Most HubSpot migrations fail on process, not import tools. Teams dump every field, keep broken stages, and go live on Monday with duplicates and no owners. Use this checklist to migrate like an implementation — not a file transfer.
Decide what “done” means before you touch data
- Name the source CRM (or sheets) and the HubSpot hubs you’ll use on day one
- Write the sales/service motion in one paragraph — stages must match reality
- Pick a cutover window and a freeze rule for the old system
- Name one business owner and one technical owner for the migration
Map objects and fields — then delete half of them
List source objects (contacts, companies, deals, tickets, custom) and HubSpot targets. For every field: keep, transform, or drop. Empty historical fields that nobody uses are noise — don’t migrate noise.
- Required: email (or other unique key), lifecycle/stage, owner, create/close dates
- Transform: picklists that don’t match HubSpot stages or deal pipelines
- Drop: deprecated fields, personal scratch notes, unused custom objects
- Park: attachments and email history — migrate if legally needed, else sample first
Clean before you import
- Dedupe on email + company domain in the source
- Normalize countries, currencies, and phone formats
- Archive closed-lost junk and test records
- Export a sample of 50 records and validate the mapping by hand
Build HubSpot first, then move data
Stand up pipelines, properties, permissions, and basic automation in an empty (or clean) portal. Import into a designed CRM — never design around a messy import.
- Pipelines and stages locked with the sales lead
- Required properties and validation where HubSpot allows
- Teams, seats, and permission sets tested with two real users
- Reporting definitions agreed before go-live (what “won” means)
Cutover without chaos
- Freeze source CRM for net-new writes (or define a short dual-run with clear ownership)
- Final delta export → import → spot-check owners and open deals
- Redirect forms, sequences, and integrations to HubSpot
- Train on the new process the same week — not “docs later”
- Keep read-only access to the old system for 30–90 days
If you only do three things: map stages to real work, dedupe before import, and train in the week of cutover. Everything else is detail.