Conjoint Analysis: How many respondents do I need?

Client asked: What is the minimum sample size for a conjoint/discrete choice study? This is the most common question - nearly universal.

Most people are surprised to learn that the number of attributes and levels in the experiment have little impact on the need for more or less subjects. So what does drive sample size requirements?

Sample requirements depend on a few factors.

• The number of subgroups or segments you have
• The strategic use of the subgroups

You need to start with the subgroups and build your way up to the total sample requirements. How many subgroups do you have? What is the incidence (percent population) of each of them?

When there is going to be a single strategy based on the study results, subgroups play a small role. Subgroups are still included as a way to check appeal among key segments. Cutting the results by segment helps avoid a situation where you find a broadly appealing solution that is hated by you target segment.

When there is a portfolio strategy, subgroups are more important because you are likely to build a distinct product/price strategy for some of the segments.

In a choice experiment, respondents are asked to react to multiple scenarios. When there are 12 choice tasks, the amount of data is 12 times the number of respondents.

The general rule is you need 400 respondents in total when there is no subgroups driving the need for more. This is far from a hard rule. Recently we completed a successful study with only 100 respondents because the audience size was so small.

The next rule is you want 50 respondents per subgroup. Or if you shoot for 50 respondents for the smallest subgroup, the rest will meet the minimum requirements naturally.

If you are looking for the study to inform a differentiated product line and the subgroups are key to crafting the products, you will want 200 respondents per key segment. If you don't have or don't want to rely on predefined segments, use the same 200 rule for each differentiated product that you think will you will have.