A migration is the one moment you can improve data quality at scale. Most organizations use it to copy their problems into a new system faster.
Before you export anything
- Decide what a household means for your organization and write it down.
- Decide how anonymity is currently recorded, and how it should be.
- Identify the fields you know are unreliable — they should import as review candidates, not as truth.
The dry run
- Run the import in report-only mode and read the counts: created, matched, merged, rejected.
- Sample twenty matched records by hand. If more than one is wrong, tighten the matching rules.
- Confirm gift totals per fund and per year against your old system before writing anything.
After the write
- Keep the source file. A migration without a reproducible source is not reversible.
- Work the duplicate queue in the first week, while the context is fresh.
- Re-run the totals check and record the result — this is the number you will be asked about.