How to use best-worst scaling to make better decisions
Best-worst scaling (MaxDiff analysis) shows which features or messages your audience values most. Learn how it works and how to run your own study.
At a glance
- Best-worst scaling tells you what matters most in relative terms. Pair it with conjoint analysis when you need to know how much a feature is worth in a purchase decision.
- Best-worst scaling forces respondents to pick their most and least preferred option from a set, which produces cleaner data than a standard rating scale.
- It is the method to reach for when every option looks equally important on paper and you need to prioritize features, names, or messages.
- SurveyMonkey LaunchPad automates the MaxDiff methodology and TURF analysis, so you get a stack-ranked list of priorities without R code or manual number crunching.
Best-worst scaling, also called MaxDiff analysis, ranks what your audience values most by forcing real trade-offs instead of vague ratings. Respondents pick their most and least preferred option from small sets of features, messages, or ideas. The result is a clear, stack-ranked list of priorities you can use to guide product, pricing, and messaging decisions with real evidence, not guesswork.
Skip the manual math
SurveyMonkey LaunchPad runs your MaxDiff study, scores the results, and turns complex trade-off data into a stack-ranked list of what your customers value most.
What is best-worst scaling, and how does MaxDiff fit in?
Best-worst scaling is a survey method for understanding how important different attributes, such as product features, packaging, or messaging, are to your target market.
Instead of asking people to rate every option (where everything tends to score well), it asks them to choose only the best and worst options from a set. That forces a real trade-off, so you learn what people actually prioritize rather than what they politely approve of.
Two terms come up constantly in this method:
- Attribute: a single property, item, or feature you're measuring.
- Set: a group of attributes shown to a respondent at once.
MaxDiff analysis is the specific statistical technique behind best-worst scaling. When people say "run a MaxDiff," they mean designing and analyzing a best-worst scaling study. The two terms are often used interchangeably, and that's fine. Just know that best-worst scaling is the survey design, and MaxDiff is the math behind it.
Best-worst scaling and conjoint analysis both uncover preferences, but they answer different questions. Best-worst scaling tells you what matters most, in relative terms. Conjoint analysis tells you how much each attribute is worth, in dollars or trade-off terms, when people are choosing between full product bundles. Many teams run both: best-worst scaling to narrow down a long list of attributes, then conjoint analysis to price and package the winners.
Need a dedicated market research partner?
Our experts can help design, field, and analyze studies.
How best-worst scaling compares to other methods
| Method | What respondents do | What you learn | Best for |
| Best-worst scaling (MaxDiff) | Pick the most and least preferred option from small sets | A relative ranking of what matters most | Prioritizing features, names, or messages |
| Conjoint analysis | Choose between full product or price bundles | How much each attribute is worth in a purchase decision | Pricing and product bundling |
| Rating scales (Likert) | Rate each item on a scale, independent of the others | Broad sentiment, often skewed toward "everything is important" | Quick satisfaction checks, not prioritization |
How to run a best-worst scaling stufy in five steps
- Pick the attributes you're testing. These might be product features, brand names, or campaign messages. Keep the list focused. Too many attributes overwhelm respondents and muddy the results.
- Build your question sets. Create at least six versions of your best-worst question using randomization, item balance, or paired balance, so each attribute appears at least three times, the same number of times as every other attribute, and paired with every other attribute an equal number of times.
- Send it to the right people. Survey your own contacts, or reach a global panel for a targeted sample in as little as one hour.
- Let the analysis run. Counting analysis, individual-level score estimation, and TURF simulation turn raw choices into a stack-ranked list of what your audience values most and which combination of features reaches the widest share of them.
- Present the findings. Bring stakeholders a ranked list and a clear recommendation, not a spreadsheet of raw scores.
SurveyMonkey LaunchPad's Feature Prioritization solution builds steps two and four for you. You add the features, messages, or benefits you want to compare, and the platform auto-builds statistically balanced question sets, then runs the MaxDiff and TURF analysis the moment responses come in.
"SurveyMonkey unlocks the ability for people that don't have that market research background to do a MaxDiff, which is awesome," said Joshua Guilfoyle, Senior UX Researcher at Just Eat Takeaway.
Why teams use best-worst scaling, and where it falls short
Best-worst scaling earns its place in a market researcher's toolkit for a few concrete reasons:
- It's easy for respondents. Choosing a best and worst from a small set mimics how people actually make trade-offs, so answers come faster and cleaner than a wall of rating questions.
- It cuts through bias. There's no scale to misread and no cultural association with a particular number. Forced choices eliminate the "everything is a 4 out of 5" pattern that rating scales invite.
- It supports real statistical modeling. The resulting data quantifies preference strength, not just direction, so you can build models around it.
It has real limits too. Longer attribute lists mean longer surveys, and shorter surveys get better response rates.
The method measures preferences relative to the other attributes in the set, not in absolute terms, so a "losing" attribute in one study might test well in a different context.
And because best-worst scaling optimizes individual features rather than a complete product, a clear feature winner doesn't guarantee customers will pay more for it.
If price matters as much as feature choice, pair your results with conjoint analysis before you commit resources.
Where best-worst scaling delivers the clearest wins
Best-worst scaling works anywhere you need to turn a long list of options into a short, defensible list of priorities:
- Feature prioritization: find out which features your roadmap should actually lead with, backed by data instead of the loudest voice in the room.
- Name and brand testing: see which name option fits your target market and your brand, before you commit to creative and packaging.
- Messaging and claims: learn which value propositions or ad claims your audience finds most credible and relevant.
- Segment comparisons: use the same study to see how preferences shift across age groups, regions, or customer segments, so you can tailor messaging by audience.
Examples of best-worst scaling
Now that we’ve covered why you should use best-worst scaling, let’s look at some examples.
Example 1
Question 1:
Think about what would make you choose one restaurant over another. Considering these features, which is most important and which is least important?
| Most Important | Least Important | |
| ◻️ | Uses only locally-grown ingredients | ◻️ |
| ◻️ | Restaurant supports charities | ◻️ |
| ◻️ | Options for special diets (e.g. vegan, gluten-free) | ◻️ |
| ◻️ | Fun, clean atmosphere | ◻️ |
Question 2:
Think about what would make you choose one restaurant over another. Considering these features, which is most important and which is least important?
| Most Important | Least Important | |
| ◻️ | Serves alcoholic beverages | ◻️ |
| ◻️ | Restaurant supports charities | ◻️ |
| ◻️ | Totally organic menu | ◻️ |
| ◻️ | Uses only locally-grown ingredients | ◻️ |
Example 2
Question 1:
When choosing a hotel, what are the most and least important factors in your decision?
| Most Important | Least Important | |
| ◻️ | Workout facilities | ◻️ |
| ◻️ | Restaurant on-site | ◻️ |
| ◻️ | Pool | ◻️ |
| ◻️ | Suites available | ◻️ |
Question 2:
When choosing a hotel, what are the most and least important factors in your decision?
| Most Important | Least Important | |
| ◻️ | Cleanliness | ◻️ |
| ◻️ | Free Wi-Fi | ◻️ |
| ◻️ | Complimentary breakfast | ◻️ |
| ◻️ | Suites available | ◻️ |
Example 3
Question 1:
Below are names for our new line of cookware for kids. Please indicate the name you think is the best fit for our brand and which is the worst.
| Best fit | Worst fit | |
| ◻️ | Kids Cook | ◻️ |
| ◻️ | The Mini Mix | ◻️ |
| ◻️ | Kids in the Kitchen | ◻️ |
| ◻️ | Super Cooks! | ◻️ |
Question 2:
Below are names for our new line of cookware for kids. Please indicate the name you think is the best fit for our brand and which is the worst
| Best fit | Worst fit | |
| ◻️ | Kids in the Kitchen | ◻️ |
| ◻️ | Salt and Pepper Cookware | ◻️ |
| ◻️ | Kids Kuisine | ◻️ |
| ◻️ | We Can Cook! | ◻️ |
Turning your best-worst scaling data into a decision
Once responses are in, a simple counting analysis gets you most of the way there:
(# of times attribute was selected as best - # of time attribute was selected as worst) / # of time the item appeared = score
- A positive score means an attribute was chosen as most appealing more often than least appealing.
- A score near zero means respondents were split.
From there, individual-level score estimation and Empirical Bayes analysis add more nuance at the respondent level, while TURF simulation shows which combination of top attributes reaches the broadest share of your audience, useful when you can only ship two or three of your five best ideas.
Start using best-worst scaling in your business
The more you know about what your audience truly values, the more confidently you can prioritize the roadmap, name the product, or pick the message that lands.
Work with the SurveyMonkey market research team or run your own study with MaxDiff analysis to put your next decision on solid ground.