On this page
NPS and CSAT measure two different things. CSAT asks how a specific interaction went and is scored as the percentage of people who answered positively. NPS asks how likely someone is to recommend you and is scored by subtracting the percentage of detractors from the percentage of promoters, which produces a number between -100 and +100. Use CSAT to check whether a single moment worked, and NPS to check whether the relationship is worth anything. For a business of one the practical answer is usually CSAT first, because it fires after a real event, needs about twenty responses to be readable, and tells you which interaction to fix. NPS needs a larger base before the number stops moving at random. The calculator below scores both, and the benchmark table says what counts as good.
Both metrics get described as measuring customer happiness, which is why people ask which one is better. They are not competing versions of the same measurement. They answer different questions, on different timescales, and the wrong choice gives you a number that moves without telling you anything.
Here is the difference in one line: CSAT measures a moment. NPS measures a relationship.
NPS vs CSAT: The Differences Side by Side
| CSAT | NPS | |
|---|---|---|
| The question | How satisfied were you with [this interaction]? | How likely are you to recommend us to a friend or colleague? |
| Scale | Usually 1 to 5, sometimes 1 to 7 | 0 to 10 |
| What it produces | A percentage, 0 to 100% | A score from -100 to +100 |
| How it is calculated | Positive responses divided by total responses | % promoters minus % detractors |
| When you send it | Straight after a specific event | Periodically, or after a relationship matures |
| What it measures | Whether that one moment worked | Whether the overall relationship has value |
| Responses needed to read it | Around 20 for a rough signal | Considerably more, because the middle is discarded |
| Moves when | You fix or break a specific process | The whole experience shifts, slowly |
| Main weakness | Says nothing about loyalty or repeat business | Throws away everyone who scored 7 or 8 |
The row that matters most for a small business is the second to last one. CSAT reacts to things you can actually change this month. NPS is a slow number, and a slow number on a small base is mostly noise.
NPS and CSAT Calculator
Enter your response counts to score both. For CSAT, positive responses are the top two boxes on a five point scale, meaning the 4s and 5s.
CSAT
NPS
What Counts as a Good Score
This is the question everyone asks second, and the honest answer is that both scores are only meaningful against your own previous number. An NPS of 30 is unremarkable in one industry and excellent in another. That said, the broad conventions are worth knowing so you can tell a disaster from a normal week.
| Band | CSAT | NPS | What it usually means |
|---|---|---|---|
| Strong | 80% and above | +50 and above | The thing is working. Do not redesign it. |
| Workable | 60 to 79% | 0 to +49 | Normal. Improvement comes from fixing specific moments, not from a strategy change. |
| Problem | Below 60% | Below 0 | More people are unhappy than happy. Find the interaction causing it before doing anything else. |
A detail that catches people out: because NPS discards everyone who scored 7 or 8, a score of 0 does not mean nobody liked you. It means promoters and detractors cancelled out. You can have a perfectly reasonable business sitting at 0 with most respondents parked in the passive band.
Aziz's take: The benchmark question is mostly a distraction. What you want is the same number measured the same way three months apart, because the direction is the only part you control. I would rather see a CSAT of 64% that was 55% last quarter than a one-off 80% with nothing to compare it against. Chasing somebody else's benchmark is how you end up redesigning something that was fine.
Which One Should You Actually Use
For most businesses of one, the answer is CSAT first, and NPS later or never. Three reasons.
CSAT works at a small scale. Around twenty responses gives you a readable signal. NPS discards the passive middle before scoring, so a small sample swings wildly on single answers. If ten people reply and one moves from a 9 to a 6, the score drops by twenty points and nothing has actually changed about your business.
CSAT points at something fixable. It attaches to a specific interaction, so a low score tells you which moment to look at. A low NPS tells you the relationship is weak without telling you where. That is a fine problem for a company with a research team and a bad one for a person with an afternoon.
NPS needs a relationship to measure. If you deliver a project and part ways, there often is not one yet. NPS earns its place when there are repeat customers or a subscription, because then the question of whether people stay actually has an answer.
Run NPS when you have recurring customers, enough of them to survive the discarded middle, and a reason to track loyalty as a trend rather than a moment. Until then it is a number that moves for reasons you cannot trace. If the reason you are reaching for a loyalty metric is that people keep leaving, the underlying problem is usually better attacked directly through the reasons customers churn than through a score that describes the symptom.
Running Either One Without a Research Budget
The mechanics are simpler than the tooling around them suggests.
- Pick the trigger. For CSAT, send it right after the event you want to measure, meaning project delivery, a support reply, or an onboarding call. For NPS, send it on a fixed cycle rather than after an interaction, or it becomes a slow CSAT.
- Ask one question. Adding a second question reliably costs you responses. Ask the score, then a single optional free text box.
- Make the free text box the point. The number tells you something is wrong. The sentence someone types tells you what. On a small base the comments are worth more than the score.
- Keep the wording identical every time. Changing the scale or the phrasing resets your history and you lose the only comparison that matters.
- Record the count, not just the percentage. A 100% CSAT on three responses and an 82% on forty are different facts, and in six months you will not remember which was which.
Whatever you learn from either metric belongs in the process it points at. If CSAT drops on projects that started badly, the fix is usually further upstream in how the work is set up rather than in the delivery itself.
Five Mistakes That Make the Number Meaningless
Tracking both from day one. Two metrics on a small response base gives you two unreliable numbers instead of one. Start with the one that matches your question.
Reading a score built on under twenty responses. Both metrics need a base before the number stops being an accident. The calculator above flags this because it is the single most common way these get misread.
Sending the survey long after the event. CSAT measures a moment, and people stop remembering the moment quickly. A survey sent a fortnight later measures their general mood.
Comparing your score to a published industry average. Different question wording, different scales, different response rates. Compare against your own last reading instead.
Treating the score as the deliverable. Neither number improves anything. They tell you where to look. If nothing changes after you read one, the survey is costing you goodwill for no return. Satisfaction scores also sit awkwardly next to money: they belong alongside the metrics that actually track the health of the business rather than standing in for them.
Aziz's take: The free text box is the part I would keep if I had to drop everything else. On a small business the score is a rounding error on twenty replies, but one customer writing "I never knew when it would be finished" is a specific, fixable fact. I have changed more about how I work from single sentences in that box than from any score I have calculated.