# MaxDiff

MaxDiff estimates relative preferences by asking respondents to choose the most and least preferred items from repeated subsets.

- Type: method
- Domain: Research & behavior
- Level: Advanced
- Last reviewed: 2026-10-20
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/maxdiff/>)

## What it means

MaxDiff, also called maximum difference scaling or best-worst item scaling, is a survey method for comparing a list of items. Respondents make repeated best-and-worst selections from small sets. Analysis estimates relative preference or importance scores within that item universe; it does not directly estimate willingness to pay.

## Why it matters

When people rate every feature as important, ordinary rating scales provide little separation. Forced comparisons expose tradeoffs and can help a team understand which benefits or messages stand out for different groups.

## When to use it

Use MaxDiff when comparing distinct items under one consistent question. Use conjoint analysis for tradeoffs among multi-attribute packages. Poorly understood items or a list containing multiple ideas per item should be resolved before fielding the study.

## How it works

1. Define the decision and write clear, comparable items.
2. Choose a sample and a survey design with adequate item exposure and comparisons.
3. Pilot the wording and cognitive workload.
4. Collect most-and-least choices across the designed sets.
5. Estimate scores with a suitable model and examine uncertainty and segments.
6. Use the relative results alongside cost, feasibility, and behavioral evidence.

## A practical example

A collaboration product compares twelve benefits, including finding previous decisions and reducing status meetings. Its target users repeatedly favor decision retrieval. The team investigates that need in interviews before selecting a feature implementation.

## Where it breaks

Results depend on which items were included and how they were worded. Relative scores are not the percentage of people who will buy a feature. Design and estimation choices need expertise, and an aggregate ranking can conceal conflicting segments.

## Common mistakes

- Interpreting preference scores as purchase probabilities.
- Comparing scores from different item lists as if their scales were identical.
- Combining several benefits into one attractive item.

## Sources

1. [MaxDiff](<https://sawtoothsoftware.com/maxdiff>) — Sawtooth Software · Method documentation
2. [MaxDiff Analysis Technical Overview](<https://www.qualtrics.com/support/conjoint-project/getting-started-conjoints/getting-started-maxdiff/maxdiff-analysis-white-paper/>) — Qualtrics · Method documentation

## Related topics

- [Conjoint Analysis](<https://wiki.soroush.uk/topic/conjoint-analysis/>)
- [Survey Design](<https://wiki.soroush.uk/topic/survey-design/>)
- [Sampling in User Research](<https://wiki.soroush.uk/topic/sampling-in-user-research/>)
