Why people leave in the first 90 days: the signals from week two
Echipa HR 365 · reviewed 2026-09-04 · 5 min read
A departure at three months is not decided in month three. The decision usually forms in the first two weeks, when the person compares their expectations with what they find — and the signals are observable long before the resignation appears.
The five causes, in order of frequency
| Cause | When the decision forms | What would have prevented it |
|---|---|---|
| The role is not the one described | Weeks 1–2 | A job description written from the real work, not from titles |
| They have nothing to do | Weeks 1–3 | The first two tasks prepared before they arrive |
| They have nobody to ask for help | Week 1 | An assigned mentor who knows they are a mentor |
| They get a harsh correction too early | Weeks 2–4 | Feedback in private, on the fact, not on the person |
| Something promised at hiring does not happen | When it should have happened | Not promising, or writing it down with a deadline |
The observable signals
You do not need intuition, you need attention to visible things. All of the following appear before the decision, not after:
- They stop asking questions. In the first weeks, silence does not mean they understood — it usually means they stopped asking.
- They stop showing up for the optional things: lunch with the team, the morning chat. Social withdrawal precedes professional withdrawal.
- They deliver exactly what is asked, not a step more. In somebody new, the strict minimum is a sign, not a preference.
- They ask about leaving procedures, transfers or paperwork — even phrased neutrally.
- The manager cannot say, at one month, what they do well. If there is no answer, they probably have not worked together enough.
The fifth signal is about the organisation, not the person, and is the most useful of all. A manager who cannot name, at four weeks, one concrete thing the new joiner does well has not observed them enough — and that explains the first four signals too.
What you do when you see a signal
A ten-minute conversation, with three questions
- Manager — What has been different from what you expected when you joined? (not “are you happy?”)
- Manager — What would you like to do that you have not got to yet?
- Manager — What would be useful to you over the next two weeks?
The first question brings the real information. “Different from what you expected” is a neutral phrasing that does not ask the person to complain — so it gets an answer even from people who would never complain.
Why the exit interview does not help here
The exit interview is useful for patterns, at company scale. For early departures it is too late twice over: the information arrives after the person has gone, and it comes from somebody who stayed three months — so they cannot say what could have been changed, only what bothered them.
The right instrument is a 30-day check, with the same three questions for everybody. It costs ten minutes per person and catches the problem in the only window where it can still be solved.
What a departure at three months costs
A role with a gross salary of 1,200 EUR. The full cost of hiring: roughly 3,350 EUR. The post stays empty for another 6 weeks until it is refilled.
- Hiring cost, lost entirely
- 3,350 EUR
- The cost of a new hire
- 3,350 EUR
- Uncovered work, 6 weeks
- ≈ 1,800 EUR
- Total cost of a departure at 3 months
- ≈ 8,500 EUR
Nu intră în calcul:
- the effect on the team, which absorbs the difference and has already invested in the integration
- the reputational cost, hard to measure and real in a local market
The comparison that matters: a 30-day check for ten new joiners is roughly two hours of manager time. If it prevents a single departure out of ten, the ratio between effort and saving needs no further argument.
What is not the cause, although it looks like it
Two explanations appear almost automatically in discussions about early departures and are usually wrong. Worth ruling out from the start, because they send the intervention in the wrong direction.
| What is said | Why it is convenient | What checking usually shows |
|---|---|---|
| They were not the right person | It moves the cause outside the organisation | The same departures repeat on the same post, with different people |
| They got a better offer | It is uncontestable and demands nothing | The offer was actively sought, after the decision was already made |
The test that separates them: if two or three people have left the same post within a year, the cause is not in the people. A post that consumes good candidates in succession has a problem with the role, the manager or the context — and recruiting a fourth time does not fix it.
Where to start, if you have one intervention
With the 30-day check, for everybody, with the same three questions. It is the only intervention that needs no budget, needs nobody’s agreement, and produces information in the first month. The rest — better job descriptions, mentors, integration plans — gets built afterwards, on what you learn there.
Where to start
Look at last year’s departures and separate the ones under 90 days. If there is more than one on the same post, the cause is not in the people and the next recruitment will not solve it.
Introduce the 30-day check for everyone hired from now on, with the three questions. It costs ten minutes per person and is the only intervention that produces information in the window where you can still change something.
After ten checks, read the answers together. You will almost certainly find the same cause in several — and that is what you repair, not the individual cases.
Checks scheduled at 14, 30 and 60 days from hiring
The check-in conversations appear as tasks for the manager, with the questions already prepared — so they happen in the window where they still change something.
Free account, every module for 7 days, no card required.