Estimate the graph's precision and recall with no ground truth

Measurement & estimationHard

Problem. You must report the graph's precision (are merges correct) and recall (does it catch a person's devices), but you never observe true identity. How do you estimate them?

Before you reveal: say your answer out loud, as if you were in the real interview — get your reasoning across clearly first. There is no single correct answer: reading what the interviewer is really after and defending your own thinking is what makes an answer strong.

← You can't randomize identity — how do you prove the graph lifts ads?Your match model has AUC 0.95 but ships badly — why calibration? →