Matching records.
A similarity score of 0.93 does not say which side of “the same person” it is on, and a merge cannot be undone quietly. A person reads the fields that decide it and answers the question the pipeline actually has: same, or not.
- Reads what decides itThe field that breaks the match, not the fields that agree.
- Same, or notA decision, where fuzzy matching only gives you a score.
- Priced by the mistakeMerging two people’s records costs more than a duplicate row.
- Only what you sendThe records you put in the question, and nothing from your database.
Are these the same person?
What happened
A shared family email, not one customer. The records stay apart, and neither sees the other’s history.
How it works
Step by step.
Send only the pairs your matcher could not settle. Quorum takes the uncertain middle, not the whole table.
Real-world example
A father and son, one inbox.A customer-data agent deduplicates accounts. Two records share an email address and a surname, and its matcher is 0.93 sure they are one person. Merged, a son would see his father’s orders and addresses. A person notices the birth years, and the merge does not happen.
For crypto agents
Where this saves a transaction.
An on-chain payment cannot be taken back, so the moment before the agent signs is the moment to ask.
deadline_ms to the window you actually have: if nobody answers in time you are refunded, and the agent takes its safe path, which is not to sign.