Ask detractors what happened
Not “how can we improve”, which invites a wishlist. Ask what went wrong, in their words. It is the shortest route from a number to a fix, and the one answer worth putting in front of a human the same day.
TemplateThe NPS survey, cloned onto your account in one click
NPS survey software with unlimited responses
Ask the 0–10 question on its own screen, branch the follow-up by score, and read every answer. NPS needs volume to mean anything, and the Upform free plan has no response cap at all.
Unlimited responses on the free plan. Conditional logic and all 21 question types included.
0–10
The scale, on one screen
±100
The range a score can land in
∞
Responses on the free plan
$0
To run NPS every quarter, forever
The definition
What NPS actually measures
Before the software, the arithmetic. It takes two minutes to learn and most teams still get the denominator wrong.
Net Promoter Score is a single number between −100 and +100 that summarizes one question: “How likely are you to recommend us to a friend or colleague?”, answered on a 0 to 10 scale. Fred Reichheld introduced it in a 2003 Harvard Business Review article, and it became the default loyalty metric largely because it is short enough that people answer it.
Every answer falls into one of three groups, and the boundaries are not negotiable if your number is to mean what everyone else's means.
Promoters · 9–10
People who would put their own name next to yours. They renew and they refer, and their reasons are worth writing down word for word.
Passives · 7–8
Satisfied and completely unattached. They count toward your total but not toward your score, which is deliberate: a 7 is not half a promoter, it is somebody who would switch for a discount.
Detractors · 0–6
Everything from mildly disappointed to actively warning people off. A 6 counts as a detractor, which surprises everyone the first time, and it is the most useful strictness in the method.
The formula, and a worked example NPS = the percentage of promoters minus the percentage of detractors. Passives sit in the denominator and are then ignored. So if 200 people answer and you get 110 promoters (55%), 50 passives (25%) and 40 detractors (20%), your score is 55 − 20 = +35. It is expressed as a whole number, never a percentage: a score of +35 is not 35% of anything.
Net Promoter, NPS and Net Promoter Score are registered trademarks of Bain & Company, Satmetrix Systems and Fred Reichheld.
The question
Eleven options, one screen, four seconds
The NPS question is the easiest survey in the world to build and the easiest to ruin. Put all eleven points on one screen, label both ends, and ask nothing else until the score is in.
- Use the Slider Scale question type with a minimum of 0, a maximum of 10 and a step of 1. The Rating type counts from 1, and the original wording needs the 0 anchor.
- Label the ends in the help text and leave the middle bare. Words on every point turn a numeric scale into a vocabulary test.
- One question per screen, so nobody sees the follow-up before committing to a number. A score anchored by the question sitting under it is not the same score.
- On a phone every target is full width with a sticky back and next bar; on a laptop Enter moves on and the arrow keys move across the scale.
1 →
How likely are you to recommend us to a friend or colleague?
0 is not at all likely, 10 is extremely likely.
Next →press Enter ↵
The 0–10 screen as respondents see it: the Slider Scale question type from 0 to 10 in steps of 1, and a two-step form so the follow-up is the only thing left.
Conditional logic
The follow-up question is the whole point
The score tells you the size of the problem. The sentence underneath tells you what it is. Ask everyone for that sentence, but never ask all three groups the same thing.
Show field
Conditions
Recommend scoreLess or equal6
ActionShow fieldWhat went wrong for you?
Rule Summary
When Recommend score less equal "6" → Show field: What went wrong for you?
Show field
Conditions
Recommend scoreBetween7,8
ActionShow fieldWhat would move that up by two?
Rule Summary
When Recommend score between "7,8" → Show field: What would move that up by two?
Show field
Conditions
Recommend scoreGreater or equal9
ActionShow fieldWhat should we never change?
Rule Summary
When Recommend score greater equal "9" → Show field: What should we never change?
Three rules, and the survey is still two questions long Each rule fires on the score and shows exactly one open-text step, so nobody answers a question written for somebody else. Upform's conditional logic is a flat list of rules, each an AND or OR group of conditions, with 20 operators and four actions: show, hide, jump and end form. The builder reads every rule back to you as a plain sentence.
Ask passives for the gap
The most ignored and most movable group on your list. “What would move that up by two?” gets a concrete answer far more often than “why that score”, because it asks for a change rather than a justification.
Ask promoters what to protect
Promoters get asked for referrals and almost never get asked what they value. Their answers are the words to put on your site, and the list of things not to redesign next quarter.
Timing
When to send it
Most bad NPS data is not badly worded, it is badly timed. Five decisions settle it, and once they are written down you never argue about them again.
Decide relationship or transactional
Relationship NPS runs on a cycle and measures the whole relationship. Transactional NPS follows one interaction and measures that. Pick one per form, or your average is of two different questions.
Wait until there is something to judge
Thirty to ninety days in is the usual window. Ask in week one and you are measuring your onboarding email, not your product.
Put it on a cycle, not on a whim
Quarterly is standard, and the interval matters less than never moving it. A score from a good month compared to one taken after an outage tells you about your calendar, not your customers.
Rotate the sample, and leave people alone
Send to a slice of your list each quarter, and leave 90 days before the same person sees it again. Fatigue shows up as a falling response rate long before it shows up in the score.
Match the channel to the moment
A public link by email for relationship NPS, an embedded form on a confirmation page for transactional. Same form either way, so the completion rate tells you which channel actually works.
Being straight with you
Reading the number honestly
NPS is a good tracking device and a terrible strategy. Three things nobody tells you when they sell you the dashboard.
Sample size decides what you are allowed to say
NPS is the average of a value that is +1 for a promoter, 0 for a passive and −1 for a detractor, so it carries an ordinary margin of error that shrinks with the square root of your sample.
At about 100 responses the 95% band sits roughly 15 points either side of your score; at 400, closer to 8. A move from +31 to +36 on 90 responses is not an improvement, it is Tuesday.
The people who answer are not your customers
They are your customers who answer surveys, which is a different population. Delight and fury both raise the odds of a reply; ordinary contentment does not. An in-product prompt after a real action reaches a broader slice than email ever will.
Watch the response rate as closely as the score. If it halves, your sample changed even when the number did not.
One number is not a strategy
A score cannot tell you what to do. Two very different companies land on +35: one with a calm, unremarkable middle, one with a devoted core and a loud unhappy edge. The distribution matters more than the average.
And once the score is attached to somebody's bonus, you start collecting scores rather than answers. Report the trend, act on the comments.
The practical version of all three Do not publish a score under 100 responses. Report it with the response count and the date range beside it, every time. Compare it only against your own last reading, collected the same way. And when it moves, find out which of the three groups moved before you tell anyone why.
It is also why a capped free plan quietly breaks NPS. Typeform's free plan is widely reported to stop at 10 responses a month, an order of magnitude short of a readable score. Upform caps responses on no plan at all.
Competitor plan limits checked September 2026 on their public pricing pages; Upform's on our own pricing page. Plans change, so verify before you decide.
After the send
Where the scores go
Upform collects and exports the raw 0–10 answers. It has no scoring engine, so the subtraction happens where you already keep your reporting, which is also where it can be checked.
Google SheetsEvery score appended as a new row, ready to bucket
SlackA message per response, so a 2 never waits a week
HubSpotCreates or updates the contact behind the scoreEvery integration is per-user OAuth, so you connect your own account and the data lands in your workspace. All of them are on the free plan. See all integrations.
The arithmetic
Four columns and one subtraction
Computing NPS in a spreadsheet takes a minute and leaves an audit trail, which matters the first time somebody asks where the number came from.
- Each response arrives as one row: timestamp, score, the open answer and whatever else you asked.
- Add one column that turns the score into promoter, passive or detractor. That is the only formula anyone has to check.
- Count each group, divide by the total, subtract detractors from promoters. Keep the response count next to the score.
- Prefer to work offline? The response table filters by date and exports to CSV.
Q3 relationship NPS · 200 responses
55%
Promoters, 9–10
25%
Passives, 7–8
20%
Detractors, 0–6
+35
Score, 55 − 20
- At 200 responses the 95% band is roughly ±11 points, so report this as “+35, n=200” and compare it only to your own last quarter.
Example figures, shown to illustrate how the groups add up.
Templates
Start from a survey that is already wired up
Preview any of these live, clone it in one click, then change the wording to sound like you.
How likely are you to recommend us?
A9 · Very likely
B6 · Maybe
Next →NPS survey template
The 0–10 question and three branched follow-ups, with the logic rules from this page already built.
Feedback · FreePreviewUse Template
How satisfied were you today?
AVery satisfied
BSatisfied
Next →Customer satisfaction survey
A 1–5 rating for a single interaction, when NPS is the wrong instrument for the moment.
Feedback · FreePreviewUse Template
What's the main reason you're leaving?
APrice
BA missing feature
Next →Exit survey
Where your detractors end up if nobody reads the follow-up. Five reasons and one open box.
Feedback · FreePreviewUse Template
Frequently asked questions
What is a good NPS score?
Any score above 0 means you have more promoters than detractors, and that is the only benchmark that travels cleanly between industries. Above +30 is usually treated as strong and above +50 as excellent, but those bands come from surveys that mix wildly different businesses, audiences and question wordings. The number actually worth watching is your own, measured the same way every quarter, with the same audience and the same words.
How many responses do I need for an NPS survey?
Enough that the score moves less than the change you are trying to detect. NPS is the average of a value that is +1 for a promoter, 0 for a passive and -1 for a detractor, so its margin of error shrinks with the square root of your sample. At around 100 responses the 95% band sits roughly 15 points either side of your score; at 400 it is closer to 8. Under 100, treat the number as a conversation starter and read the comments instead.
Is Upform's NPS survey free?
Yes, and free here means permanent rather than a trial. The free plan includes unlimited forms, unlimited responses, all 21 question types, conditional logic and every integration, so you can run NPS every quarter for years without paying anything. Pro is $12 a month and only adds unlimited file storage, branding removal and up to five team members.
Does Upform calculate the NPS score for me?
No. Upform has no scoring or calculation engine, so it collects and exports the raw 0 to 10 answers rather than turning them into a score. In practice that is one subtraction: send responses to Google Sheets, bucket each answer into promoter, passive or detractor, then subtract the detractor percentage from the promoter percentage. Doing the math in a sheet also makes the number auditable, which matters the first time someone questions it.
How often should I send an NPS survey?
Quarterly is the usual cycle for relationship NPS, and it is frequent enough to see a trend without wearing people out. Leave at least 90 days between sends to the same person and rotate a sample rather than mailing your whole list at once. Transactional NPS is different: it follows the interaction it is measuring, not the calendar.
Should I ask the follow-up question of everyone or only detractors?
Ask everyone, but do not ask everyone the same thing. Detractors get an open box about what went wrong, passives get asked what would move them up, and promoters get asked what you should keep doing. Conditional logic branches on the score, so each group sees exactly one follow-up and the survey stays two questions long for everybody.
What is the difference between relationship NPS and transactional NPS?
Relationship NPS asks how likely someone is to recommend you overall, on a fixed cycle, and tells you whether the relationship is holding. Transactional NPS asks the same question straight after a specific interaction, such as a closed support ticket or an onboarding call, and tells you about that interaction. They use the same question and the same math, so keep them as separate forms or the two averages will contaminate each other.
Get started
Send the 0–10 question this afternoon
Two questions, three logic rules, one link. Every response is yours, and no cap stops the survey halfway through your sample.
No credit card. Unlimited forms and responses on the free plan.
Keep exploring
Survey templates
NPS survey templateCustomer satisfaction survey templateProduct feedback formExit survey templateSurvey templatesSurvey guides
Rating scale questionsLikert scale explainedSurvey questions to askSurvey design principlesHow to make a surveyProduct
Conditional logicForm analytics and drop-offConversational formsGoogle Sheets integrationSlack integrationSolutions and alternatives
Customer feedback surveysProduct research surveysSurveyMonkey alternativeTypeform alternative