The most honest conversation in the employment relationship happens after the relationship has ended. That is a design flaw.
Exit interviews are honest and useless in roughly equal measure. Honest, because leaving removes the cost of candour — the departing worker finally says what the stay would not permit. Useless, because the decision the interview explains was made months or a year earlier, quietly, somewhere between a workload that stopped fitting the day and a concern that went nowhere.
By the time the truth is spoken at exit, it has been operating unaddressed in the organization for a long time. The interview does not reveal new information; it declassifies old information. And the organization that receives it usually files it as a story about one person — a move, a family reason, a better offer — rather than as a reading of itself.
The information leaders actually need exists much earlier, in what might be called the near misses of resignation: the month someone starts absorbing a vacancy without being asked, the supervision session that keeps being rescheduled, the role that has drifted so far from its description that the description has become a historical document. None of this is hidden. It is simply not being read as data.
Reading it as data requires a few disciplines, none of them expensive. Supervision that asks about load — not "how are you" but "what part of the work as designed no longer fits in the day". Attention to role drift, because people rarely resign from the job they were hired into; they resign from the job it quietly became.
Turnover is data. Most organizations receive it as weather.
A stay conversation done properly — which does not mean retention theatre, but a genuine inquiry into what would make leaving rational for this person, asked while the answer can still change something. People generally know what is wearing them down. They are rarely asked while it matters.
And pattern reading. One departure is a story; three departures from the same role in two years are a description of the role. The discipline is to read departures together, at least twice a year, looking for what they share — a manager, a workload profile, a career ceiling, a client group without adequate support. The pattern is usually visible in hindsight. The practice is to schedule the hindsight.
It also helps to count what a departure costs — recruitment, onboarding, lost continuity, the load on the people who stay, the relationships that leave with the person. Most organizations feel this cost without counting it. Counted, it makes the case for earlier action arithmetically rather than emotionally.
None of this makes the exit interview worthless. Its proper job is verification: a check on whether what the organization heard earlier was the truth. An exit interview that contains surprises is not evidence that the interview worked. It is evidence that everything before it did not.