Should we clean data before or after import?
Clean obvious structural problems before import, then use the destination system to review relationships and uncertain duplicates. Do not rely on one stage to solve everything.
Lynka guide
Move the useful history. Leave the mess behind.
A CRM migration is a chance to clean the customer model, not a requirement to reproduce every field and duplicate from the old system.
Most migration risk is created before the import button is pressed. The team does not know which source is authoritative, company and contact records are mixed together, owners no longer work at the business, or the same customer exists under several spellings.
A controlled migration treats data as a set of business records with owners and relationships, not as one giant spreadsheet that must be copied exactly.
01
1. Inventory every source before combining anything
List the current CRM, spreadsheets, accounting exports, contact lists and team-owned files that contain customer or sales data. For each source, record who owns it, what date range it covers and whether it is still actively maintained.
Do not merge files yet. First identify overlap so you know why the same customer may appear in three places.
02
2. Separate record types
Companies, people, leads, opportunities and products are different records even when an old spreadsheet stores them in the same row. Decide what each column actually represents before mapping it.
A company name should not become a contact name simply because the old sheet had one “Name” column. Relationships deserve deliberate mapping.
03
3. Clean values that drive matching and reporting
Standardize important fields such as email, phone, country, industry, status and owner. Remove obvious formula artifacts and malformed rows before the migration becomes responsible for interpreting them.
Leave original source files unchanged and perform cleanup in a working copy so the team can always trace a migrated value back to its source.
04
4. Decide how duplicates will be recognized
Exact email matches are useful for people, while company duplicates may require name, domain or other business context. Do not automatically merge records only because their names are similar.
Create a review queue for uncertain matches. A wrong merge can be more damaging than leaving two records separate for a short period.
05
5. Map only fields with a destination and purpose
Create a mapping table from source field to destination record and field. Mark columns as migrate, transform, archive or discard.
If a custom field has not been used for years and no report depends on it, migration is not a reason to preserve it forever.
06
6. Import a representative test batch
Choose records that exercise the difficult cases: companies with several contacts, active opportunities, products with suppliers, foreign currencies and incomplete fields.
Review the result inside the actual UI. A technically successful import can still be operationally wrong if relationships, ownership or field meanings were mapped badly.
07
7. Reconcile counts and spot-check relationships
Compare source and destination counts by record type and status. Then inspect individual accounts end to end rather than trusting totals alone.
Check that contacts belong to the right companies, active opportunities have the right owners and product values did not shift because of formatting.
08
8. Plan a short, explicit cutover
Choose when the old system becomes read-only for the records being migrated. Capture the final delta after that point and make the new system the place where fresh work begins.
Keep the old data available for reference until the team is confident the migration is complete, then retire duplicate working files so data does not immediately split again.
Migration control list
- All source systems and spreadsheets are inventoried.
- Record types are separated before mapping.
- Owners and statuses are normalized.
- Duplicate rules are documented.
- Every source field has migrate, transform, archive or discard status.
- A representative test import was reviewed in the UI.
- Counts were reconciled by record type.
- Relationships were spot-checked.
- Cutover and read-only timing are agreed.
- Original exports are retained according to company policy.