Confidence, not coincidence
Matching “John Smith” by name alone is how bad data gets made. Every inferred relationship in the graph carries a confidence level and shows the evidence behind it: shared appointments, shared addresses, filing patterns. That lets you judge the link before you act on it.
Built on the same record
The graph is built on the point-in-time event store, which means relationships can be reconstructed as they stood on any date, not just as they look today. That matters when you need to show what was knowable at the time of a decision.