What is MaxDiff analysis?

Also called: MaxDiff, Maximum difference scaling, Best-worst scaling

MaxDiff analysis is a survey method that shows people small sets of items, asks them to pick the most and least important in each set, and uses those choices to rank every item on one scale of relative preference.

Updated

Where it comes from

MaxDiff, short for maximum difference scaling, is also called best-worst scaling. It was developed by Jordan Louviere, a researcher in choice modelling, and is now a standard method in market research for ranking features, messages and benefits.

Why teams use it instead of rating scales

Ask customers to rate 15 features from 1 to 5 and most features come back as "important". People don't have to trade anything off, and some respondents rate everything high while others rate everything in the middle. MaxDiff forces a choice in every set, so you get clear separation between items and less noise from how each person uses a scale.

A worked example

A product team has 12 candidate features for next year and wants to know which matter most to admins. Each respondent sees a series of screens with four features at a time, for example:

  • Screen 1: SSO, audit log, bulk user import, custom roles. Most important: SSO. Least important: bulk user import.
  • Screen 2: audit log, usage reports, API access, custom roles. Most important: custom roles. Least important: usage reports.
  • The design rotates features across screens so each one appears about the same number of times, usually three or more per respondent, and appears alongside different features each time.

How to score it

  • Counting: for each item, take (times picked most − times picked least) ÷ times shown. It's quick, works in a spreadsheet and is fine for a first read.
  • Modelled scores: most survey platforms with MaxDiff estimate preference scores per respondent with hierarchical Bayes or a logit model. These let you compare segments and turn scores into shares that add up to 100.
  • Report the ranking with the gaps. If the top three items are far ahead of the rest, that's a clearer story than a list of 12 in order.

What MaxDiff won't tell you

MaxDiff gives relative preference, not absolute value. The bottom item may still matter to some customers, and the top item may still not be worth building. It also says nothing about why. Follow up with interviews on the top and bottom items to find out what people expect from them.

Go deeper

Questions

How many items can a MaxDiff study include?

There's no fixed limit, but every extra item adds screens per respondent, which raises survey fatigue. If your list is long, cut obvious duplicates first or split it across respondents with a design built for that.

What's the difference between MaxDiff and conjoint analysis?

MaxDiff ranks a list of separate items, such as features or messages. Conjoint tests combinations of attributes, such as price, plan and feature set together, to estimate how people trade one attribute against another. Use MaxDiff to prioritise a list, and conjoint to design a package.

How many items should each MaxDiff screen show?

Sawtooth Software, which makes MaxDiff survey software, recommends four or five items per set. Fewer gives you less information per screen, and more makes each choice harder, especially on mobile.

Next step

Run the study, not just the method

Great Question handles recruiting from your own customers, scheduling, incentives, interviews, tests and analysis in one place.

Book a demo

No credit card required.

Related terms

Full glossary
  • Van Westendorp price sensitivity meter

    The Van Westendorp price sensitivity meter is a survey method for pricing research that asks people four questions about when a product would feel too cheap, a bargain, getting expensive and too expensive, then plots the answers to find a range of acceptable prices.

  • Acquiescence bias

    Acquiescence bias is the tendency for survey respondents and interviewees to agree with statements or answer yes regardless of what is being asked, which inflates agreement on agree/disagree items and yes/no questions.

  • Survey fatigue

    Survey fatigue is the drop in willingness or attention that sets in when people are asked to complete too many surveys, or surveys that are too long, so they stop responding, quit partway through or rush their answers.