Switching Insurance CRMs Without Losing Your Book

Contact migrations almost never fail. Name, phone and email move cleanly between any two systems on earth.

What fails is everything that makes the record insurance data: when consent was obtained and in what language, which carrier issued the policy and on what date, and when the chargeback window closes. Those three live in custom fields, get exported as unlabelled columns, and arrive in the new system as strings nobody maps. Six months later an agent needs to prove consent for a call made last March and the field is empty.

Plan the migration around those three, and the rest takes care of itself.

The field map

Build this spreadsheet before you export anything. One row per field in the old system, four columns: old field name · old field type · new field name · transform required.

The fields that matter, roughly in order of how badly they break:

Consent and compliance

• Consent obtained: date and time, not just date
• Consent method (web form, recorded call, written, inbound)
• Exact consent language shown to the prospect, verbatim
• Source URL or campaign the consent was captured on
• IP address, where captured
• Do-not-call flags, and the date each was set
• Revocation date and channel, where applicable

Policy

• Carrier, product, policy number
• Application date, issue date, effective date — three different dates that migrations routinely collapse into one
• Face amount, modal premium, mode, annualized premium
• Status (applied / approved / issued / active / lapsed / declined)
• Chargeback window length and window end date
• Commission level applied at time of sale

Lead

• Original source and sub-source
• Cost of the lead, if tracked
• Date acquired, distinct from date first dialed
• Disposition history, not just current disposition

Activity

• Call recordings, and the recording URL or file reference
• Notes, with their timestamps and authors intact

The single most common irreversible loss is consent timestamps degraded to dates. A date alone cannot establish that a call was placed after consent was obtained on the same day. Check the export’s time component before you commit — this is not recoverable once the old system is cancelled.

What exports actually do to your data

Dates get reformatted, silently. US systems export MM/DD/YYYY, some export DD/MM/YYYY, and the new system imports whatever it assumes. Any day-of-month under 13 imports without error and is simply wrong. Test with a record dated the 3rd of a month and verify it lands correctly before importing 40,000 rows.

Dropdowns become free text. A “Status” picklist exports as strings. If the old system used “Issued” and the new one uses “In Force,” every record lands unmapped or in a default bucket. Map every picklist value explicitly.

Multi-value fields flatten. Disposition history, note threads and activity logs frequently export as one concatenated cell, or not at all.

Recordings usually do not move. Most platforms export a URL that stops resolving when your account closes. If your retention obligation is three years and your recordings live in a system you cancelled in month two, you have a compliance gap that no amount of contact data fixes. Ask specifically: can I export the audio files themselves, in bulk, and in what format?

Attachments and documents — applications, illustrations, signed forms — have the same problem and are frequently overlooked entirely.

The sequence

Weeks 1–2: map and test. Build the field map. Export 100 records — not all of them — and import into the new system’s sandbox. Compare field by field against the source. This is where you find the date bug, the picklist mismatch and the missing consent time, while it costs you an afternoon rather than a book.

Week 3: full export, staged import. Import in this order: contacts, then policies, then activity, then consent records. Each layer references the one before, and importing out of order creates orphans that are painful to reconcile.

Weeks 3–5: parallel run. Both systems live. New activity goes into the new system; the old one stays read-only and paid for. Two weeks minimum, and do not shorten this because everything looks fine on day three. The failures you are protecting against surface when someone needs a record from four months ago, which does not happen on a schedule.

Week 5: cut over. Old system read-only for a further 30 days. Then export a final full archive — as files you hold, not as an account you log into — before cancelling.

Do not cancel the old system the day the import completes. The cost of one extra month is trivial against the cost of discovering in week six that recordings did not come across.

Rollback criteria, decided in advance

Write these down before you start, because in the middle of a migration everything looks like a small fixable problem.

Roll back if:

• More than 2% of policy records have a missing or wrong status after import
• Any consent record lost its timestamp and cannot be re-derived
• Chargeback window end dates are absent for in-force policies inside their window
• Agents cannot answer “who do I call next” in the new system by end of week 3

The third one is financial risk carried in the dark — the mechanics of why are in persistency and chargebacks and how CRMs track commissions and chargebacks.

Migrating from a spreadsheet

Easier in one way, harder in another. Easier because you control the format. Harder because spreadsheets accumulate inconsistency — three spellings of the same carrier, dates entered as text, phone numbers in five formats.

Clean before importing, in this order: normalise phone numbers to E.164, standardise carrier and status values against the new system’s picklists, split any combined name column, and convert every date column to ISO YYYY-MM-DD. An hour of cleaning saves a week of reconciliation.

If you are still on a sheet and wondering whether to move at all, the threshold test in do you actually need a CRM comes first — the field map above is deliberately written so a spreadsheet built to it migrates cleanly later.

Before you switch at all

Switching costs two to five weeks of partial attention. Make sure the thing you are switching to solves the problem you actually have — insurance CRM pricing for what the second invoice contains, and insurance CRM vs generic CRM for whether the object model is the real difference.