Determining the ideal sample size
You want to set up a NPS survey, but you are not sure how many respondents you need? The number of respondents needed to get reliable results can be calculated based on three factors:
How sure do you want to be in the results, i.e. your preferred confidence level.
For example, when a decision can have fatal consequences (for instance while testing a new drug) you want a very high level of confidence in the results. When you are only exploring about rough designs or feelings without much impact of being wrong, a confidence level of 80% may be sufficient.
The confidence level refers to the percentage of all possible samples that can be expected to include the true population parameter. A 95% confidence level implies that 95% of the confidence intervals would include the true population parameter. The higher the confidence level, the less likely it is that the sample statistic is not reflecting the true population parameter.
How precise your answers need to be, i.e. your preferred confidence interval.
For example, a final poll on the day before a tight-running election would ask for a smaller confidence interval in order to predict the result with the most precision.
The confidence interval is the figure indicating the degree of uncertainty associated with a sample statistic. The confidence interval consists of a range of values with an acceptable margin of error around an unknown population parameter.
How large is the group representing your potential customers, i.e. your population size (aka universe).
For example, a sample of 50 respondents might be more relevant in a population of 75 than in a population of 100000. Of course in reality also cost considerations will be a decisive factor, as larger samples cost more to survey.
The population size is the magnitude of the group that represents your sample. This can be the number of people who have bought a specific product, or all potential clients for a service.