Employee Net Promoter Score — eNPS — asks your people one question: on a scale of 0 to 10, how likely are you to recommend this company as a place to work? Subtract the percentage who answer 0–6 from the percentage who answer 9–10, and you have a number between -100 and +100.
It's the most widely adopted engagement metric in existence, and it earned that position honestly: it's one question, it takes eleven seconds, and it produces a single number an executive can hold in their head. This guide covers how to calculate it, what a good score looks like, and — the part most eNPS articles skip — the structural things it cannot tell you, and what to pair it with.
How to Calculate eNPS
Responses sort into three buckets:
- Promoters (9–10) — genuinely enthusiastic. They refer candidates and defend you in public.
- Passives (7–8) — content, not attached. They'd take a 15% raise elsewhere without agonizing.
- Detractors (0–6) — actively disengaged or on their way out.
The formula:
eNPS = (% Promoters) − (% Detractors)
Passives are excluded from the arithmetic but stay in the denominator. A worked example on 100 responses: 45 promoters, 35 passives, 20 detractors → 45% − 20% = eNPS of +25.
Two things follow from that structure and they matter more than people realize. First, passives are invisible — a company with 45/35/20 and a company with 45/35/20 arranged differently across teams score identically. Second, the same score can describe very different companies. An eNPS of +20 could be 40% promoters and 20% detractors (polarized, half the building unhappy) or 25% promoters and 5% detractors (bland but stable). Those demand opposite responses, and the headline number can't distinguish them. Always report the three-way split alongside the score.
What Counts as a Good eNPS?
Benchmarks vary by source and industry, but the broadly used bands look like this:
| Score | Read |
|---|---|
| Below 0 | Detractors outnumber promoters. Treat as urgent, not as a data-quality problem. |
| 0 to +20 | Typical. Most organizations live here. |
| +20 to +40 | Good. Meaningfully above average. |
| +40 and up | Strong. Rare and usually deliberate. |
| +60 and up | Exceptional — and worth auditing for anonymity problems before celebrating. |
Industry drags this around considerably. Software and professional services skew high; frontline retail, healthcare, and hospitality skew low for reasons largely structural rather than cultural. Comparing your hospital system's eNPS to a 200-person SaaS company tells you almost nothing.
The comparison that actually matters is you against yourself, last quarter. Trend beats absolute level, because trend controls for all the industry and workforce-composition variables you can't adjust for anyway.
Running the Survey Without Poisoning the Data
Quarterly is the right cadence for most companies
Annual is too slow to act on — you learn in November about a problem that started in March. Monthly produces survey fatigue and, worse, a score that jitters with news cycles and reorg rumors rather than tracking anything real. Quarterly is frequent enough to catch a trend and slow enough that people take it seriously.
Anonymity has to be real, and demonstrably so
This is where most eNPS programs quietly fail. If people suspect the score is attributable, you don't get bad data — you get flattering data, which is worse, because it's confidently wrong. Suppress results for any group below a threshold (8–10 responses is a common floor), say the threshold out loud, and never break it for a curious executive.
Always ask the follow-up
One open text field: "What's the main reason for your score?" The number tells you the temperature; the comments tell you what's burning. Any eNPS program without a qualitative field is generating a metric nobody can act on.
Segment, but carefully
Company-wide eNPS is nearly useless for action because it averages away the thing you need to find. A +25 overall can hide a +45 in engineering and a -10 in support. Cut by department, tenure band, and manager where headcount permits — and where it doesn't, don't fabricate precision by reporting a score for a team of four.
What eNPS Cannot Tell You
Here's the honest section, and it's the reason this post exists rather than another calculation walkthrough.
It's a lagging indicator. By the time someone rates you a 3, the disengagement is finished, not forming. The behaviors that produced that 3 — withdrawal, reduced collaboration, going quiet — started weeks or months earlier. eNPS confirms; it doesn't warn.
It's four data points a year. Quarterly sampling means an issue that appears in week two of a quarter goes undetected for eleven weeks. For a team of eight, that's an entire attrition cycle.
It measures sentiment, not behavior. People report what they believe about their workplace in the moment they're asked, which is influenced by recency, the last all-hands, and whether they've just had a bad sprint. Behavioral data doesn't have this problem — it records what people actually did.
It can't tell you where to intervene. A score of +10 doesn't identify the manager, the team, or the practice that produced it. The comments help, but comments are self-selected and skew toward the most articulate and most aggrieved.
None of this means don't run it. eNPS is cheap, comparable over time, and universally understood. It means don't run it alone.
Pairing eNPS With a Leading Indicator
The gap in every survey-only measurement program is the eleven weeks between surveys. The fix is to pair the quarterly sentiment number with something that updates continuously, in the flow of work, without asking anyone anything.
Recognition activity is one of the better candidates, because it's a genuinely voluntary behavior. Nobody is required to thank a colleague. When someone who gave and received recognition weekly for eight months goes quiet for three weeks, that's a behavioral signal with a date attached — and it typically shows up well before the survey does. The same is true of the reverse: a person who stops receiving recognition is often a person whose work has become invisible to their team, which is one of the more reliable precursors to a resignation.
We've written up the mechanics of this separately: using recognition data to detect quiet quitting and building an early-warning system for attrition both go deeper than we can here.
Our bias, stated plainly: we build Propsly, a Slack-native peer recognition tool that produces exactly this kind of behavioral data, so of course we like this argument. What we'd argue regardless is the pairing principle — one lagging sentiment metric plus one leading behavioral metric beats either alone, and it doesn't much matter whether the behavioral one comes from us.
A Practical eNPS Stack
| Layer | Cadence | Answers |
|---|---|---|
| eNPS + open comment | Quarterly | How do people feel, and is it moving? |
| Full engagement survey | Annually | Which specific drivers are weak? |
| Recognition / collaboration activity | Continuous | Who is disengaging right now? |
| Voluntary turnover rate | Monthly | What did all of the above actually cost? |
That last row is the one that converts the whole exercise into a business conversation. Employees who don't feel adequately recognized are about twice as likely to say they'll quit within a year (Gallup / Workhuman), organizations with strong recognition cultures report up to 31% lower voluntary turnover (Deloitte), and the Work Institute has repeatedly found roughly three in four voluntary departures were preventable. Replacement runs 33–150% of salary depending on role — a 100-person company at $65k average salary and 15% turnover spends about $487,500 a year on it.
An eNPS that drops from +30 to +15 is interesting. An eNPS that drops from +30 to +15 next to a turnover line that just moved three points is a budget conversation.
What To Do With a Bad Score
Three failure modes to avoid, in order of frequency:
- Explaining it away. "Response rate was low," "it was a hard quarter," "the survey went out right after the reorg." Sometimes true, always tempting, and the fastest route to a program nobody believes in.
- Sharing nothing back. Asking for feedback and going silent is worse than not asking. The next survey's response rate is the receipt for how you handled this one.
- Announcing an initiative. A low eNPS reliably produces a committee. Pick one specific thing the comments actually named, fix it visibly, and say that you did.
The strongest move after a bad score is unglamorous: publish the result, name one change, ship it before the next survey, and reference it when you ask again. That does more for the following quarter's number than any engagement program — and it's roughly the same discipline that makes broader engagement work stick.
The Short Version
eNPS is a good metric being asked to do a job it was never designed for. As a cheap, comparable, quarterly read on sentiment, it's excellent. As the sole instrument on the dashboard, it's four blurry photographs a year of something that changes daily.
Calculate it, report the three-way split rather than just the headline, segment it where the numbers support segmenting, always pair it with a comment field — and put something continuous next to it so you find out about problems in week two rather than week eleven.