MaxDiff Tells You What People Say They Want. Not What Drives the Sale.

MaxDiff is the best ranking tool in survey research — and still measures the wrong thing for high-stakes decisions. Here's how it works, and where it quietly misleads.

What MaxDiff actually does

MaxDiff — maximum difference scaling, also known as best-worst scaling — is a method for ranking a list of items by importance or preference. It is the cleanest ranking instrument survey research has produced, and on its own terms it is genuinely excellent.

Here is the mechanic. Instead of asking respondents to rate every item on a scale, you show them small subsets — usually four or five items at a time — and ask two questions: which item in this set is best (most important, most appealing), and which is worst. Each respondent sees many such sets, with the items rotated so every item is shown a balanced number of times. From those repeated forced choices you reconstruct a full importance ranking for the entire list.

Why researchers love it — and they're right to

MaxDiff earns its reputation. Compared with the rating scales it replaces, it fixes three real problems:

• It forces discrimination. On a 1-to-5 grid, respondents rate everything "important." MaxDiff makes them choose, so you get a real hierarchy instead of a wall of fours and fives.
• It kills straight-lining. You cannot speed through a MaxDiff by ticking the same box down a column. Every screen demands a fresh decision, which raises data quality.
• It travels across languages and cultures. There is no scale to interpret — no "does a 4 mean the same in Japan as in Germany?" A best-and-worst choice is unambiguous, so it is robust in multi-country studies where scale-use styles differ wildly.

If your job is to take a long list and produce a defensible, bias-resistant ranking, MaxDiff beats Likert ratings every time. None of what follows takes that away.

The turn: it measures what people say, not what they do

Here is the part survey vendors rarely put on the slide, because they sell survey software and MaxDiff is what they sell.

MaxDiff measures stated importance inside a survey grid. It does not measure causal impact on real choice or revenue. Those are different things, and the gap between them is exactly the say-do gap — the most expensive blind spot in consumer research.

A respondent can rank "sustainability" as the #1 feature in your MaxDiff and then buy the cheaper, less sustainable competitor an hour later without a flicker of guilt. The ranking is sincere. It is also irrelevant to the purchase. Stated importance is not the same as a revealed driver of behavior. MaxDiff is very good at the first and silent about the second.

The three gaps

When a MaxDiff result misleads a real decision, it is almost always through one of these:

• Stated vs. causal importance. What people nominate as important and what actually moves their choice are different variables. MaxDiff captures the nomination, not the cause.
• Context-free. The grid has no price, no competitor on the shelf, no real money at stake, no friction. Strip those away and "importance" inflates for the things people wish mattered to them.
• Importance ≠ impact. An attribute can rank highly and still have zero leverage on the outcome you care about — because everyone already delivers it, or because it never tips a decision. Importance is a ranking; impact is a business result. They are not interchangeable.

What to do instead — or alongside

The fix is not to abandon MaxDiff. It is to stop treating a stated ranking as if it were a causal one.

The decision-grade question is never "what do customers say matters?" It is "what actually moves choice, conversion, and willingness to pay?" That answer comes from derived, causal importance — modeling real behavioral data to see which drivers genuinely shift the outcome, rather than asking people to nominate their own reasons.

This is the core of SUPRA's approach: pairing Deep Implicit measurement of gut-level response with Causal AI that isolates true drivers from the noise of correlation. When the stakes are high, use MaxDiff to prioritize the list, then validate the top of that list against what behavior actually shows. When the two agree, you proceed with confidence. When they disagree — and on the attributes people most love to over-claim, they often do — the behavioral signal wins.

When MaxDiff is fine — and when it isn't

The bottom line

MaxDiff is the best ranking tool in the survey toolbox. Use it when you need a fast, robust, bias-resistant ordering of a list — that is real value, and the vendors are right to sell it.

Just don't mistake the ranking for the answer. A list of what people say they want is the starting point of a decision, not the end of one. When the cost of being wrong is real, find out what actually drives the sale.

Frequently asked questions