---
name: exit-interview-synthesis
source: https://app.decimal.ai/s/exit-interview-synthesis@1/SKILL.md
source_sha256: b8a57e435d51
---

# Exit-interview synthesis: surface reason vs driver, and regrettable vs not

When you turn an exit interview or departure conversation into an analysis, the job is not to
transcribe what the person said — it is to classify *why they actually left* into a consistent
taxonomy the org can trend over time, and to tell leadership whether this exit is a problem they
should have prevented. Left to itself, the base model records the stated reason at face value, hangs
a single ad-hoc label on it, and treats any sympathetic reason as a "sad loss." The disciplines below
are the difference between a paraphrase and an analysis.

## 1. Standard departure-reason taxonomy

Classify every departure into exactly one **primary** category from this closed list (plus any
contributing categories). Use these categories — do not invent new ones per case:

- **Compensation & benefits** — pay, equity, or benefits below market or below expectation.
- **Management** — the relationship with their direct manager: poor management, lack of support,
  no recognition, feeling undermined or micromanaged.
- **Career growth & development** — no advancement path, blocked or stalled promotion, no learning
  or skill growth, feeling stuck.
- **Role fit & work content** — the actual work: unengaging, mismatched to skills or interest,
  unclear scope, not what was expected.
- **Workload & work-life balance** — hours, burnout, on-call load, lack of flexibility.
- **Culture & belonging** — values misalignment, exclusion or inclusion failures, interpersonal or
  peer conflict, a toxic environment.
- **Company direction & stability** — strategy, leadership churn, layoffs, financial or product
  uncertainty, lost confidence in the company.
- **Location & commute** — return-to-office mandate, commute burden, on-site requirement.
- **Relocation & personal circumstances** — moving, family, health, returning to school, other
  life changes outside the org's control.
- **Better external opportunity** — accepted another role. Treat this as a **surface** label almost
  always, not a primary driver (see §2) — probe what made them receptive.
- **Involuntary / performance** — termination, PIP-driven exit, layoff, or non-renewal.

## 2. Separate the surface reason from the underlying driver

The reason the person *states* is often not the reason they *left*. Record both, and make the
primary category the **underlying driver**, not the surface reason.

- **"I got a better offer" is a symptom, not a root cause.** People who are happy, growing, and
  fairly paid rarely take interviews. Ask what made them receptive: a comp gap, a blocked promotion,
  a manager they'd stopped trusting. That receptiveness is the primary driver; "better offer" is the
  surface reason.
- **A polite reason can be a cover.** Exit interviewees soften or launder the truth to protect a
  reference or avoid conflict — especially about their manager. "Just time for a new chapter" over
  vague, guarded answers, next to a team that has churned repeatedly, is a signal to look harder.
- **But do not fabricate a driver the evidence doesn't support.** When the interview genuinely lacks
  signal, say the underlying driver is unclear and name what would confirm it — do not invent one to
  fill the slot.
- **The test for the primary driver:** the factor that, if it had been different, plausibly retains
  the person. That is the primary category; everything else is a contributing factor.

## 3. Regrettable vs non-regrettable — keyed to the employee, not the reason

This axis is about **whom you lost**, not **why**.

- **Regrettable** — a strong performer, high-potential, or hard-to-replace person the org wanted to
  keep and would gladly rehire. Losing them is a talent loss.
- **Non-regrettable** — a low performer, poor fit, redundant role, or involuntary exit; the org is
  neutral or relieved.

The reason does not decide this. A **top performer leaving for relocation is still regrettable**. A
**weak performer leaving for higher pay is non-regrettable** — same "comp" reason, opposite verdict.
Read the employee's performance/value signal to set this flag; if it's absent, say so.

## 4. Preventability — a separate flag

Independently of who left, mark whether the org **could have prevented** it:

- **Preventable / actionable** — comp, management, growth, workload, role fit, RTO/location. The org
  had levers it chose not to pull.
- **Largely unavoidable** — relocation, health, family, returning to school, genuine life change.

The combination that matters most: **regrettable + preventable** = a talent loss the org caused and
could have stopped. That is the highest-priority signal for retention action, and it should be called
out explicitly, not buried among the other tags.

## Output shape

For each departure, report:
1. **Primary category** (one, from §1) — the underlying driver.
2. **Surface reason** — what the employee stated, when it differs from the driver.
3. **Contributing factors** — secondary categories, if any.
4. **Regrettable / non-regrettable** — with the performance signal you keyed it to.
5. **Preventable / unavoidable** — and an explicit flag when it's both regrettable *and* preventable.

## Worked example

> **Situation:** A senior engineer we rated a top performer is leaving. In the interview she says
> she "accepted an offer with better pay." She also mentions, more quietly, that she'd asked to lead
> a project twice and it never happened, and she wasn't sure where she'd grow here.
>
> **Face-value (avoid):** "Reason for leaving: better compensation elsewhere. A sad loss."
>
> **Synthesized (aim for):** "**Primary driver: Career growth & development** — two declined
> opportunities to lead and no visible growth path. **Surface reason: better external offer** — the
> pay bump is what let her leave, but the receptiveness came from being stuck. **Contributing:
> compensation.** **Regrettable** — top performer we wanted to keep. **Preventable** — a growth
> conversation and a real project could plausibly have retained her. This is a *regrettable,
> preventable* loss and should be flagged for retention review."

Same exit. The second version tells leadership what to fix and what it cost them; the first just
records what she said.