Conjoint and Maxdiff Analysis – Webinar Recap

Last week, we co-hosted a special webinar all about how to run and use conjoint and maxdiff analyses. We started with a review of the two types of analysis, then moved into a case study on how Zynga used maxdiff to help narrow down things like the name of upcoming game releases. Last, but not least, we watched a demonstration of conjoint analysis and maxdiff in SurveyAnalytics (and released this last weekend on QuestionPro Enterprise). Let’s review a bit more in-depth.

Conjoint vs Maxdiff: a review

Let’s start with the difference between a maximum differential analysis (maxdiff) and a conjoint analysis.

Maxdiff: the question type for a maxdiff analysis is the single list of attributes with radio buttons on each side for respondents to select the attribute that matches for each end of a measurement spectrum. For example, you might set up a list of attributes for a product and ask a respondent to select which they would most list to see in the product versus which they would least like to see in the product.

Conjoint: the question type for a conjoint analysis is two different sets of attributes with a radio button at the bottom of each, and the respondent is asked to select which of the two they would prefer (or by more likely to buy, watch, etc.). In this case, the survey designer is using a set of attributes that have varying levels (such as two price points, two colors, two sizes, etc.). With the online tool, you can list the attributes you want to measure, and then the levels for each attribute, and how many tasks you want the respondent to complete (i.e., how many variations do you want to show each respondent).

Generally speaking, if you are looking at creating a new product, you would want to start with a maxdiff to determine the attributes that are most important to your potential consumers, then follow up with a conjoint analysis that shows those attributes to the respondent and helps you measure things like figure configuration and price sensitivity.

Best Practice: one question that kept coming up during the webinar was how many attributes and levels were recommended? For a maxdiff analysis, you could have up to 50-100 attributes, but only show respondents sets of 5 (and limit how many sets you show each respondent). You’d want to have about 300 responses total to have a good measure. For conjoint analysis, about 5 attributes with 2-3 levels per attribute seemed to be the general consensus to follow so that you don’t end up with respondent fatigue.

Zynga and Maxdiff

Rob Aseron, former director at Zynga, shared with us how he had used maxdiff to help identify things like what version of a game they should release (including what name to give the release). We got to see an example of share preference being measured using maxdiff (i.e., what was the preference for any given item over the others in the list). A great best practice he described following was anticipating what his stakeholders were going to be asking next, and creating an extra set of questions to come prepared with the answers so that he didn’t need to go back to the drawing board and field another study again each time.

The demo

We finished with a demo of how to set up a conjoint analysis from within SurveyAnalytics. (For the QuestionPro audience – we just released these features on QuestionPro Enterprise this last weekend; if you’re interested, please contact us!) The demo itself is best viewed on the recording, but you can also follow along using screen shots on the slides.


More Information: