Retention risk scoring: the 5 signals that show up before the resignation

60-Second Summary

Quits are rarely sudden. Five observable signals — participation, performance, manager relationship, calendar, output texture — show up weeks before notice.
Score each signal 0-2. A 6+ score is a 30-day watch list candidate. 8+ means act this week.
The conversation isn't 'are you leaving?' (they'll say no). It's 'what would make this job better in the next quarter?' (they'll tell you).
Manager has 30 days to test one specific intervention. If nothing changes, accept the loss and start succession.

The phrase 'I had no idea they were unhappy' is almost always wrong. The signals were there, you just weren't looking. The five-signal model gives you a structured way to look, and the conversation script gives you a way to act before the offer letter lands on their desk.

The five signals

Signal	What to watch for	Where to look
Participation drop	Quieter in meetings, fewer questions, fewer initiatives volunteered	Meeting recordings, 1:1 notes
Performance plateau	Output stable but no growth; declining hard projects	Reviews, project tracker
Manager-relationship distance	Shorter 1:1s, more cancellations, formal tone in async	Calendar, Slack/email texture
Calendar shift	More personal-marked blocks, recruiter-shaped LinkedIn activity, mid-day gaps	Calendar (within policy), public LinkedIn
Output texture change	Less polish on shipped work, more 'minimum viable' submissions	Code reviews, design crits, written work

Scoring rubric

0-2 per signal, max 10

1
0
No change from baseline.
2
1
Noticeable change for 2+ weeks.
3
2
Sustained 4+ weeks AND others on the team have commented.

Action thresholds

Score 4-5: watchful. Score 6-7: 30-day watch list, schedule the conversation within 2 weeks. Score 8+: act this week. Above 8 you usually have 14-30 days before the resignation conversation.

The 30-day conversation

Script — opening (45 min, manager to report)

'I want to use this 1:1 differently than usual. I've noticed a few things in the last six weeks — [specific 1, specific 2]. I'm not trying to put you on the spot, and I'm not asking whether you're looking. I am asking: what would make the next quarter of this job genuinely better for you? Not in a wishlist way — in a specific, would-actually-keep-you-engaged way. I want to commit to one change in the next 30 days. Tell me what would matter most.'

Two rules: don't ask if they're interviewing (they'll lie, and you've just told them you noticed). Don't promise more than one change (you'll fail and confirm what they suspected). One specific, deliverable change is more powerful than a list of intentions.

Interventions that actually work

Move the needle / waste time

Often works

Off-cycle scope change — a project that uses an underused skill.
Manager change (if the relationship is the issue, comp won't fix it).
Visible learning investment — paid course, conference, mentor pairing — within 60 days.
Public stretch assignment with skip-level visibility.

Usually doesn't

Counter-offer with no scope change ('we'll match the new offer'). Buys 6 months max.
Vague promises of future promotion.
Generic recognition without scope change.
Asking them to 'hold on' until the next reorg.

When the intervention doesn't work

Re-score at day 30. If still 6+, the leverage is gone.
Begin succession quietly — knowledge transfer pairings, documentation, hire backfill if applicable.
Continue the human relationship — surprise resignations land worse than expected ones, but expected resignations are still resignations. Don't punish them in the meantime.
If they do resign, the exit interview is the calibration on whether your model needs an update — what signal did you miss?

Written by Pawan Joshi.Sources cited inline.

First published 16 Jun 2026See site changelog →