HR metrics that matter vs vanity metrics: the eNPS critique and the time-to-fill trap
Which HR metrics actually predict business outcomes and which are theatre — with a deep, honest critique of eNPS, time-to-fill, and other metrics that quietly…
On this page▾
- A metric is real if (a) it's tied to a decision, (b) it correlates with a business outcome, and (c) it can't be gamed without changing the underlying behaviour. Most HR metrics fail at least one of these tests.
- eNPS is a single-question advocacy proxy. It's noisy (±10 swing from one detractor in a 30-person team), conflates engagement with brand, and is gameable by all-hands messaging. Useful as one signal among many; dangerous as a primary KPI.
- Time-to-fill is the most-quoted recruiting metric and the most-misleading. Optimising it incentivises sloppy hiring. Use time-to-productive-hire or quality-of-hire instead.
- Real HR metrics with proven business linkage: regretted attrition (segmented), quality-of-hire (90-day rating × retention to 12mo), first-year retention, manager effectiveness, internal mobility rate, span of control, time-to-productive-hire.
- The honest rule: if a metric can move 5+ points in a quarter without anything actually changing in the business, it's noise. Demand confidence intervals, not point estimates.
Every HR team measures things. Very few measure the right things. This guide separates the metrics that genuinely predict business performance from the ones that have spread because they're easy to count — including a full critique of eNPS and time-to-fill, the two most-misused metrics in modern HR.
What makes a metric real
- 1Does it drive a decision?If the metric moves, does anyone do anything different? If not, it's reporting, not measurement.
- 2Does it correlate with a business outcome?Is there evidence — internal or external — that movement in this metric corresponds to movement in revenue, productivity, retention, or customer outcomes?
- 3Is it ungameable?Can someone improve the metric without improving the underlying reality? If yes, it'll be gamed within 18 months and become meaningless.
Apply the test ruthlessly. Most HR dashboards keep 60% of metrics that fail at least one of these. Trim aggressively — fewer metrics, more weight, more decisions.
The eNPS critique (in full)
Employee Net Promoter Score, adapted from Fred Reichheld's customer NPS, asks one question: 'On a scale of 0–10, how likely are you to recommend [company] as a place to work?' Score = % Promoters (9–10) − % Detractors (0–6).
What's actually wrong with eNPS
- Statistical noise. In a 30-person team, one extra detractor swings eNPS by 7 points. In a 100-person team, by 2 points. Most quarter-over-quarter movement is noise, not signal — but reported as signal.
- Wrong construct. The question measures *advocacy intent*, not engagement, satisfaction, or retention probability. Three different things. eNPS is a weak proxy for any of them.
- Gameable. All-hands the week before the survey. CEO sends a heartfelt email. Promote a feel-good initiative. eNPS moves; nothing real changed.
- Ordinal-treated-as-interval. The math (subtracting %s) is statistically incoherent — a 0 and a 6 are both 'detractors' even though they're vastly different states.
- Survivorship bias. The people most likely to give 0–3 have already left. eNPS measures the people still here, which is a sample skewed toward people who haven't (yet) decided to go.
- Crowds out richer measurement. Teams that have an eNPS often skip Q12 / engagement / Kahn-style diagnostics because 'we already measure engagement'. They don't.
As one signal alongside richer engagement data, with confidence intervals, segmented by team and tenure, tracked over 4+ quarters before reading a trend. As a single quarterly KPI in a board deck? It's misleading at best, manipulative at worst.
The time-to-fill trap
Time-to-fill (TTF) measures days from req-open to offer-accepted. It's quoted on every recruiting dashboard. The problems are structural:
- Optimising TTF incentivises lowering the bar. The fastest way to fill a role is to accept the first hireable candidate, not the best candidate. Quality drops within 6 months.
- It excludes ramp time. A role 'filled' in 30 days takes another 16 weeks before the person is productive. The metric leaders actually care about is time-to-productive-hire, which TTF doesn't measure.
- It's gamed by closing reqs. 'We closed 40 reqs this quarter' often means 'we cancelled 12 reqs we couldn't fill and ignored 8 stale ones'. TTF goes down; nothing improved.
- It punishes hard hires. A 90-day senior leader hire scores worse on TTF than a 20-day junior hire — even though it was the better outcome.
(1) Time-to-productive-hire: req-open to person hits defined productivity threshold (varies by role). (2) Quality-of-hire: hiring-manager rating at 90 days × retention to 12 months. (3) Funnel ratios: pass-through rates at each stage (sourced → screened → onsite → offer → accept), with quarterly trend. These three together replace TTF entirely.
Other vanity metrics to retire
| Metric | Why it's vanity | Replace with |
|---|---|---|
| Training hours delivered | Measures effort, not outcome | Skill-acquisition rate (assessed pre/post) or business-outcome lift |
| Headcount growth % | Growth ≠ progress; can mask declining productivity | Revenue per employee, gross-profit per employee |
| Diversity hire % | Pipeline-stage metric; hides where the funnel actually leaks | Funnel pass-through by demographic at each stage |
| Engagement survey participation % | Measures survey hygiene, not engagement | Engagement scores themselves, with segmentation |
| Offer acceptance rate | Often 95%+ for any decent recruiter; little signal | Decline reasons (segmented) and pipeline drop-off |
| Cost per hire | Useful annually, misleading per-req (high-cost hires are often the best ones) | Hiring spend as % of payroll added |
| Manager-to-IC ratio (alone) | Without span analysis, says nothing | Span of control distribution + layer count |
The metrics that actually matter
- 1Regretted attrition (segmented)% of voluntary exits classified as 'regretted' (high performers, key roles). Segment by tenure, function, manager. Correlates strongly with team performance over the next 2 quarters.
- 2First-year retention% of new hires still employed at 12 months. Best single predictor of hiring quality. Industry benchmark: 85–90% is healthy, < 75% means the hiring process is broken.
- 3Quality-of-hireHiring-manager rating at 90 days (1–5) × retention-to-12mo flag. The cleanest measure of whether you're hiring well.
- 4Manager effectivenessComposite of team engagement, team retention, team performance distribution, 360 feedback. Correlates ~0.4–0.5 with team productivity in published research.
- 5Internal mobility rate% of open roles filled internally. Healthy: 25–40%. Below 15% suggests no development pipeline; above 60% suggests pipeline closure to external talent.
- 6Time-to-productive-hireReq-open to defined productivity threshold. Replaces time-to-fill.
- 7Span of control + layer countSpans 6–10 healthy, layers ≤ 7 for orgs under 5k. Correlates with decision speed and engagement.
- 8Revenue / gross-profit per employeeThe ultimate efficiency metric. Track over time; segment by BU. Useful for board-level workforce conversations.
Why HR metrics need confidence intervals
HR metrics are almost always reported as point estimates ('engagement is 72%'). But every people metric is a sample with sampling error. A 30-person team's eNPS has a 95% confidence interval roughly ±15. A 100-person team's about ±8. Reporting point estimates implies precision that doesn't exist.
(1) Always report sample size with any score. (2) Don't compare scores across small teams without confidence intervals. (3) Don't claim a 'trend' from one period to the next unless the change exceeds the noise margin. Most HR 'trends' reported to executives are statistically meaningless.
FAQ
Frequently asked questions
Are we supposed to drop eNPS entirely?
Not necessarily. Keep it if it gives leaders a simple shared number — but pair it with Q12 or equivalent, always show confidence intervals, and never make it the only engagement KPI.
How many metrics should an HR dashboard have?
5–9 at the company level, 3–5 per team. More than that and no one reads them. Pick the ones that pass the 3-question test and kill the rest.
What about leading vs lagging indicators?
Most useful HR metrics are lagging (attrition, retention, quality-of-hire). The few useful leading indicators: engagement scores (predicts 3–6mo attrition), manager 360 scores (predicts team performance), candidate decline reasons (predicts hiring quality).
Read next
All playbooksA short, defensible dashboard for leaders. Skip the vanity metrics and track what predicts hire quality and retention.
Most HR dashboards are unread. The reason is consistent: too many metrics, no decision context, no benchmarks, and no narrative.
Annual cadence is the wrong cadence for the only signal that actually predicts attrition. Here's the cadence that does.