeline for implementation, and each closed-won deal created a twin there with no amount. In an unfiltered count of paid-search wins, nearly half were twins. HubSpot exported what our configuration created; the double count was ours. We caught it because the average deal size came out far below what we knew a first deal was worth.

Filter to the sales pipeline before you count wins or sum amounts, in the old system and the new one.

3. Test the flags you plan to map

Status fields are where migrations go wrong without anyone seeing. In our portal the current-customer flag was true on almost every company, an older customer flag had been set by name matching and was wrong in both directions, and the churn flag disagreed with deal history on a hand-checked sample. Rebuild these from billing or deal history and leave the old flags behind.

4. Find what writes each field

Our attribution fields were almost always blank, and two connected sync tools wrote to the same ones. Before you map a field, list every integration that writes it, and switch off the ones the new CRM will replace.

5. Carry over the history

Lost reasons, contacts per deal and closed-lost records are hard to rebuild. In our export, won deals had a median of 6 contacts against 2 on lost deals, and about a third of closed-lost deals had gone dark. That history is worth moving even if no report uses it yet.

6. Reconcile before you retire the old CRM

Run the same counts in both systems on the same date, then freeze writes to the old one.

Check Old CRM New CRM Match
Companies
Contacts with a job title
Closed-won deals, sales pipeline only
Closed-won amount, same date range
Open deals by stage
Fields mapped, each with its source system

Keep the field map beside it: old field, new field, source system, and whether the value was checked.

Comparing destinations: Attio vs HubSpot, HubSpot vs Pipedrive and Close vs HubSpot. Before moving data you might buy again, read CRM data enrichment.

Questions buyers ask

How do I migrate from HubSpot to another CRM?

Export every property through the API, filter deals by pipeline, test status flags against a sample, then reconcile counts in both systems on the same date before switching off HubSpot.

Why are won deals double counted in a HubSpot export?

In our case, a second pipeline. Our setup created an implementation-pipeline twin for each closed-won deal. Filter to the sales pipeline before counting.

What data gets lost in a CRM migration?

The fields nobody reads, and the history: lost reasons, contacts per deal and closed-lost records. Export all properties, not only the ones you map.

Sources

Sources and evidence behind this page