Discover-MaxDiff selects items to show across multiple questions according to the following goals: a) each item should appear an equal number of times, b) each item should appear with each other item an equal number of times, and c) each item should appear in each position (top, middle, bottom) of MaxDiff questions an equal number of times. Depending on your list of items and number of questions to show, it may not be possible to achieve perfection on all three goals; but MaxDiff questionnaires that are almost-but-not-quite perfect are still very efficient and you will obtain excellent results in practice. For more information about Discover-MaxDiff designs, please see the white paper at: https://sawtoothsoftware.com/resources/technical-papers/discover-maxdiff-how-and-why-it-differs-from-lighthouse-studios-maxdiff-software.
Number of MaxDiff Questions
For the methods used in Discover-MaxDiff, if respondents can give you the time, we recommend you ask enough sets (questions) so that each item appears three or more times per respondent.
Note for advanced users: There are instances where you might purposefully field a MaxDiff study where each item appears less than two times per respondent, because you decide that you don't need high-resolution individual-level scores, but rather plan to conduct your analysis at the segment or population level via aggregate logit or latent class estimation. Discover’s MaxDiff analysis capabilities do not include aggregate logit or latent class estimation. But, you can export the MaxDiff data from Discover-MaxDiff to run logit or latent class estimation using Sawtooth Software’s external standalone latent class software tool (under Data & Analysis choose More | Advanced Export ).
Number of Items per Question
For most situations, we recommend displaying three to five items at a time (per question) in MaxDiff questionnaires. Showing more than about five items at a time can lead to greater respondent fatigue and response error. The software will not let you display more than half as many items as there are items in your study as this reduces the precision of your MaxDiff scores. For example, if your study has just eight total items, you cannot display more than four items per set.
In the vast majority of cases, researchers will show the items in randomized position order within each MaxDiff question. That way, each item has an equal chance of appearing at the top, middle, or bottom of the question. This reduces potential order bias that could alter the utility scores.
On rare occasion (advanced applications), researchers may wish to show the items in the order they are listed on the Items tab. For example, best-worst conjoint (BW Case 2), where the items are mutually-exclusive conjoint levels. With best-worst conjoint, you might have three brands, three colors, and three prices (nine total items), and you want exactly one brand, one color, and one price to show in each MaxDiff question. You do this by setting up prohibitions. After showing Brand listed first and Price last in each MaxDiff question, you ask respondents which item makes them most want to buy the product and which makes them least want to buy. It is not a substitute for CBC, but it can achieve similar results.