Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

How to improve accuracy in a MaxDiff exercise for a sample of 30-40 people?


We do like to like to perform a MaxDiff exercise with a sample of 30-40 people. There are 11 choices and we understand that most MaxDiff surveys include more than 100 respondents.

What should we do about our survey design to make this MaxDiff survey more accurate with 11 choices & 40 respondents?

Thank you for your support.
Best regards,
asked Jun 20 by cemerdemli (150 points)

1 Answer

0 votes
Best answer
Actually, MaxDiff was originally designed to be able to work for individual respondents, so it scales down very nicely.  We can get reliable utilities at the respondent level if we show each item say 4 times per respondent.  So if you show 11 quads (sets of 4 items each) then each respondent will see each of the items 4 times.

Now, there's still the problem of generalizability (statistical inference):  with a small sample there's an amount of sampling error that you're just not going to remedy but for increasing sample size.  

But as far as your 40 respondents go, you'll be measuring their utilities very well.
answered Jun 20 by Keith Chrzan Platinum Sawtooth Software, Inc. (111,275 points)
selected Jun 20 by cemerdemli
Thank you for your response.