How to use best-worst scaling to make better decisions

Best-worst scaling, also known as MaxDiff (maximum difference) analysis, is useful when gauging preferences on messaging, brand names, product features, and more. It’s easy to use and offers clear insights into what customers truly like—and dislike. Let’s take a look at best-worst scaling and how you can use it to make better decisions for your business.

Best-worst scaling is a type of survey research conducted to understand the relative importance of attributes such as product features, packaging, messaging, etc., to your target market. By identifying what consumers really want most, you can maximize your business investments of time and capital on efforts that will appeal to your target market.

Best-worst scaling asks respondents to choose among several options at once—selecting only the best and worst options. This type of survey question allows you to collect the information you are seeking quickly and definitively. There’s no guesswork about what respondents mean when they choose a score near the middle of the range as in Likert scales or rating scale questions. Your respondents simply choose the most and least important options to them. MaxDiff questions can be asked in a single survey or as part of a longer questionnaire.

Two terms are frequently used in best-worst scaling:

How to conduct your best-worst scaling study: 

Best-worst scaling is very similar to conjoint analysis for determining respondents’ preferences for various items or features. However, MaxDiff analysis is easier to use and less refined than conjoint analysis. 

Here are examples of the two methods for comparison:

Conjoint analysis studies mimic shopping trips, where participants review products, features, attributes, and prices to make purchase decisions. The analysis is complex and considers multiple factors.

For example, you could ask “If you were in the market for a new smartphone, which of the following would be most appealing to you?” Respondents would compare brand, price, storage and more. 

In best-worst scale survey questions, respondents are asked to choose the least and most important factors within the answer options. Data analysis is faster and cleaner.

For example, you could ask, “If you were in the market for a new smartphone, please indicate the feature that would be the most important in your purchase decision and which feature would be the least important.” You would then list various features like camera, display, face ID, and more–and respondents would decide which feature was most and least important. 

As you can see in the examples, both methods are seeking what customers prioritize when purchasing a smartphone. The best-worst model data will reveal the most preferred feature and the least preferred feature. Conjoint analysis shows how much each feature influences the final purchasing decision. 

Often, these methods are used together to create a more detailed picture of what customers want and are willing to pay for.

In addition to the ease of creation and data analysis, best-worst scaling has several advantages.

Best-worst scaling questions are easy for respondents to answer. They are, in effect, simulating real-world behaviors in making choices and trade-offs, eliminating options that they don’t feel strongly about. Answering several of these questions makes the strength and importance of each choice known.

By forgoing a ranking scale, best-worst scaling avoids biases from cultural differences—some cultures have associations with different numbers—or perceptions of ratings. Respondents may also have list order bias, where they indicate that all features are equally important. When bias occurs, the business may waste resources on improvements or features that aren’t truly important to customers.

MaxDiff survey data is well suited for creating statistical models. These models make it possible to quantify preferences and understand what your target market values.

Determining what features are deemed best and worst by your customers provides you with actionable insights for product features and improvements, new products, and obsolete product features. Use this data to give your customers what they truly want and value. This will result in increased customer satisfaction overall.

As with any process, there are some downsides to best-worst scaling.

Depending upon the number of attributes you are testing, it may take more time to take your survey. Shorter surveys tend to have higher response rates. 

MaxDiff analyses measure respondents’ preferences based on the attributes presented in relation to each other. This does not account for whether the attributes offered on the survey are good or bad based on an absolute perspective.

Best-worst scaling is typically used to optimize features, not the product as a whole. For example, your MaxDiff survey may show a definite preference for one feature, but in reality, customers may not be willing to pay a higher price for the feature. To dig deeper into your results, perform a conjoint analysis to test the product as a whole.

Is the level of preference indicated for an attribute conditional upon the alternatives it is compared to? This is the context effect. For example, if you’re conducting a best-worst scaling smartphone study and your attributes are Samsung, Nokia, Google, Sony, and Apple, respondents are probably thinking about Google in terms of hardware. But if you’re doing a study that includes Google and Yahoo in its attributes, people will more likely think of Google in terms of its search engine.

It’s important to be aware of the context effect even though it is impractical to try to avoid attributes because of this possibility.

Standard rating questions can have user scale bias, scale meaning bias, and lack of discrimination between answer choices. MaxDiff surveys eliminate the issue of bias because participants are asked to choose their preference, not rank or rate.

As we mentioned in our earlie