"We have over 50 features in our solution, and we're not sure which ones are most valued by customers — or even by different customer segments. That's making some of our bundling decisions difficult." That's a version of a conversation we have often with SaaS product and pricing teams. A mature platform accumulates features faster than anyone can track customer sentiment on all of them, and by the time someone asks what should actually sit in the core plan versus a premium tier, nobody in the room has a confident answer.
Max-Diff (maximum difference scaling) is built exactly for that moment. Customers can't reliably rank a list of 50 features when you ask them outright — most people call everything "important." Max-Diff forces genuine trade-offs by asking respondents to pick the best and worst of a small handful of items at a time, and rolls those choices up into a single, clean priority ranking across the entire list. For a photography platform, that might mean sorting out where custom website templates, richer client galleries, built-in e-commerce, 4K video support, expanded storage, and a dozen other ideas actually rank against each other.
Where conjoint analysis is built to test how customers trade off price against features, Max-Diff answers a narrower, but no less important, question first: out of everything on the list, what actually matters most — and what matters least?
How Max-Diff Studies Work
In a Max-Diff exercise, respondents are shown a series of small sets of items — typically four or five at a time, pulled from a larger master list — and asked two simple questions for each set: which item is most important (or most preferred), and which is least important. Because respondents choose extremes rather than rank an entire list or rate every item on a scale, the exercise avoids the "everything is a 9 out of 10" problem that plagues simple rating-scale surveys.
Across enough sets — each showing a different combination of items — every item in the master list ends up compared, directly or indirectly, against every other item. The resulting best/worst choices allow researchers to calculate a relative preference score for each item, producing a single clean ranking from most to least important across the full list.
Why It Works Well
Max-Diff forces genuine trade-offs. Because respondents can't rate everything highly, the output reflects real relative priority rather than survey fatigue or a tendency to rate everything favorably.
It also scales well to long feature lists. A single well-designed study can rank a dozen or more items using a relatively modest number of choice sets per respondent, since every set generates two data points — a most and a least — rather than one.
And the output is unusually easy to act on: a single ranked list of features, from highest to lowest relative importance, that maps directly onto roadmap and packaging decisions.
Where It Falls Short
Max-Diff tells you relative importance, not absolute value. It can tell you that client galleries matter more than 4K video support, but not what customers would actually pay for either one — that's a job for conjoint analysis or a price-specific method like Gabor-Granger.
It also depends on the master list being complete and well-defined. Leaving an important feature off the list, or defining items inconsistently in scope or detail, will distort the ranking even if the exercise itself is executed well.
And because it only produces a ranking, it doesn't tell you how customers would react to a specific bundle or packaging structure built from the top-ranked items — only that those items ranked highest individually.
When to Use Max-Diff Studies
- You have a long list of features, benefits, or messaging ideas and need to separate what genuinely matters from what merely sounds nice.
- You need a single, clean priority ranking to guide a roadmap, packaging decision, or messaging strategy — not a price recommendation.
- Simple rating scales have let you down before. If a prior survey came back with everything rated 8 or 9 out of 10, Max-Diff is built to force the differentiation that rating scales can't.
- You're preparing for a conjoint study. Max-Diff is a common and efficient way to narrow a long feature list down to the handful of attributes worth carrying into a full conjoint design.
That last point is worth dwelling on. Conjoint studies are typically limited to somewhere around five or six attributes — beyond that, respondent burden and design complexity make the exercise unreliable. If your master list has 50 features, you can't test all of them in a conjoint design, and you don't want to guess which handful deserve a seat at the table. Running Max-Diff first tells you which features actually move the needle, so the conjoint study that follows spends its limited attribute slots on the features that will genuinely influence a plan choice decision — not ones customers barely notice either way.
Key Considerations When Designing a Study
- Master list quality. Everything you want ranked needs to be on the list, defined clearly and consistently, before fielding begins — you can't add an item after the fact.
- Set design and balance. Items should appear a roughly equal number of times across the choice sets, and no two items should be paired together more often than necessary, so every item gets a fair, comparable read.
- List length and number of sets. Longer master lists need more choice sets per respondent to keep every item well represented; beyond a certain length, consider splitting the list into two related studies or trimming it first.
- Segment-level differences. Like any preference study, consider whether you need to read results by segment (plan tier, company size, use case) rather than only in aggregate, and size your sample and quotas accordingly.
- Pairing with conjoint. Treat Max-Diff as a scoping step, not a substitute, when the ultimate decision also involves price or feature trade-offs — use it to shortlist attributes, then bring the shortlist into a conjoint design.
Try it yourself. The simplified exercise below mirrors a real Max-Diff study for a photography platform choosing between seven potential features. Work through five rounds, picking the most and least appealing feature in each set, then see how individual answers like yours would roll up into a full best-to-worst ranking.
From Ranking to Roadmap
A Max-Diff ranking won't tell you what to charge for a feature, or how customers would react to a specific bundle — but it will tell you, with real confidence, which ideas are worth building first. Used well, it turns a roadmap debate into an evidence-based priority list, and gives a conjoint study that follows a much shorter, sharper list of attributes to work with.
The segment-level view is often where a Max-Diff study gets genuinely interesting. On one engagement, a client was struck by just how differently feature preference broke down across their customer segments — enough that it eventually led them to build a couple of niche, segment-specific plans alongside their standard plan structure, built around the particular features that mattered most to those groups.
