Data migration: checks to run before cutover

Data migration is more than importing a file. Information may be spread across old software, spreadsheets and a database. Different names may describe the same thing, or identical fields may follow different rules.

Inventory the sources before transforming data

List the sources, volumes and data that teams still need. Some information must move into the new system for daily work; other information can stay in a searchable archive. The people who use the data should approve that choice alongside the technical team.

Look for identifiers, duplicates, empty fields and values that no longer meet current rules. The same company, for example, may appear under several names. The new system cannot resolve that ambiguity on its own.

Document mappings and exceptions

For each field, record its source, destination and any transformation. A date format may change; an old status may map to several new statuses. Cases without a clear mapping should be flagged for a decision rather than assigned an arbitrary value.

The mapping table becomes a shared reference. It helps check the migration script and later explain why a value changed. Sensitive data needs special care: limit working copies and restrict access to the people involved.

Rehearse the migration with representative data

A first trial should include straightforward records and difficult cases: missing data, old values, duplicates and links between entities. Review the result in the new system with business teams. An imported row can be technically valid yet unusable in their day-to-day work.

Useful checks include:

• the number of records read, imported and rejected;
• comparison of key fields in a chosen sample;
• the expected links, such as those between customers and orders;
• the ability to rerun the import without creating duplicates.

Prepare cutover and final checks

Decide when changes to the old system will stop, how to handle updates since the last trial and who will approve the migration. Back up the source and plan a rollback if the result is not usable.

After the final import, compare volumes and have users check important workflows. A successful migration is judged by whether people can find and use the data in the new system, not just by an “import complete” message.