Sample sizes

When using small sample sizes to understand your customer perspectives clearly every negative comment counts and you’ll want to be informed to be able to follow up and address.

To extract overall themes with a high degree of confidence that these themes are representative of the views of all your customers we are looking for high degree of agreement in responses, i.e. low variability in the nature of responses. The more there is variability in the responses the more customers we need to complete the survey to have a stable estimate of results that we can have a high degree of confidence in.

Some rules of thumb that we use for this number of responses are:

< 10 : Mostly anecdotal, not statistically meaningful for all customers but of course every individual is important, and this is especially true to address any single complaint

10–30 : Tentative themes and insights here; statistically there is still a high margin of error; that said, these numbers are okay for exploratory research

30–50 : The industry acknowledged minimal threshold for confidence in patterns and applying basic inferential statistics

>100 : Statistically meaningful for most purposes, even without knowing the total population

> 300 : Ideal for stable probability estimates and generalizations