Auto-repair launched in waves across institutions. Estimate its effect.

Causal inferenceHard

Problem. Automatic repair (fixing a broken connection when the user logs in to the same bank in another app) was turned on for institutions in three waves: months 7, 9 and 11. You want the effect on the share of broken connections repaired within 30 days. A colleague runs a regression with institution and month fixed effects and a "treated" indicator and gets +16 points. Is that right?

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.

← Link redesign wins overall, loses at one big customer, and the split is off. Decide.12,000 institutions checked daily: how do you avoid drowning in false alarms? →