There is a particular kind of survey result — from an NPS survey or any other — that tells you nothing: 400 responses, an average satisfaction score of 4.2 out of 5, and a slide in the quarterly deck saying customers are happy. Nobody does anything differently. Next quarter it is 4.1 and nobody does anything differently about that either.
Customer feedback is only useful when it changes a decision. That means asking a question people will answer honestly, at a moment when they remember the experience, and having a route from the answer to someone who can act on it. NPS is the most common vehicle for this and also the most commonly misused. This guide covers running it properly — plus the two other surveys that often matter more.
The NPS survey: what it measures and what it does not
Net Promoter Score asks one question: how likely are you to recommend us to a friend or colleague, on a scale of 0 to 10.
Responses split into three groups:
- Promoters (9–10) — enthusiastic, likely to refer.
- Passives (7–8) — satisfied, unattached, easily poached.
- Detractors (0–6) — unhappy, and a churn risk.
The score is the percentage of promoters minus the percentage of detractors, producing a number between −100 and +100. Passives count for nothing, which is deliberate and also the source of most confusion about why the number moves oddly.
What it is good for: tracking a relationship trend over time, flagging specific unhappy customers for follow-up, and — through the follow-up question — generating the qualitative material that actually tells you what to fix.
What it is not good for: benchmarking against other companies. Cross-industry comparisons are close to meaningless. Response scales are culturally variable, sampling differs wildly, and a "good" NPS in enterprise software is a poor one in consumer retail. Compare yourself to yourself.
The score is the least valuable output of an NPS survey. The comment box underneath it is the point.
The follow-up question does the work
A score alone tells you a temperature, not a cause. One open question afterwards changes the exercise entirely.
Better than "why did you give that score?" — which produces short, unhelpful answers — is a question tuned to the group:
- To promoters: "What would you tell a colleague who asked about us?" You get your own marketing language, written by customers. This is where the best testimonial quotes come from.
- To passives: "What would make this a 9 or 10?" The single most actionable question in the whole survey, because it produces a roadmap ranked by the people who nearly love you.
- To detractors: "What went wrong?" Then, crucially, someone replies.
The two surveys that often matter more
NPS measures the relationship. These two measure the experience, and they are usually faster to act on.
CSAT — customer satisfaction
"How satisfied were you with [specific thing]?" Asked immediately after a defined interaction: a support conversation, an onboarding call, a delivery.
CSAT is transactional and narrow, which is its strength. It tells you whether a particular process is working, and because it is asked at the moment, recall is accurate. If support CSAT drops in a week, you know where to look.
CES — customer effort score
"How easy was it to [do the thing]?" Effort turns out to predict loyalty better than delight in most service contexts. People rarely leave because you did not amaze them; they leave because something was harder than it should have been.
CES is particularly good on signup flows, first-time setup, returns and cancellations. A high-effort onboarding shows up as churn two months later, and CES catches it in week one.
Finally know what’s actually working.
Track reach, engagement, and ROI across every channel in one dashboard. 15-day free trial, no card required.
See your analytics →Asking at the right moment
Timing does more for response quality than any wording change.
- NPS: 30 to 90 days after onboarding, then on a regular cycle — quarterly for high-touch, twice yearly otherwise. Too early and they are rating the sales process; too late and you have missed the churn window.
- CSAT: immediately after the interaction. Within the hour, ideally in the same channel it happened in.
- CES: right after the task, whether it went well or not.
Two timing mistakes worth avoiding. First, surveying only happy moments — asking for NPS right after a successful purchase inflates the score and teaches you nothing. Second, surveying the same people repeatedly; set a suppression window of at least 90 days so your most engaged customers are not your only data source.
Getting responses without begging
Response rates vary a lot by channel and audience, and in-product surveys usually get more answers than emailed ones. Things that move the number:
- One question on screen. The visible length of the survey is what determines abandonment. A single question with an optional comment beats a five-question form every time.
- In-product beats email for anything about the product itself. The person is already in the context you are asking about.
- No incentive for NPS. A prize draw attracts people who want the prize and skews the result. Incentivise long-form research interviews, not score surveys.
- Say what you did last time. "Last quarter you told us search was slow; here is what changed." Response rates climb when people have evidence their answers went somewhere.
- Do not ask in the first week. They have no basis to answer, and you will get noise.
From answers to changes
This is where most feedback programmes quietly fail. Four habits fix it.
Close the loop on detractors within 48 hours. A person, not an automated apology. This single practice does more for retention than any change to the survey itself — a detractor who gets a thoughtful reply frequently becomes a promoter, and they will tell you things in that conversation they would never write in a box.
Tag the comments. Read them and code them into themes: pricing, performance, missing feature, onboarding confusion, support quality. Without tagging you have anecdotes; with it you have counts, and counts win arguments about priorities.
Route themes to owners with a number attached. "Nineteen of the last hundred detractor comments mention the mobile app" is a product conversation. "Customers say the app is bad" is a complaint.
Report the change, not the score. In the quarterly review, lead with what you changed because of feedback and what happened next. The score is a footnote. This is what stops the programme becoming a slide nobody reads.
Plan, approve, publish — together.
A shared calendar, approvals, and roles so your whole team ships on time. 15-day free trial, no card required.
Try Skyfliq free →Reading the score without fooling yourself
Four things that make NPS numbers look like they mean more than they do.
Small samples swing wildly. With 40 responses, three detractors moving to passive shifts the score by roughly 8 points. That is noise, not a trend. Below about 100 responses per period, watch the direction over several quarters rather than reacting to each reading.
Who you surveyed decides the answer. Survey everyone and you get a relationship score. Survey only people who opened a ticket and you get a support score with an NPS label on it. Keep the sampling rule identical between periods or the comparison is meaningless.
Non-response is data. If your response rate falls from 22% to 9%, the score may look stable while the underlying picture worsens — disengaged customers stop answering before they leave. Track response rate next to the score, always.
Segment before you conclude. A flat overall score frequently hides a rising number from one group and a falling one from another. Split by plan, tenure and acquisition channel. The most useful finding in most NPS datasets is that one segment is quietly unhappy while the average looks fine.
Where the answers should live
Feedback that sits in a survey tool is feedback nobody uses. The score and the comments belong on the customer record in your CRM, where they are visible to whoever next talks to that person.
The practical difference is immediate: a salesperson opening an account before a renewal call should see that this customer gave a 4 and complained about onboarding in March. If that sits in a separate dashboard nobody opens, the call goes badly for reasons that were entirely knowable.
This is the same argument as first-party data generally — the value appears when the information is in the place where decisions get made, not in the tool that collected it.
A programme you can actually sustain
Start smaller than you think:
- Pick one survey. If you have nothing, start with CSAT after support — the fastest feedback loop and the easiest to act on. Add NPS once that habit exists.
- One question plus one open box. Nothing else.
- Set the suppression window so nobody gets surveyed more than quarterly.
- Assign the loop-closing to a named person with time for it. This is the step that gets skipped and it is the one that matters.
- Tag comments monthly and bring the top three themes to whoever owns the roadmap.
- Publish what changed — to customers, in the changelog or a newsletter. It is the cheapest trust-building exercise available.
Turn followers into customers.
Capture leads, nurture them by email, and close — all where you already post. Free for 15 days, no card required.
Start free →The honest limitations
Surveys over-represent the strongly opinionated at both ends. The quietly-considering-leaving majority often does not respond at all, which is why an NPS trend should never be your only churn signal — usage data catches the silent ones.
They also measure stated intent rather than behaviour. Someone scoring 9 is not guaranteed to refer anyone; if you want referrals, build a referral programme rather than assuming promoters will act.
And they cannot tell you about the people who never became customers. The most valuable feedback in most businesses is from prospects who chose something else, and no NPS survey will ever reach them.
The short version
Ask one question at a moment people remember, always with an open box. Reply personally to anyone unhappy within two days. Tag the comments so themes become counts. Report what changed rather than what the number was.
Do that for a year and you will have something more useful than any score: a documented record of what your customers asked for and what you did about it.