A 90-Day Retention Score Is Not Quality of Hire

Why the cleanest metric can be the most misleading

A new hire still employed on day 90 is easy to count. But staying is not the same as performing, and leaving is not automatically evidence of a bad selection decision. A person may exit because a manager changed expectations, the role disappeared or onboarding failed. SHRM's recruiting research reports that comparatively few organizations formally measure quality of hire. Treat a 90-day retention rate as one outcome with a defined observation window, not a quality label for a person.

The survivor-only performance trap

Suppose a company reviews only employees who remain at six months. Its performance average can rise after weaker performers leave—but it can also rise because high performers who disliked the manager left and are now missing from the dataset. Comparing only survivors creates selection bias. Keep all hires in the cohort, mark departures and record missing performance measures rather than silently dropping them. Never impute an exit as a zero performance score without a justified model.

Build a four-lens outcome card

Use four separate dimensions: job-relevant performance evidence, time to independently deliver agreed tasks, retention/exit context, and the candidate's assessment of expectation accuracy. Do not average these automatically into a universal index. A team with longer onboarding may look worse on ramp speed while hiring people into more complex jobs. Define role-specific milestones before the start date and record when the manager changed the target.

A concrete counterfactual: same recruiter, different managers

Two recruiters supply similarly qualified hires to two teams. Team A provides structured onboarding, clear ownership and frequent feedback; Team B has shifting priorities and no training time. If early attrition differs, attributing all of the gap to the recruiters confuses selection with environment. Segment outcomes by job family, manager, location and onboarding model before drawing causal conclusions. Small groups require caution: a single exit can move a percentage dramatically.

Do not punish candidates for taking a better offer after joining

An early departure might reveal misleading job previews or broken promises, not inadequate skill. Review the offer-to-work reality: responsibilities, schedule, tools, supervision, growth expectations and compensation. The relevant question is 'what was observable at selection time and what changed afterward?' A post-hire audit should improve the hiring process without creating an unfair permanent label attached to an individual.

A 30-hire retrospective with a blind spot register

Choose thirty hires from a closed start-date cohort, not thirty current employees. Record which have reached 90 days, which have reached 180 days, and which outcomes remain pending. Add performance evidence only where a job-specific rubric exists. For departures, use a neutral reason taxonomy with 'unknown' available. Keep a separate count of missing ratings and manager changes. Report the number eligible for each metric so later hires are not treated as failures.

The early-warning measure nobody asks for

Track expectation variance: how often the actual first-month work differs materially from the realistic job preview given during hiring. A high variance can damage both retention and ramp time even if the selection rubric was excellent. Sample new-hire feedback privately and aggregate before sharing. This is an operational hypothesis; do not present it as a validated predictor of resignation.

What a credible leadership report looks like

Publish retention at a defined horizon, performance coverage, ramp milestone attainment and onboarding context side by side. Include cohort size, missingness, time window and the number of people not yet eligible. Show sensitivity to one additional exit in small cohorts. Use the tool below to expose eligibility and denominator effects. Never turn the output into an automated ranking of recruiters or a claim of causation.

Run the operational check

This calculator uses only synthetic or aggregate values and runs entirely in your browser. It cannot score an individual candidate.

Implementation checklist and guardrails

Assign a named owner, agree the data definitions, document exclusions and check a small representative sample before changing the workflow. Review outcomes by job family and process stage. Protect candidate privacy, provide accessible alternatives and do not automate adverse decisions from these calculators. Recheck the process after two cycles and retain the results that contradict the initial hypothesis.

Research basis and further reading

Twenty topic-specific and adjacent methodological references. Some describe regulations or assessment standards, others are vendor studies or operational definitions. They do not all validate the proposed framework; source scope and jurisdiction must be checked before applying it.

1. SHRM — 2025 recruiting benchmarking report
2. CIPD — Resourcing and talent planning
3. CIPD — Turnover and retention
4. SHRM — Onboarding success measures
5. SHRM — Measuring quality of hire
6. SHRM — Effective recruitment teams in 2026
7. Ashby — Recruiting operations benchmarks
8. CIPD — Workforce planning
9. ICO — Recruitment and selection data
10. CIPD — Workforce planning topics
11. CIPD — Recruitment overview
12. EEOC — Employment tests and selection procedures
13. OPM — Structured interviews
14. OPM — Structured interview guide
15. OPM — Assessment strategy
16. OPM — Hiring process analysis
17. OPM — Interview scoring principles
18. OPM — Competency selection
19. Ashby — Source attribution definitions
20. Ashby — Candidates versus applications

The decision worth changing this week

Pick one measurable, reversible change, state what success and harm would look like, and publish the result with the denominator. Recruiting quality improves when evidence survives the handoff from recruiter to hiring manager.