Incremental matching
Re-running entity resolution on a source that grew by a small amount without re-scoring the entire dataset.
When a source adds 5,000 new rows to an existing 500,000-row dataset, you don't want to re-score every pair from scratch. Incremental matching processes only the new rows against the existing cluster index, deciding for each new row whether it joins an existing cluster or starts a new one.
The trade-off: pure incremental matching can miss cluster reshapes that a from-scratch re-resolve would catch (e.g., a new row that bridges two previously-separate clusters into one combined one). Most production MDM systems run incremental hourly + a full resolve nightly or weekly to catch those.
Golden Suite's incremental matching uses the per-row `notebook_event_id` lineage column on entity members to skip already-clustered rows on subsequent resolves.