Research Sample Size: Segmentation and Data Quality

If you are planning a research project that involves breaking down respondents into segments, as we do with our RealPersonas™, then you’ll want to sample enough people that will allow you to work with any major segments that exist and emerge in the population.

This is important because as we break down the group into smaller segments (and thus smaller groups of people), the confidence intervals for each segment will become larger . So if you’re doing a segmentation study, you want to get responses from as many people as you can afford.

As an example, suppose that you’re studying how supplement users feel about your brand direction.

You know that there’s a segment of supplement users out there who spend a lot on supplements in general and your brand specifically,  and you want to see what these high-value consumers feel about your proposed branding direction.

If you sample only 150 people, you might capture 10-15 of these high-value consumers – hardly a large enough group to say much about, especially since the confidence intervals for a group of this size would be around +/- 30%! However, if you go out and sample 1,000 people, you might find 60-100 of these high-value supplement users, which enables you to have higher certainty about how they feel as a group about your brand direction.

This can be important for a brand, since decisions about which segment to target and how to do so depend on accurately understanding the segment’s value and its preferences in the first place.

Sometimes even when you can’t get a large sample size due to budget or logistical restrictions, a small group of people falls out naturally through segmentation as being very different from the rest of the group, and it’s worth looking at them anyway. But starting out with a large enough sample size helps to guard against low confidence levels by getting enough respondents that each major segment is likely to be adequately represented.

As a rule of thumb, we try to sample at least 100 people for each RealPersona™ segment that we anticipate will be present, and if time and costs allow, go out to 1,000 or more people, which enables us to have 3-6 comfortable segments emerge when performing segmentation analysis. This gives us enough data to pick up on any small but important segments.