Attrition Driver Analyst
Finds why people leave, not what they said on the way out
19.3% of people left — and it is not spread evenly.
11 of 57 people left.
Tenure (years) <1 years leaves at 80% — 4.1 times the overall rate.
<1 years alone accounts for 73% of all leavers.
This shows where people leave, not why: reasons are not in the data. A driver needs 5+ people, 2+ leavers and 1.5× the overall rate.
Attrition by tenure (%)
Attrition by department (%)
- Support38.9%
- Sales15%
- Engineering8.3%
- Finance0%
Every segment
| Segment | People | Left | Rate | × overall |
|---|---|---|---|---|
| Tenure (years): <1 years | 10 | 8 | 80% | 4.1× |
| Manager: R. Hale | 9 | 7 | 77.8% | 4× |
| Department: Support | 18 | 7 | 38.9% | 2× |
| Level: Associate | 37 | 9 | 24.3% | 1.3× |
| Manager: A. Patel | 6 | 1 | 16.7% | 0.9× |
| Department: Sales | 20 | 3 | 15% | 0.8× |
| Manager: M. Chen | 14 | 2 | 14.3% | 0.7× |
| Level: Senior | 20 | 2 | 10% | 0.5× |
| Department: Engineering | 12 | 1 | 8.3% | 0.4× |
| Tenure (years): 3–5 years | 13 | 1 | 7.7% | 0.4× |
| Tenure (years): 5+ years | 14 | 1 | 7.1% | 0.4× |
| Manager: K. Novak | 12 | 1 | 8.3% | 0.4× |
What to do this week
- Talk to the people in Tenure (years) <1 years first: 8 of 10 left.
- Ask for the exit reasons of R. Hale: the data shows where, not why.
Do
- Compare like with like: look at rates, not counts.
Don't
- Don't treat one manager's rate as proof of fault: check team size and role mix.
How to improve
- Record the real reason each person leaves, at the exit interview.
- Re-run this every quarter on the same definitions.
Demo
Overview
Exit interviews are conducted by the organisation being left, in the final week, by someone the leaver may need a reference from. What they produce is "a better opportunity" and "for personal growth" — polite, unfalsifiable and useless. Every HR function knows this and most still build retention strategy on it, because it is the only thing that looks like data.
What is informative is structural and already recorded. When people leave in their tenure: a cluster at eight months is a hiring or onboarding problem, a cluster at three years is a progression one, and they need completely different responses. Whether leaving concentrates around a team, a change, or the period after a specific event. Whether leavers had been passed over, had no internal move available, or sat at the bottom of a pay band while new hires came in above them.
That last one is the finding organisations most reliably miss: paying new people more than existing ones for the same work is invisible until someone finds out, and then it costs several people at once.
This agent reports drivers with the evidence and what each would cost to address. It analyses patterns and never individuals: predicting who will leave leads to managers treating people differently on a model's say-so, which is a worse outcome than the attrition.
Benefits
The real driver, with evidence
Rather than a word chosen to keep a reference intact.
The right intervention
Onboarding and progression are different problems with different fixes.
A cost that justifies acting
Replacement priced against the fix.
What it does
Reads tenure, not exit interviews
A cluster at eight months is onboarding. A cluster at three years is progression. The stated reason distinguishes neither.
Checks pay position against new hires
Paying new people more for the same work is invisible until someone finds out — and then it costs several people at once.
Tests internal movement
Whether people who left had anywhere to go. Often the whole story, and never in an exit interview.
Finds concentration and timing
Around a team, a change, or the weeks after a specific event.
Costs the replacement
Recruitment, notice, ramp and lost output, so retention competes for budget properly.
Never predicts an individual
A leaving-risk score leads to managers treating people differently on a model's say-so. That is worse than the attrition.
What this changes
What this changes about your week
- It takes over
- Finding the structural reasons people leave.
- Done today by
- Exit interviews, conducted by the organisation being left, in the final week.
- How often
- Reviewed as a percentage, never as a pattern.
- Typical manual effort
- Tenure, pay position, internal movement and concentration all analysed together.
Why HR & People in particular. Exit interviews produce 'a better opportunity' — polite and unfalsifiable. The real driver is usually structural: new hires entering above existing people, or an internal move that was never available.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Attrition Driver Analyst has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolsextract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Requirements
Integrations
No integrations required.
Before you start
- Leavers and current staff with role, team, tenure and pay
- Internal moves and promotions
- Pay bands and new hire pay for the same roles
- Exit interview reasons, where captured
- Requires the All-Access plan or higher
Use cases
Rising attrition
Where it concentrates and what it has in common.
Annual pay review
Where existing people sit against new hires in the same role.
After a reorganisation
Whether leaving follows the change.
Board reporting
Retention with drivers and a cost, not a percentage.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Does it predict who will leave?
No, and it refuses to. An individual leaving-risk score changes how managers treat people, on a model's say-so, usually to that person's detriment. It reports structural patterns.
Does it use exit interview data?
It reads it and reports it beside the structural evidence, so you can see how far apart they are. That gap is usually the most instructive thing in the report.
How does it protect privacy?
It suppresses groups below the size you set and reports no individual. Where a team is too small to report on without identifying someone, it says so instead.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent