Exit interviews name overload first and the boss second.

An HR business partner has done thirty exit interviews this year. She remembers the three angry ones. The other twenty-seven said something about workload, something about a promotion that did not come, and something kind about the team. The quarterly report says people leave because of managers.

What the study found

Hommelhoff, Keller and Stemmler reviewed 78 studies, surveyed 197 people and coded 312 real exit-interview transcripts. Leavers named three or four reasons each. The most common avoidance reason was stress from overload; problems with the boss came second and were seldom the only reason given. Reasons to go somewhere else, advancement above all, were as important as reasons to leave. The authors' title says it: turnover reasons are more complex than “people quit bosses”.

The same group's comparison between exit interviews and anonymous online reviews is the part HR should sit with. Advancement dominated what people said to HR; boss problems dominated what they wrote anonymously. People tell the employer the reason that costs them nothing to say.

Why one-at-a-time reading fails

1. Memory keeps the vivid interview, not the typical one. Three angry conversations outweigh twenty-seven measured ones.
2. Reasons arrive bundled. A leaver who names overload, no promotion and a new manager is coded under whichever the interviewer wrote first.
3. The sample is who agreed to talk. The people who left without an interview are the ones most likely to have had the boss problem.
4. Nobody compares across sites. The same reason at three plants reads as three anecdotes.

Reading them as a record

The useful unit is not the interview but the theme, counted across all of them: how often each reason appears, how the mix changes by population, site and manager, and how it moves over time, with anonymised quotations as illustration. Two rules make it safe. A minimum segment size below which nothing is reported, so that no theme can be traced to a person; US federal survey practice suppresses any cell with fewer than four responses, and that is a reasonable floor. And a stated purpose: manager-level views serve improvement, never appraisal, and that purpose is written down where the works council can read it.

That is the Coworker we build for exit interviews: themes by population, site and manager, above an anonymity threshold, with the weight of each theme and the quotations purged of anything identifying. It never reports an individual, never identifies who said what, and never produces anything about the people still employed. The individual conversation stays where it belongs, with the HR business partner who had it.

What to change this quarter

1. Code every interview against one short list of reasons, allowing several per leaver.
2. Report the mix quarterly by site and population, with the segment floor applied.
3. Compare the exit-interview mix with anonymous channels. Where they diverge, the anonymous one is closer to the truth.
4. Act on the first reason the record shows, which is likely to be workload, before the one the room expects.