Duplicate Detection and Matching

The same person arrives as "Rob Smith", "Robert Smith" and "R Smith, 14 High St", and a system treating them as three people gives three partial histories and no complete one.

Where to find it

These features have no dedicated Architect Panel section of their own. They are configured through their datastores, opened from All Datastores, and most of what a caseworker sees appears on the record itself rather than on an admin screen.

Fuzzy rather than exact

Exact matching finds only the duplicates nobody would have missed. Matching allows for spelling variation, name order, transposed digits and formatting, because that is what real data contains.

Scoring, not deciding

Matching produces a score and a candidate list rather than a verdict. That is deliberate: automatic merging on a confident-looking score is how two different people with the same name and street become one person, which is far harder to unpick than leaving them separate.

Tuning the threshold

Set it so review is realistic. Too loose and the queue is full of pairs that are obviously different; too tight and you only catch what you would have caught anyway. Start conservative, review what it finds, and loosen once you trust it.

Conflict checking

For a legal practice this is also conflict checking. Run matching on the parties before a matter opens rather than after — a conflict found at file opening is an awkward conversation, and the same conflict found three weeks in is a professional problem.

The sweep

The Duplicate Detection Sweep task looks for new candidates in the background. It ships disabled; daily is usually right, because duplicates accumulate steadily rather than suddenly.