Customer experience
Net promoter score (NPS) survey
One number you can follow for years, and the sentence that explains what moved it.
Why this survey gets sent
“My manager asks me for the score every quarter. I give him the number, he asks why it went up, and all I can tell him is that people seem happier than they were in March.”
The score itself is arithmetic anybody can copy. What separates a measure a team acts on from a slide nobody reads is the question sitting underneath the number — and the discipline to word it the same way every single time you send it.
What is in it
- How likely are you to recommend us? (0 to 10)
- The main reason for that number (required long answer)
- What would have to change to raise it (long answer)
- How long you have been dealing with us (single response)
- Have you recommended us to anyone lately? (yes / no)
- Email, only if you want a reply (optional)
About 3 min to answer
Net promoter score (NPS) survey
About 3 min
How to write this survey
Keep the scale at 0 to 10
The eleven points are not decoration; they are what makes the arithmetic work. Everyone who picks 9 or 10 is a promoter, 7 and 8 are passives who are counted in the base and nowhere else, and 0 through 6 are detractors. The score is the share of promoters minus the share of detractors, so it runs from -100 to +100 and is a percentage of nothing — write it as a bare number, never with a percent sign beside it.
This is the one place where a shorter scale costs you something real. Compress eleven points into five and you lose the line between a 7 and a 6 — the line the whole measure is built on — and you can no longer compare your number with the one you recorded last year or with anything published anywhere else. Pick the scale once, label only the two ends, and leave the middle as bare digits so nobody has to read nine descriptions to answer one question.
The number is the headline; the reason is the story
Put the open question immediately after the scale and make it the only required text field in the survey. Ask about the number the person has just chosen — “what is the main reason for the number you chose?” — rather than about their feelings in general. A reader who has a specific digit fresh in mind writes a specific sentence; a reader asked for “any comments” writes nothing, or writes a paragraph about something they remembered from last year.
Add a second, optional box that asks what would have to change to raise the number. It is a different question and it produces different answers: the first tells you what happened, the second tells you what to build. Between them they turn a quarterly figure into a short, ranked list of work, which is the only kind of output a score like this has ever justified.
One segmenting question, and only one
The temptation is to bolt on age, city, product line and channel so that the score can be sliced every way at once. Resist it. Each extra question costs you responses, and a score computed from a hundred answers is a rumour. Ask the single thing that changes what you would do with the answer — usually how long the person has been a customer, because a detractor in their first month and a detractor of three years are two different problems with two different owners.
If you already know who you sent the survey to, do not ask them to tell you again. Anything sitting in your own records — the branch, the plan, the date they joined — belongs in the invitation, not in the survey. Every question you delete for that reason is a question someone actually answers.
Same wording, same rhythm, every quarter
A score only means something against the last one, so the moment you rewrite the question you have reset your history without being told. Save the survey, send the identical copy next quarter, and keep a note of the date you sent it — a score collected in a quiet week and a score collected the day after a price change are not comparable, however carefully the question was worded.
Once a quarter is enough for a relationship measure, and more often than that quietly destroys it — people who are asked monthly stop answering, and the ones who keep answering are not a sample of your customers any more. If you want a reading after every single visit or order, that is a different, shorter survey, and it deserves its own page rather than a faster version of this one.
A detractor who left an email is a customer you can still keep
Keep the contact field last and optional, and say plainly what it is for: someone will read this and get back to you. Phrased that way it stops being a demand for personal data and becomes an offer, and the people most likely to accept it are exactly the ones with a low number and a sharp reason — the responses worth a phone call.
Then answer them within the week. The single largest gain most teams get from this survey is not the trend line; it is the handful of people every quarter who said something was wrong, were called back, and stayed. Everyone else is allowed to remain anonymous, and the page should say so above the field rather than leaving it to be guessed.