Most CRM migrations that go badly fail on data, not on software. Duplicate accounts, free-text fields used as makeshift pick-lists, contacts attached to the wrong company and ten years of abandoned leads all travel faithfully into the new platform unless someone stops them. This article is a working checklist, organised around the three moments that matter: before the load, during the test loads, and across the cutover window.

Why Data Quality Is a Cost Question

Cloud CRM platforms commonly charge in ways that relate to users, storage, API usage or add-on capabilities. Check each vendor’s current documentation for how they meter these. Migrating records nobody will ever use means paying to move, validate and store them. Cleaning first is therefore both a quality measure and a cost-control measure, which is why it belongs on a finance leader’s radar as much as an IT one.

Stage One: Decide What Deserves to Move

Start with a retention decision rather than a technical one. Sort records into three groups.

• Active: open deals, current customers, recent contacts. These migrate in full.
• Reference: closed deals, past customers, historical activity. These migrate, possibly with reduced detail.
• Archive: stale leads, long-dead opportunities, test data and duplicates. These go to cold storage or are deleted, depending on your retention policy.

Retention rules depend on your industry and jurisdiction. Where personal data is involved, ask your legal or compliance advisers what you must keep, what you must delete and what consent covers, before deleting or moving anything.

Stage Two: Clean Before You Map

Work through these in order, because later steps depend on earlier ones.

1. Remove obvious test and junk records.
2. Standardise values that should be controlled lists: country, state, industry, lead source, stage.
3. Deduplicate accounts first, then contacts, then leads. Duplicates at account level cause the worst downstream confusion.
4. Repair ownership: records owned by departed employees need a new owner or a holding queue.
5. Fix relationships, such as contacts with no account or opportunities with no contact.
6. Decide the fate of every custom field. If nobody can say why a field exists, flag it for retirement.

Keep a log of every bulk change. If someone later asks why a record changed, you need an answer.

Stage Three: Build the Mapping Document

A mapping document pairs each source field with a target field, a transformation rule and an owner. Dates, currencies, multi-select fields and notes are the usual trouble spots. Pay special attention to record IDs: external systems often reference the old CRM’s identifiers, so decide whether to preserve them in a legacy-ID field on the new records. That one field makes reconciliation, and later troubleshooting, much easier.

The Checklist

Item	Owner	Done When
Retention decisions documented	Compliance and business owner	Written sign-off
Duplicates resolved	RevOps	Agreed match rules applied
Picklists standardised	RevOps	Target values approved
Ownership repaired	Sales managers	No records owned by inactive users
Field mapping complete	Migration lead	Every field has a rule or is retired
Legacy-ID field created	Admin	Field exists in target
Test load reconciled	Data analyst	Counts and samples match
Integrations repointed in sandbox	IT	End-to-end tests pass
Rollback plan rehearsed	IT lead	Written steps and decision-maker named
Users trained	Team leads	Sign-off per team

Stage Four: Run Test Loads, Plural

Do at least two full test loads into a sandbox or trial environment before the real one. The first will expose mapping errors and rejected records. The second proves your fixes work and gives you a realistic timing for the load itself. After each, reconcile in two ways.

Quantity: do record counts per object match what you expected after exclusions?

Quality: pull a random sample of records, including awkward ones such as accounts with many contacts, and compare them field by field with the source. Ask sales users to spot check their own accounts; they notice problems analysts miss.

Check the vendor’s current documentation on import tools, batch sizes and API limits before estimating how long the production load will take.

Stage Five: The Cutover Window

Plan the window around business rhythm, not around convenience for IT. Avoid quarter-end, major campaigns and anything that coincides with month-end finance close. A typical shape looks like this:

• Announce a change freeze for the legacy CRM.
• Take a final backup and export.
• Run the delta load of any records changed since the last test load.
• Repoint integrations and email or calendar sync.
• Run the validation scripts and the sample checks again.
• Let a small group of power users run through real tasks before opening to everyone.
• Hold a go or no-go checkpoint with the rollback decision-maker.

Keep the legacy system in read-only mode afterwards for an agreed period so users can look things up while confidence builds.

A Hypothetical Walkthrough

Suppose a company has a large list of leads accumulated over many years. During stage one, the team finds that most leads have had no activity for a long time and cannot be tied to any consent record. Rather than load them all, they move active leads, archive the rest to inexpensive storage and note the decision in the retention log. The load is smaller, validation is faster, and the new platform starts life with data the sales team actually trusts.

Frequently Asked Questions

Should we clean data in the old CRM or in a staging area? Either works. Cleaning in a staging area keeps the live system untouched and makes your changes easier to repeat across test loads.

How do we handle attachments and notes? Decide whether they must live inside the CRM or can sit in document storage with links. Attachments often drive storage consumption, so check the vendor’s current storage terms.

Do we need to migrate full activity history? Not always. Many teams migrate recent history in full and summarise or archive older history. Choose based on how users actually search for it.

What is the most common cutover mistake? Forgetting an integration, such as a quiet billing or reporting feed, that only breaks after the weekend. Your integration inventory should be checked line by line.

Who signs off that the data is correct? Business owners, not only IT. Sales and finance leaders should confirm that pipeline totals and key accounts look right before the legacy system retires.

Next Step

Pick one object, usually accounts, and run the five cleanup steps on a copy of it this week. The duplicate rate and the effort it takes will tell you more about your real timeline than any estimate.