You sometimes hear a rule like: “If three direct reports resign within three years, the manager gets three strikes and is out.”
It sounds satisfying. It is also too crude.
Three resignations from a 20-person department over three years are not the same event as all three members of a three-person team leaving.
Now consider a sharper pattern:
3 people → all 3 leave → 2 replacements join → 1 of those 2 leaves within six months.
At that point, the most informative evidence is not the polished sentence written on each resignation form.
It is the sequence showing that people do not stay.
1. “Four people resigned” throws away the most useful information
If HR records this only as “four voluntary exits,” the structure disappears.
The original cohort went from three people to zero. Its retention rate was 0%.
Two people were then hired, and one left within six months. The replacement cohort also showed early attrition.
The useful representation is:
3 → 0 → 2 → 1
Imagine products repeatedly falling off the same production line, while management reports only, “Four products fell.” The total matters, but the location, timing, recurrence, and operating conditions matter more.
Turnover should be read the same way.
2. “Three people in three years” means little without a denominator
Use person-years as a rough illustration.
Three exits over three years would be:
- an average five-person team: 3 / 15 person-years = 20% per year
- an average ten-person team: 3 / 30 = 10%
- an average twenty-person team: 3 / 60 = 5%
The exact same count creates very different pictures.
That is why an absolute “three resignations and the manager is out” rule is closer to importing baseball into HR than doing workforce analytics.
The meaningful dimensions are rate, time, concentration, and recurrence.
3. In tiny teams, the sequence can matter more than the percentage
Small-team percentages are noisy.
One departure from a three-person team instantly creates a 33% rate. Two creates 67%. So a big percentage alone does not prove a broken manager or workplace.
What strengthens the signal is recurrence.
The original group disappears.
The company replenishes the team.
A newly hired person also leaves quickly.
The first wave might be explained by timing: several people happened to be ready for a change.
But when the replacement cohort also starts leaking, there is evidence that hiring alone did not remove the underlying condition.
That condition could be the manager, but it could also be workload, pay, job design, poor onboarding, hiring mismatch, weak career prospects, or team culture.
The correct conclusion is not “manager guilty.”
It is “this is enough to justify an investigation.”
4. Exit-reason labels are filtered self-reports, not system logs
Organizations often collect explanations such as:
“career growth,” “family reasons,” “a new challenge,” or simply “personal reasons.”
Then someone says: “See? Everyone had a different reason.”
That inference is too fast.
A 2025 Journal of Vocational Behavior study found that turnover reasons commonly occur in combinations. In its survey, respondents selected roughly three to four reasons on average.
When asked how much of their true reasons they had disclosed in resignation conversations, respondents reported an average of 73%.
Among people who reported boss issues, the average disclosure level was 64%, compared with 78% among people without boss issues.
So “I found a better career opportunity” can be true.
“I also wanted to get away from this manager” can be true at the same time.
Reasons are not mutually exclusive.
Treating one polite exit label as proof that no workplace factor existed is like turning off the security camera and trusting a single questionnaire.
5. But “people quit bosses” is too simple too
The opposite slogan can also mislead.
“People do not quit companies; they quit managers” is memorable, but the evidence is more complicated.
The same 2025 research found important roles for overload-related stress, better advancement opportunities elsewhere, and boss issues. Reasons often appeared together.
So “four exits means the boss did it” is also a bad inference.
A better question is:
“Why is turnover clustering in this management unit, and why did replacement hiring fail to stabilize it?”
That is process diagnosis, not a witch hunt.
6. What does the Japanese baseline look like?
Japan’s Ministry of Health, Labour and Welfare reported an annual 2025 separation rate of 11.4% for regular workers.
For regular workers in manufacturing, it was 9.2%.
The rate attributed to personal reasons was 10.2%.
Those national figures are useful context, not a courtroom threshold for a three-person team.
They combine employers, occupations, ages, regions, and workforce structures.
For a small team, the stronger comparator is usually another similar team inside the same company: same site, similar work, similar hiring pool.
If one manager’s unit repeatedly loses people while comparable units do not, that concentration is informative.
7. What HR should monitor instead of one exit-reason field
A serious review would track at least:
- voluntary turnover by manager over rolling 12- and 24-month windows
- six- and twelve-month retention of new hires
- transfer requests, leave, absence, overtime, and other strain indicators
- multiple contributing reasons rather than one forced exit category
- confidential or anonymous channels separate from the direct manager
- matched internal comparison groups
- before-and-after patterns when management changes
- stay interviews with employees who remain
This turns turnover from paperwork into an observable operating system.
8. When does “huh?” become “investigate this”?
There is no universal magic count.
The signal becomes stronger when several of these occur together:
- nearly the whole original team leaves
- replacements also leave within roughly six to twelve months
- departures cluster under one manager
- comparable teams do not show the same pattern
- transfer requests, absence, leave, or overtime problems rise too
- similar themes appear through confidential channels
- retention improves after a manager or structural change
A pattern of 3 out of 3 leaving, then 1 out of 2 replacements leaving within six months does not automatically justify discipline.
But it is well beyond the point where doing nothing is an analytically comfortable response.
9. Five mistakes to avoid
Do not automatically punish a manager based on a raw count.
Do not automatically clear a manager because every resignation form has a different polite label.
Do not apply a national average mechanically to a tiny team.
Do not study only leavers and ignore the people still inside the system.
And do not wait for a beautiful statistical significance test while the department turns into an empty room.
In small teams, repeated operational patterns can become useful before formal inference becomes strong.
10. Conclusion: exit reasons are notes; turnover patterns are logs
An exit reason matters, but it is one person’s explanation in one social situation.
By contrast,
3 people → 0 → hire 2 → 1 leaves within six months
is a log emitted by the organization itself.
If management reads only the notes, every departure can be made to sound unique.
Family reason.
Career growth.
Personal choice.
New opportunity.
Meanwhile, the same seat keeps emptying.
The answer is not a crude “three strikes and fire the manager” rule.
It is simpler:
When people repeatedly disappear from the same place, investigate the structure, not just the label.
