Syntha Lab RU

NPS confidence interval and significance

NPS is the difference between the shares of promoters and detractors, and it has a spread that depends on the sample size. The calculation gives NPS with a confidence interval and compares two periods: whether the change is significant or just survey noise. It also shows the minimum difference that can be noticed at all at your sample sizes.

Customers and loyalty For whom: Customer experience heads, analysts and marketers who report NPS and its change.

Calculate on your data →

At a glance

What question does it answer?
How uncertain NPS is at your sample and whether its change between periods is significant.
What data is needed?
The shares of promoters and detractors of each period.
How to read the result?
First the intervals: if they overlap and z is small, the change is not proven. Then the minimum noticeable difference: it shows whether a larger sample is needed. Example on the default values: NPS of the first period: 32 pts; Interval of the first period (±): 7.7 pts; NPS of the second period: 38 pts; Interval of the second period (±): 8.2 pts. The figures are illustrative: enter your own to see your result.

Why and how to reason

NPS is a difference of shares with a spread that depends on the number of responses. In the example 32 ± 7.7 at 400 responses and 38 ± 8.2 at 350; z = 1.04, so a rise of 6 points is within the survey noise. The minimum noticeable difference at such samples is 11.3 points.

When to use it

Apply when reporting NPS and when comparing periods or segments.

How to read the result

First the interval and z, then the conclusion; with a small sample increase the number of responses.

Common mistakes

They compare NPS without an interval and work with small samples.

What data is needed?

  • The shares of promoters and detractors of each period.
  • The number of responses in each period.
  • The confidence level (usually 95%).
  • An example: 52% and 20% at 400 responses and 57% and 19% at 350: NPS 32 ± 7.7 and 38 ± 8.2, z = 1.04, the change is not significant (illustrative numbers).

Inputs

FieldUnitValue in the exampleMeaning
Promoters%52The share of respondents giving 9–10 points. The example value is illustrative.
Detractors%20The share of respondents giving 0–6 points. The example value is illustrative.
Responsesresponses400How many people answered in the first period. The example value is illustrative.
Promoters%57The share of respondents giving 9–10 points. The example value is illustrative.
Detractors%19The share of respondents giving 0–6 points. The example value is illustrative.
Responsesresponses350How many people answered in the second period. The example value is illustrative.
Confidence level%95Usually 95%: the interval covers the true value in 95 cases out of 100. The example value is illustrative.

How it is calculated

StepFormula
NPSpromoters − detractors, in percentage points
Spreadσ² = (P + D − (P − D)²) ÷ n, interval NPS ± z × σ
Comparing periodsz = (NPS₂ − NPS₁) ÷ √(σ₁² + σ₂²); significant if |z| is not below the quantile of the confidence level

Worked numbers

Default values from the app; in the interactive version you enter your own.

NPS of the first period
32 pts
Interval of the first period (±)
7.7 pts
NPS of the second period
38 pts
Interval of the second period (±)
8.2 pts
Change in NPS
+6 pts
z statistic
1.04
p-value
29.6 %
Minimum noticeable difference at these samples
11.3 pts
NPS and interval bounds, points
Bar chart: First period 24.3, Second period 29.801020304050First periodFirst periodLower bound: 24.324.3NPS: 3232Upper bound: 39.739.7Second periodSecond periodLower bound: 29.829.8NPS: 3838Upper bound: 46.246.2
  • Lower bound
  • NPS
  • Upper bound
Data as a table
MeasureLower boundNPSUpper bound
First period24.33239.7
Second period29.83846.2

The change is not significant: it is within the survey noise. Do not announce a rise or fall of NPS on such a difference.

NPS does not replace retention and repeat purchases; survey after the order is received and compare the same segments and survey methods.

What to do with the result: report NPS with the interval; for a significant difference enlarge the sample (the survey calculation counts the sample size).

A typical mistake: comparing NPS without an interval and working with small samples.

Nuances and limits

  • Comparing NPS without an interval.
  • Relying on small samples.
  • Taking NPS for a replacement of retention.

Terms used

Calculate on your data →