Data quality

The record matcher finds the records that stand for the same customer, supplier or product - in a list merged from several systems, or in one system that has been filled in by hand for years. Each field is compared the way that suits it: people's names with their initials and typing slips, organisation names with the legal form left out and the words in any order, e-mail addresses, telephone numbers however they are written, postcodes and other codes, numbers, dates and names that sound alike. Each agreement counts for as much as it tells, by the Fellegi-Sunter method, with weights set by hand or learned from the data, and only records that share a blocking key are compared, on a background thread, so a list of thousands takes moments.

Pairs that score between the two thresholds wait in a review queue, side by side with each field's similarity and weight, for a person to call the same or not the same from the keyboard. Matches join into clusters, and each cluster gets a golden record: every column taken by its own survivorship rule - the most recent, the most trusted source, the most complete record, the largest value and others - or from a record chosen by hand. Every run, decision and choice goes into an audit trail. Records come in as CSV or JSON; golden records, scored pairs and the records with their cluster go out as CSV, and the work saves as a JSON file to carry on with later or to run again on a server.

Try it with your own records in confidence: they are opened and matched by your browser, on your device. Nothing is uploaded, no copy reaches us or anyone else, and nothing is kept once you close this tab.