Using MaxDiff to make better comparisons

Karsten Shaw - Analytics Partner

When booking your summer holiday, what’s the most important factor in helping you choose? Perhaps you’re on a budget and it’s price that concerns you the most. Maybe you want to go back to your favourite destination. Or it could be a need to be by the sea. Perhaps it’s all three.

Take this next scenario: you want to lie back and watch some TV, but with so many different providers (Netflix, Apple TV, Prime etc.), how do you choose which subscription to exchange your hard-earned cash for? Is it the variety of shows? Does it have something your kids love watching? Is there a special deal on offer? 

How do we measure the importance of these endless different possibilities?

If you’re conducting a survey, then this should be a pretty straightforward exercise, right? You list out all the possible motivations and ask respondents how much they contribute to the decision. You use a five-point scale, or a ten-point scale; throw in an ‘Other’ for good measure. Simple.

But how useful is the information you’ve collected?

Think about this question: When booking a flight, how important is it that the seats are comfortable? You think of your back, which has been playing up, so this becomes a very important factor. Onto the next question: When booking a flight, how important is it that there’s a personal in-flight entertainment system? You’re going to be flying for hours. You don’t want to be bored. So, again, this becomes very important.

Suddenly you have a dozen ‘Very important’ answers and you’ve learned nothing.

This is where MaxDiff comes in.

What is MaxDiff analysis and how can it benefit businesses

Short for Maximum Difference Scaling, MaxDiff is a simple and effective method which allows you to understand the priorities and quantify the preferences or importance consumers attach to a product or service.

The approach allows you to rank items by doing what the brain does best. As consumers, when we choose from a list, say a menu, we tend to be very good at picking out both our favourite, but also our least favourite option. The ones in the middle are often a bit harder to rank. Research has demonstrated that surveys that feature rating questions are vulnerable to factors such as scale bias. A chef looking to create a new menu might ask how they would rank different dishes, but these dishes could all end up rating highly if they are popular dishes. Similarly, a ranking scale survey is susceptible to the respondents’ mood or if they are feeling hungry/full at that point. This makes the survey difficult to analyse efficiently and accurately. 

Instead of asking about attributes in isolation, MaxDiff pits them up against each other in the form of a set of attributes (this is the term that is generally used for a list of options researchers wish to rank). From our previous TV example, a survey question might look like this:

Thinking only about the items below, which of them is most likely to get you to subscribe to this TV service, and which one is least likely:

Most likely   Least likely
 A variety of programming for children 
 New blockbusters every week 
 Option to cancel subscription anytime  

We can have more than three attributes; four is the standard. But it’s important to be careful to not to overload respondents with options.

How to analyse MaxDiff data

If you’re wondering how to analyse MaxDiff data, the great news is it can be done in a number of different ways. The most common way is using a 0—100 scale. Each respondent will get a score between zero and a hundred for each attribute. This means we can treat them like any other variable and filter the scores by other characteristics such as demographics.

MaxDiff can also be used to segment based on the patterns of response, or we can run a TURF analysis, to find out the combination of items which have the broadest appeal. All without needing any extra questionnaire length.

Many studies have demonstrated that the MaxDiff approach gives superior predictions in real life, which is why we often recommend it to our clients when decisions of the business are based on the outcome of the research.

The great thing about MaxDiff is its flexibility as a tool. The examples above are about importance of factors, but MaxDiff can be used for anything which can be ranked. For instance, one MaxDiff analysis example could include testing the effectiveness of different messages or where a business should prioritise its investment. We’ve even used it to test a list of things from everyday life that people can’t live without. The possibilities are limitless.

The right data analysis for you and your clients

So, before you go ahead and book that holiday, you can rest in the knowledge that you’ve used the best possible method for your client’s business.

At Yonder Data Solutions, we combine advanced analytics with our high quality data collection solutions to provide seamless integration of stats solutions within our surveys. MaxDiff is just one of a wide suite of Advanced Analytics approaches that we can offer to meet a range of different needs and gain deeper insights. 

To get in touch, please contact Shashi Lakhani: