# Analyze attrition risk and recommend retention actions

Finds who is likely to leave and why, and what to do about it.

- Topic: HR & Hiring
- Framework: RCTFVV
- Level: Advanced
- Tags: Analyze, Fill-in Template
- Source: https://www.thinkwithprompt.com/prompts/analyze-attrition-risk-and-recommend-retention-actions

## Prompt

```text
Role: Act as a people analytics specialist analysing voluntary attrition.

Context: Population and period: [POPULATION_PERIOD]. Attrition data by segment (team, tenure, level, location, manager, performance band): [ATTRITION_DATA]. Exit interview themes: [EXIT_THEMES]. Engagement / comp / workload data available: [OTHER_DATA]. Benchmark or prior year: [BENCHMARK]. Constraints on interventions: [CONSTRAINTS].

Goal: Identify where attrition is genuinely elevated and why, and recommend actions that are proportionate and fair, without profiling individuals.

Task: Produce: overall vs benchmark; segments where attrition is meaningfully elevated (with the numbers and whether the sample supports the conclusion); likely drivers triangulated from exit themes and other data (labelled as hypotheses); 3-5 retention actions targeted at the biggest drivers with expected effect, owner and cost; and what further data would confirm the drivers.

Constraints: Report at segment level only; do not score or flag named individuals as flight risks. Do not infer protected characteristics. State sample-size limits for every segment claim. Interventions must be equitable across the affected group.

Examples: Hotspot shape: 'Segment: [team], tenure 12-24 months - attrition [X]% vs [Y]% overall, n=[N] (sufficient / caution). Exit themes: [themes].'

Format: Sections: Overall vs benchmark | Elevated segments (table) | Driver hypotheses | Retention actions (table) | Data to confirm. ~600 words.

Verification: Recompute the segment rates and show n for each; flag every segment where n is too small to conclude.

Validation: Pressure-test as an HR director: would acting on this be fair to everyone in the affected segment, and is the driver evidence strong enough to spend on? Note what you would confirm first.
```

## What to fill in

- POPULATION_PERIOD
- ATTRITION_DATA
- EXIT_THEMES
- OTHER_DATA
- BENCHMARK
- CONSTRAINTS
- team
- themes

Free to copy and use. From ThinkWithPrompt (https://www.thinkwithprompt.com/), a library of 2,663 curated AI prompts.