MaxDiff Analysis
Finding the Features That Truly Matter
Executive Summary
"Standard rating scales fail because users say 'everything is a 10'. MaxDiff (Best-Worst Scaling) forces trade-offs to reveal the true hierarchy of consumer needs."
01.The Failure of Rating Scales
If you ask a customer to rate 20 features from 1 to 5, you will almost always encounter Scale Inflation. Respondents, wanting to be helpful or fearing the loss of a feature, will rate 15 of them as a '5'. This provides zero signal for a product manager or pricer.
MaxDiff (Maximum Difference Scaling) solves this by removing the numbers entirely. Instead of rating, the user is shown a subset of 4-5 items and asked: 'Which is most important?' and 'Which is least important?'
This forced trade-off mimics real-world decision-making. We don't evaluate features in a vacuum; we choose them at the expense of others.
02.How MaxDiff Math Works
The Choice Task
A respondent sees multiple screens. Each screen has a balanced rotation of attributes. No item is seen too many or too few times.
Hierarchical Bayes (HB)
The backend math uses HB estimation to determine individual-level utility. It works backwards from the choices to find the underlying 'value' of each item.
Probability of Choice
Utility scores are converted into a percentage. A score of 10% means that if all features were presented, that specific feature would be chosen 10% of the time as #1.
03.Attribute vs. Price
While MaxDiff is primarily a feature prioritization tool, it is essential for pricing strategy. By including 'Price' as an attribute, you can see where it sits in the hierarchy of needs. If price is at the bottom (low importance), you have massive pricing power. If it is at the top, you are in a commodity war.
04.The 4-Step Analysis Protocol
Utility Ranking
Calculate the raw utility. This tells you the order of preference. Focus development on the top 20% of the list.
Segmentation
Analyze by persona. Your 'Enterprise' users might value Security at 40%, while 'Free' users value Speed at 40%.
Value per Dollar
Cross-reference MaxDiff scores with implementation cost. High utility + Low cost = The Pro Plan core.
TURF Analysis
Total Unduplicated Reach and Frequency. Identify the set of 3 features that satisfy 90% of the market.
Common Pitfalls
The Too-Long List
Limit your MaxDiff to 20-25 items. Beyond that, respondent fatigue causes random clicking.
Vague Descriptions
Don't test 'Good Support'. Test '24/7 Phone Support' vs 'Email Support (48h)'. Specificity is key to signal.
Ignoring the Bottom
The bottom 20% of the list are your 'Negative Value' drivers. Consider removing them to save cost without hurting conversion.
Industry Benchmarks
Recommended for stable segment-level utilities.
The sweet spot before data quality degrades.
Standard target for a winning feature subset.
Expert Q&A
Q: Can I use MaxDiff for pricing?
Yes, by including discrete price points (e.g. $10, $20, $30) as attributes, but Conjoint is generally superior for price/feature trade-offs.
Q: How is this different from Kano?
Kano measures satisfaction vs presence. MaxDiff measures relative priority. Use MaxDiff to find what to build, and Kano to find how users feel about it.