Skip to main content
All reports

Report 11 · People analytics

The eNPS Benchmark Report

How to run eNPS so the number means something: the honest benchmarks and why they disagree, the four ways the metric gets corrupted, and what to actually do with detractors, passives and promoters.

AudienceHR, People Ops and executives
Reading time13 min read
Length8 sections, print-ready
PublishedOmie · tryomie.com
Executive summary

The short version

eNPS is a thermometer, not a diagnosis.

One question ("How likely are you to recommend working here, 0 to 10?"), promoters minus detractors, on a range from minus 100 to plus 100. Global averages sit around plus 14 in the broadest database and plus 32 in a more self-selected 2025 study. The disagreement itself is the lesson: benchmark against your own trend first, industry second, vendor decks never.

Run it quarterly, anonymously, with a follow-up "why," and never wire it to anyone bonus. The moment you do, it stops measuring sentiment and starts measuring campaigning.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025) / 2 MIT Sloan Management Review and practitioner analyses (2024)

+14Global average

Perceptyx benchmark database of more than 20 million employees.

1 Perceptyx, QuestionPro and AIHR, 2025
+322025 US study

QuestionPro survey of 5,000 full-time employees, up from 25 in 2024.

1 Perceptyx, QuestionPro and AIHR, 2025
65–70%Response floor

Target response rate. Below roughly 60 percent you are sampling the willing.

3 SHRM, Journal of Applied Psychology and People Insight, 2025
4.2 moEarlier warning

How much sooner quarterly pulses catch problems than annual surveys.

3 SHRM, Journal of Applied Psychology and People Insight, 2025

What's inside

The mechanics, precisely: the arithmetic and the three properties people forget.
Reading a score: the interpretation bands and why a plus 30 is genuinely good.
Why the benchmarks disagree, and which number should decide anything.
Segment or learn nothing: the manager cut that hides in the average.
The four ways the number gets corrupted, and the guardrails.
The survey kit, a self-audit, and what to do with each group.
Decision

Benchmark against your own last four quarters. Treat plus 14 as the floor and the bands as the language for your exec slide.

Section 01

The mechanics, precisely

Get the arithmetic beyond dispute.
The eNPS calculation
eNPS = %promoters − %detractors

Respondents score 0 to 10. Promoters score 9 or 10, passives 7 or 8, detractors 0 through 6. Passives count in the denominator but not the subtraction, so the score can run anywhere from minus 100 to plus 100.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025)

A team of 20 with 8 promoters (40 percent), 7 passives and 5 detractors (25 percent) scores plus 15. The passives sit inside the denominator, drag on nobody, and quietly decide the number.

Three properties people constantly forget. It is harsh by design: a 6 is a detractor, and employees grade their workplace tougher than customers grade products, so an employee plus 30 is genuinely good where a customer plus 30 is average. Passives are invisible but decisive: they are typically 30 to 40 percent of respondents and the swing vote, moved into a promoter by a good manager or into a detractor by a messy reorg. Small teams are noisy: on a 12-person team, one person changing their mind moves the score 8 points, so below roughly 20 respondents you read direction, never decimals.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025) / 2 MIT Sloan Management Review and practitioner analyses (2024)

Decision

Below roughly 20 respondents, read the direction of travel, not the decimal.

Section 02

Reading a score

What counts as good, in plain bands.
Fig. 01. Interpretation bands (Perceptyx)
Below zero, more detractors than promotersAct now
0 to +10, modest positiveImprove
+10 to +30, healthyGood
+30 to +50, strongStrong
Above +50, exceptionalRare

Source: Perceptyx interpretation bands, scores range from minus 100 to plus 100

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025)

Calibrate the room before the first readout. Because a 6 counts against you and employees judge their workplace harder than they judge a product, an eNPS that would look mediocre as a customer score is a healthy employee score. Use the bands as the shared language on the executive slide: a plus 18 is not a crisis, and above plus 50 is rare rather than expected.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025)

Section 03

Why the benchmarks disagree

Same metric, different samples.

Two credible sources, two different numbers. The broad benchmark database includes reluctant measurers and lands near plus 14. The self-selected survey study skews toward organizations that already care and lands at plus 32. Neither is wrong, and neither is your benchmark.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025)

SourceSampleAverage eNPS
Perceptyx databaseMore than 20 million employees+14
QuestionPro 20255,000 US full-timers+32
Tech sectorWide spread+26

Treat plus 14 as the floor, the bands as the language for your exec slide, and your own last four quarters as the only benchmark that decides anything.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025)

Decision

Your own trend beats every published average. Compare yourself to yourself first, industry second.

Section 04

Segment or learn nothing

The manager cut hides in the average.

An org-wide plus 18 that hides a sales team at negative 20 is a resignation cluster with a good average. Always cut the score by team, tenure and manager. The manager cut matters most, because most of the difference between a thriving team and a leaving one sits with the person running it.

Evidence: 4 Perceptyx and Gallup (2025)

~70%Manager effect

Share of team-level engagement variance explained by manager effectiveness.

4 Perceptyx and Gallup, 2025
52%Preventable exits

Voluntary leavers who say their exit could have been prevented.

4 Perceptyx and Gallup, 2025
The rule

Never let the org average travel alone

Report the headline number next to the manager-level breakdown, and suppress any segment under 5 respondents so the cut stays anonymous. A single average is a press release, not a diagnosis.

Evidence: 3 SHRM, Journal of Applied Psychology and People Insight (2025)

Section 05

Four ways the number gets corrupted

And the guardrail for each.
Corruption 1

Tying it to compensation

The killer. In a study of 140 organizations, wiring survey scores to manager pay pushed scores up 12 percent while voluntary turnover rose 8 percent, so the metric measured campaigning, not sentiment. Report by team, coach on it, bonus no one on it.

Evidence: 2 MIT Sloan Management Review and practitioner analyses (2024)

Corruption 2

Breaking anonymity

Identifiable surveys inflate scores 15 to 23 percent, worst where psychological safety is lowest, which is precisely where you most need the truth. Keep it anonymous, and suppress any segment under 5 respondents.

Evidence: 3 SHRM, Journal of Applied Psychology and People Insight (2025)

Corruption 3

Timing it tactically

Surveying the week after bonuses, or skipping the quarter after layoffs, turns the trend line into fiction. Fixed cadence, same week each quarter, no exceptions. The discipline is the data.

Corruption 4

Changing the words

Swapping "How likely are you" for "Would you" shifts response patterns. Freeze the exact wording, the scale labels and the channel. A trend requires a frozen instrument.

Decision

Report, coach, and freeze the instrument. Never bonus anyone on the score, at any level.

Section 06

What to do with each group

The score is the start, not the deliverable.

The number is the headline; the action is the deliverable. Because roughly half of voluntary leavers say their exit was preventable, the value sits in what you do with each of the three groups once the score lands.

Evidence: 4 Perceptyx and Gallup (2025)

GroupWhat they are telling youThe move
Detractors (0 to 6)Something specific is broken, usually manager, load or stalled growth.Theme their reasons and fix the top theme visibly. Where a team skews detractor, run stay interviews across the team within a month.
Passives (7 to 8)It is fine. The swing vote, 30 to 40 percent of your workforce, one experience away from either direction.Growth is the passive mover: visible skill progress and a named next step. This is where learning investment shows up in eNPS.
Promoters (9 to 10)Your culture carriers and referral engine.Ask them the one-change question too, route them into buddy and mentor roles, and protect what they name as the reason they stay.

Evidence: 2 MIT Sloan Management Review and practitioner analyses (2024)

Decision

Passives are the program: 30 to 40 percent of your people, moved most reliably by visible growth.

Keep this page

The survey kit, the audit, and the takeaways

Three questions, eight checks, one page.

The survey kit: three questions, ninety seconds

1

Q1. The score

On a scale of 0 to 10, how likely are you to recommend [company] as a place to work? Ask it verbatim, every quarter, forever.

2

Q2. The why

What is the main reason for your score? Open text. This is where the actual information lives. Theme the answers by manager, growth, load, pay and culture, and report the themes beside the number.

3

Q3. The one change

What one change would move your score up a point? This converts sentiment into a ranked backlog. Close the loop publicly on the top item within 30 days.

Cadence and floor

Quarterly, fixed week, 90 seconds total. Chase a 65 to 70 percent response rate. Below 60 percent, report insufficient sample rather than a misleading number. Quarterly rhythm catches problems about 4.2 months earlier than annual surveys.

Evidence: 3 SHRM, Journal of Applied Psychology and People Insight (2025)

Self-audit: does your eNPS mean anything? Check every statement that is true today.

Fixed quarterly cadence, same week, run even when the timing is unflattering.
Frozen wording and scale, an identical instrument for four or more quarters.
Anonymous, with small segments suppressed, nothing reportable under 5 respondents.
Response rate at or above 65 percent, or the readout says insufficient sample.
Segmented by team, tenure and manager, so the org average never travels alone.
The reason behind each score gets themed and reported, not just the headline number.
The top requested change is closed publicly within 30 days, or explained.
No compensation is tied to the score anywhere, at any level.

Takeaways, the one-page version

Your own trend beats every benchmark. Averages disagree (plus 14 broad, plus 32 self-selected) because samples differ.
A 6 is a detractor and plus 30 is genuinely good. Employee scores run harsher than customer NPS.
The metric dies the day it touches compensation. Scores rise 12 percent, turnover rises 8 percent.
Anonymity is measurement infrastructure. Identifiable surveys inflate 15 to 23 percent, most where safety is lowest.
Quarterly, frozen, 90 seconds. Three questions buy about 4.2 months of early warning.
Passives are the program: 30 to 40 percent of your people, moved most reliably by visible growth.
Close one loop per quarter, loudly. The top requested change keeps response rates above the floor.

Evidence: 1 Perceptyx, QuestionPro and AIHR (2025) / 2 MIT Sloan Management Review and practitioner analyses (2024) / 3 SHRM, Journal of Applied Psychology and People Insight (2025)

Decision

This week: freeze the wording, scale and cadence in a one-page charter. This quarter: run the pulse, theme the whys, and close one loop publicly.

Appendix

Sources and method

Every external numeric claim in this report points to one of these 2024 to 2026 sources. Forecasts and self-reported surveys are labelled so they are not mistaken for causal proof.

1
eNPS benchmark databases and 2025 industry study
Perceptyx, QuestionPro and AIHR · 2025

Perceptyx benchmark database of over 20 million employees puts the global average near plus 14; QuestionPro 2025 study of 5,000 US full-timers lands at plus 32, up from 25; tech sector around plus 26; interpretation bands and the employee versus customer NPS calibration.

perceptyx.com
2
Survey scores, incentives and passive dynamics
MIT Sloan Management Review and practitioner analyses · 2024

Study of 140 organizations: linking survey scores to manager pay raised scores 12 percent while voluntary turnover rose 8 percent. Passives run 30 to 40 percent of respondents and act as the swing vote.

sloanreview.mit.edu
3
Survey response, anonymity and cadence research
SHRM, Journal of Applied Psychology and People Insight · 2025

SHRM response-rate benchmarks put the target at 65 to 70 percent; Journal of Applied Psychology anonymity research finds identifiable surveys inflate scores 15 to 23 percent; People Insight cadence research finds quarterly pulses catch problems about 4.2 months earlier than annual surveys.

shrm.org
4
Manager variance and exit research
Perceptyx and Gallup · 2025

Perceptyx manager variance research attributes about 70 percent of team-level engagement variance to manager effectiveness; Gallup exit research finds 52 percent of voluntary leavers say the exit was preventable.

gallup.com