"Is our NPS good?" is the most common question asked about the metric and the least answerable one as phrased. The honest answer is always: good compared to what? This post gives you the realistic industry ranges in one table, explains where those numbers come from and why they spread the way they do — and then makes the case that the comparison you actually need is with your own score three months ago.

NPS benchmarks by industry: the table

These are the ranges where published benchmark studies cluster, cross-checked against what we see in survey programs run on Formspring. The primary public sources are Retently's annual NPS benchmark studies, the Bain & Company methodology and research that defined the metric, and the Qualtrics XM Institute's NPS guidance. Treat every range as a fog bank, not a target — the methodology section below explains why the same company can land 20 points apart in two different studies.

Industry Typical NPS range Notes / source
B2B SaaS +30 to +40 Retently's B2B SaaS panels publish averages in the high 30s; +50 and above is exceptional, and scores above +60 usually mean only the happiest segment is being surveyed.
Software & tech (consumer) +28 to +45 Wide spread between beloved consumer apps and utility software; Qualtrics XM Institute data shows tech consistently above cross-industry medians.
E-commerce & retail +40 to +50 Among the highest published ranges (Retently); simple interactions, well-calibrated expectations. Premium and niche brands run higher, mass-market discounters lower.
Banking & financial services +20 to +35 Bain's banking research shows direct/digital-first banks outscoring incumbents — partly product, partly self-selected early-adopter customer bases.
Insurance +15 to +30 Published ranges cluster in the mid-20s; the claims experience dominates the score because the product is only "used" on a bad day.
Telecom & cable 0 to +15 The reliably lowest published category; negative scores are not unusual in panel studies. An ISP at +20 is genuinely outperforming.
Healthcare +10 to +50 Extremely wide: providers often score high (people like their doctor); payers and insurers score near telecom levels. Check which side a benchmark measured.
Education & e-learning +25 to +45 Course platforms and tutoring score toward the top of the range; institutional/administrative experiences pull the bottom down.
Hospitality & hotels +25 to +45 High variance driven by premium-vs-budget positioning; published hotel averages sit in the 30s.
Airlines +20 to +40 Bain's original research used airlines as a flagship example; budget carriers and premium carriers occupy opposite ends of the range.
Professional services +35 to +50 Agencies, consultancies, accounting — relationship businesses with chosen, high-touch engagements publish some of the highest B2B ranges.
Logistics & delivery +15 to +35 Scores track delivery reliability almost one-to-one; B2C parcel delivery sits lower than contracted B2B logistics.
Streaming & media +15 to +40 Subscription fatigue and price increases have dragged published medians down in recent waves; individual beloved services still post +40s.
Fitness & wellness +35 to +55 Boutique studios and connected-fitness brands publish some of the highest consumer scores; identity-adjacent products recruit promoters easily.

Phrase to keep in mind when quoting any of these: published ranges cluster around — not "the benchmark is". No study of someone else's customers, run with someone else's methodology, produces a number your score can be precisely compared to.

Where do these numbers come from, and why do they vary?

Industry benchmarks come from two fundamentally different pipelines, and the gap between them is the first thing to check before trusting any number.

Panel studies — the Bain/Satmetrix lineage, and what research firms like Qualtrics XM Institute publish — survey the general population about brands they use. That includes people who barely interact with the brand, people locked into contracts, and people whose last touchpoint was years ago. Vendor aggregates — the Retently model — pool the self-run surveys of the vendor's own customers, who are surveying their own engaged users, usually in-app or by email shortly after meaningful interactions. The same company can score 20 points apart across the two methods, with the vendor-aggregate number almost always higher.

Three further sources of variance compound this:

  • Relationship vs transactional surveys. A quarterly "how do you feel about us overall?" relationship NPS and a "how was this support ticket?" transactional NPS produce different numbers for the same company in the same month. Benchmark tables rarely say which they aggregated.
  • Survey channel. In-app prompts reach active users; email reaches everyone including the disengaged; phone surveys add an interviewer-politeness bias that inflates scores. Channel alone can move a score double digits.
  • Response bias. Low-response-rate surveys over-represent the emotional extremes — the delighted and the furious. A benchmark built from 5%-response-rate surveys is measuring a different population than your 40%-response-rate in-app prompt, which is one reason response rate is worth improving before benchmarking anything.

So the ranges in the table are honest as ranges — directionally real, repeatedly replicated across studies — and dishonest the moment anyone reads a single integer out of them as a target. When you cite a benchmark, cite its provenance too: "Retently's vendor-aggregate puts B2B SaaS around the high 30s" is a checkable sentence; "the industry average is +38" is laundered precision.

What counts as a good NPS overall, regardless of industry?

If you need a cross-industry yardstick, the most widely cited one comes from Bain's own framing: anything above 0 means you have more promoters than detractors, above +20 is generally considered favourable, above +50 is excellent, and above +80 is world class — a tier that almost nobody sustains at scale, whatever their case studies say. Qualtrics XM Institute's cross-industry panels typically put the all-industry median somewhere in the +20s to +30s, which is consistent with that framing.

The reason this generic scale exists alongside industry tables is that both answer different questions. "Above +50 is excellent" tells you about the absolute health of your promoter-to-detractor balance. The industry table tells you whether your category makes that balance easy or hard to achieve. A telecom at +25 is doing something remarkable; a boutique fitness brand at +25 has a problem the generic scale would call "favourable". Always read your number against both — and then, as the rest of this post argues, mostly against your own history.

What explains the spread between industries?

Two structural patterns explain most of it. First, chosen products outscore endured ones — people promote things they picked (boutique fitness, their agency, a SaaS tool they championed) and punish things they feel locked into (telecom contracts, insurance, the cable company). Second, low-friction categories outscore high-friction ones, because every billing dispute, support queue, and claims process manufactures detractors regardless of core product quality.

That is why the table's ordering is so stable across years and across publishers even when the absolute numbers wobble: retail above banking, banking above insurance, insurance above telecom. The ordering is structural; the integers are methodology.

The practical implication for multi-product companies: benchmark each line of business against its own category, never against the company aggregate. A bank running both a beloved budgeting app and a mortgage-servicing arm should expect a 25-point spread between them — and treating the blended score as one number hides exactly the comparison that would have been informative.

Why do published NPS benchmarks mislead?

Four reasons to hold every benchmark — including the ranges above — loosely:

Methodology drives the number more than performance does. Survey channel (in-app vs email vs phone), timing (post-purchase vs random), scale presentation, and even question wording shift scores materially. An in-app survey of active users and an email panel of all signups are measuring different populations, and neither matches the methodology behind whatever benchmark you found.

Response bias is unevenly distributed. Low-response-rate surveys over-represent the emotional extremes. A company with a 5% response rate and a company with a 40% response rate cannot be compared on score, even in the same industry.

Sample composition is invisible. B2B scores swing on who answers — economic buyers score differently from daily users. A benchmark never tells you whose voice it aggregates.

Survivorship and selection effects. Companies publicise good scores and quietly shelve bad ones. Vendor-published aggregates skew toward companies invested enough in customer experience to run NPS programs at all. Public benchmark lists are a highlight reel, not a census.

The honest use of a benchmark: a sanity check with ±15 points of fog around it. If your SaaS scores −10, something real is wrong. If it scores +34 against a "+38 industry average", you have learned approximately nothing.

Why does your trend beat any absolute number?

Here is the reframe that makes NPS useful: your own score, measured the same way over time, has none of the comparability problems above. Same methodology, same channel, same population definition, same wording. Every confound that wrecks cross-company comparison is held constant.

That means a 10-point movement in your own trend is signal, while a 10-point gap to a published benchmark is mostly noise. The practical program:

  • Fix your methodology and never silently change it. If you alter the channel, timing, or wording, annotate the change on the chart — you have created a new baseline.
  • Plot monthly, segment relentlessly. By plan tier, tenure cohort, and acquisition channel. The aggregate number hides everything interesting; an aggregate "stable +35" can conceal new-cohort decline masked by loyal-customer survivorship.
  • Alert on movement, not level. A 10-point drop in any segment, sustained across two periods, is worth interrupting your week for. The absolute level is worth a quarterly glance.
  • Mind your sample size before reading anything. Below roughly 50 responses per period the score is noise — with 100 responses, normal sampling variation alone moves the score several points between identical periods.

And keep the follow-up question and closed-loop response at the centre of the program. The verbatims tell you why the trend moved, which is the part a benchmark could never do.

How should you actually use industry benchmarks?

The three legitimate uses, in descending order of value:

  1. Calibrating expectations when you launch a program. Knowing telecom hovers near zero stops a new team from panicking at +12 — or celebrating it in SaaS.
  2. Board and investor communication. "We're at +42 against a category that typically runs +30 to +40, and trending up 6 points year over year" is an honest sentence. Note that the trend clause is doing most of the work.
  3. Competitive context — only with identical methodology. If a research firm measured you and three competitors in the same panel study, that comparison is real. Comparing your in-app score to a competitor's press-released number is not — you are comparing your methodology to their marketing department.

The illegitimate use: setting a benchmark-derived score as a team target. Targets make the metric gameable — reps fishing for 10s, surveys timed after wins — and a gamed NPS is worse than no NPS, because it launders bad news into a green dashboard.

The one-paragraph summary

Industry NPS ranges are real but wide: SaaS roughly +30 to +40, retail +40 to +50, banking +20 to +35, telecom near zero — with the spread explained by whether customers chose the product and how much friction the category imposes. Published benchmarks carry methodology, response-bias, and survivorship distortions large enough to swallow most real differences. Benchmark once for calibration, then spend your attention on your own trend, segmented, measured consistently, with a closed feedback loop. That is the version of NPS that changes what a team builds.

Ready to start measuring? Collect NPS with Formspring surveys — branching follow-ups, consistent methodology, and segmentable results, because your own trend is what matters.

Related from this desk