eNPS for On-Call: Measuring Whether Engineers Hate the Pager (2026)
Updated May 2026. Sources: incident.io 2024 State of On-Call, DORA 2024 State of DevOps, Frederick Reichheld on NPS (Harvard Business Review 2003), public engineering blog write-ups on team-experience measurement.
Borrowing a Tool That Already Works
Net Promoter Score (NPS) was introduced by Frederick Reichheld at Bain in 2003 as a single-question instrument for measuring customer loyalty. The question, on a 0 to 10 scale, captured a remarkable amount of predictive information about customer behaviour with minimal survey burden. The HR community subsequently adapted the framework for employee satisfaction, producing eNPS (employee NPS): the same one-question instrument scoped to whether employees would recommend the workplace.
eNPS-on-call applies the same adaptation to the on-call experience specifically. The question: "On a scale of 0 to 10, how likely are you to recommend the on-call experience on this team to a peer engineer joining the company". Respondents scoring 9 to 10 are promoters; 7 to 8 are passives; 0 to 6 are detractors. The score is the percentage of promoters minus the percentage of detractors, giving a range from negative 100 (everyone is a detractor) to positive 100 (everyone is a promoter). Most well-run teams sit in the negative 20 to positive 40 range on this scoping.
The instrument is borrowed deliberately. eNPS-on-call is not a clinical-grade research instrument; it is a fast, cheap, repeatable signal that captures whether the team experience of being on-call is moving in the right direction or the wrong direction. The single-question form is the strength: response rates stay high (typically 70 to 90 percent of polled engineers respond), respondents complete in under 30 seconds, and the score is comparable across quarters and across teams.
The Question and How to Ask It
The cleanest form of the question is: "On a scale of 0 to 10, how likely are you to recommend the on-call experience on this team to a peer engineer joining the company?". Some teams prefer variants ("How likely are you to recommend this team's on-call rotation to a friend"). The phrasing matters less than the consistency: pick a wording and use it every quarter so scores remain comparable. Changing the wording from quarter to quarter compromises the trend signal that is the metric's main value.
Asked quarterly. More frequent surveys produce survey fatigue and degraded response quality; less frequent surveys miss within-year trends that matter for early intervention. Time the survey to fall between on-call rotations rather than during a particularly heavy or light week. Survey responses are anonymous, voluntary, with no individual reporting attached. Aggregate reporting goes to the engineering leadership team and to the SRE function; team-level trends inform tactical decisions, individual responses are never visible to managers.
The survey instrument can be as simple as a Google Form, a Slack poll, or a Linear cycle survey. Most modern engineering survey tools (Culture Amp, Lattice, Officevibe) support custom one-question surveys at this cadence. The infrastructure is not the work; the consistency is. Run the survey on the same week each quarter, publish the team-level result to the team within two weeks of close, and act visibly on at least one finding per survey cycle.
Benchmarks and What Scores Mean
| eNPS-on-call score | State | Interpretation | Action |
|---|---|---|---|
| +30 or above | Excellent | Team experience is genuinely good; engineers would recommend the rotation | Maintain; share patterns with other teams |
| 0 to +30 | Healthy | Most engineers are satisfied; a few have specific concerns to investigate | Standard quarterly review; address noted issues |
| -20 to 0 | Concerning | Significant fraction of engineers are detractors; attrition risk is elevated | Diagnose root cause via follow-up survey; commit to a quarter of intervention |
| Below -20 | Critical | Most engineers are detractors; structural intervention required to prevent attrition spiral | Leadership escalation; emergency on-call structure review |
Cross-organisation benchmarks are sparse because few organisations publish their on-call eNPS. Triangulating from related public data: incident.io 2024 found that 41 percent of on-call engineers had considered leaving because of alert load, which would imply a substantial detractor population in any organisation surveyed. DORA 2024 reports general developer wellbeing scores that correlate to on-call experience. The realistic conclusion is that many teams are likely in the 0 to negative 20 range; reaching positive 30 indicates a genuinely well-run on-call practice and is a worthwhile target.
Correlation With Attrition Intent
eNPS-on-call correlates well with attrition intent at the team level, less reliably at the individual level. A team with consistent negative eNPS over multiple quarters is more likely to see attrition above the team baseline; this is reasonably well-supported by the broader eNPS literature (Bain's original research, Reichheld's follow-up work) and by the more specific incident.io 2024 finding that alert-load-driven attrition consideration runs at 41 percent in noisy environments.
At the individual level the signal is weaker because the instrument is anonymous and because individual answers reflect strong-feeling responses that may not translate to actual departure decisions. Some engineers score 4 every quarter and never leave; others score 8 and leave the next month after an unrelated life event. Use the metric for team-level diagnosis (the aggregate has predictive value), not for individual intervention (the individual score does not warrant action).
Pair the eNPS score with departure interviews when departures do occur, to triangulate root causes. A departing engineer who scored low on eNPS-on-call for several quarters is a confirming data point; a departing engineer who scored high may have left for reasons unrelated to on-call, which is itself useful diagnostic information (the on-call practice is not the binding constraint at this team).
The Two-Question Follow-Up
The eNPS number alone gives you direction (improving, stable, deteriorating) but does not tell you what to fix. Pair the single-number survey with a two-question free-text follow-up, run alongside or one cycle after the eNPS instrument. Question one: "In one sentence, what is the worst part of the on-call experience on this team?". Question two: "In one sentence, what one change would most improve the on-call experience?". Free-text answers are aggregated and themed, surfacing the dominant intervention areas.
Common themes that emerge from this follow-up: alert volume (fix sequence in /alert-tuning), runbook coverage (read /runbooks-oncall), rotation cadence (read /follow-the-sun-on-call-cost and /two-tier-on-call-cost), compensation gap, manager support during on-call weeks, escalation clarity. The free-text aggregation typically points at one to three dominant themes that warrant intervention. Pick the highest-leverage theme and commit visibly to addressing it over the next quarter. The visible commitment is what makes the metric work: engineers who see real intervention following the survey become willing to respond candidly the next time, which keeps the data quality high over time.