Where duplicates come from
Different systems assign different identifiers. Migrations re-create accounts. People change names, email addresses and roles. Contractors become employees.
Rules before probabilities
Match on shared, verified identifiers first. Only then use probabilistic signals such as name, date of birth and org unit, with a confidence score.
Review, do not auto-merge everything
Set a threshold below which matches go to a human. Record every decision. A wrong merge is worse than a missed one.
Preserve the mapping
Keep the link between every original identifier and the reconciled identity. Auditors and migration validation both depend on it.
Ready to see your learning data in one place?
Tell us which systems you run and we will show how the platform connects, normalizes and preserves their records.