Start with the destination
Export a small template from your CRM and list the fields your team actually uses. Decide which object receives each row: an account, contact, or another record type. A spreadsheet that looks tidy can still put information in the wrong place.
Protect existing records
Keep a dated backup of the fields that may change. Retain stable record IDs and define how new records are distinguished from updates. Write down whether blanks leave existing values alone or clear them; do not leave that decision to an import default.
Make a mapping sheet
For each source column, record its destination property, data type, allowed values, required status, and update rule. Separate company geography from contact geography and free-text roles from controlled seniority categories. Keep unrecognized values in a review file.
Test representative exceptions
Your pilot should include an existing record, a new record, a duplicate candidate, a blank field, and an unmatched company. Use a sandbox or a customer-approved test scope. Inspect both updated records and rejected rows before widening the import.
Reconcile the result
Compare input counts with created, updated, skipped, and failed counts. Check a sample of associations and preserve the exception log. Confirm that excluded customers were not reintroduced and that unrelated fields stayed intact.
Use your platform’s current instructions
CRM import behavior depends on the platform, object type, and account setup. HubSpot’s file-format guide explains identifiers and association requirements. Review the current official documentation before choosing an import mode.
Reference: HubSpot: format import files.
Use the sample-quality checklist or review the example field definitions.