RESEARCH
July 2026 · 6 min read · Aymeric Julien, Co-founder
What is conjoint analysis?
Stop asking buyers what they want
Ask buyers what they want and they will tell you: every feature, at the lowest price. Conjoint analysis measures what they will actually choose.
Ask a buyer directly what they would pay and you get an aspirational answer, not a reliable one. People are poor predictors of their own purchasing behaviour, and every incentive in a survey pushes them to understate willingness to pay.
Conjoint analysis was built for this problem. Instead of asking what people want, it makes them choose.
What is conjoint analysis?
Conjoint analysis is a survey-based research method built on choice experiments. Respondents see a series of realistic product profiles, each combining features, brand, format, and price. In every round, they pick the one they would buy. No rating scales. No questions about how important price is. Just choices.
Because no profile has everything, every choice is a trade-off. Run enough of them across enough respondents and a statistical model can estimate how much each attribute contributed to the decisions. Researchers call these estimates utilities. They are the closest a survey gets to the priorities buyers act on but cannot articulate.
To be precise: conjoint is still a stated preference method. Respondents are choosing in a survey, not spending real money. Its advantage is the format. Forced trade-offs make honest answers the path of least resistance, which is why conjoint predicts market behaviour far better than direct questioning.
How it works
Take a CPG example. An oat milk brand is weighing two formats, three shelf prices, and whether a no added sugar claim earns its place on the pack.
A respondent sees three options. A 1 L carton, no added sugar, at $4.49. A 1.5 L carton, original recipe, at $5.29. A 1 L carton, original recipe, at $3.99. They pick the one they would buy, or none. Then the combinations change and they choose again, typically 8 to 12 times.
Each individual task looks unremarkable. Across 300 respondents and thousands of choices, the pattern is not. The model quantifies how much the claim is worth in dollars, how demand shifts between $3.99 and $4.49, and whether the larger format wins on value or loses on price.
When to use it
Use conjoint when the product itself is still a variable. Pricing a new product before launch. Designing pack and price architecture. Deciding which features justify a premium tier. Testing how your lineup holds up when a competitor cuts price.
What you get out of it
Three outputs matter.
First, the value of each attribute, in a common unit. You see whether the no added sugar claim moves choices more than a 50 cent price gap.
Second, price sensitivity. Not a single number but a curve: how preference for your product shifts at each tested price, with competitors in the frame.
Third, a market simulator. Change your price, add a feature, drop a format, and see the projected effect on preference share before committing to anything in the real market.
These are estimates from a model, not guarantees. Conjoint is strongest at relative comparisons: which option wins, which feature earns its cost, where demand starts to fall. Treat the outputs that way and it is the most decision-ready pricing method available.
When not to use it
Conjoint is the wrong tool when your product is fixed and you need a fast read on an acceptable price range. Van Westendorp answers that at a fraction of the design effort.
It also struggles when you cram in too many attributes. Past six or seven, respondents simplify their decision-making and data quality drops with them. Part of designing a good conjoint study is deciding what to leave out.
Work with Relumi
A conjoint study is only as good as its design: the attributes you test, the prices you span, the sample you recruit. That is the work we do.
You deal directly with the person running your study, from the first scoping call to the final recommendation. One month from kick-off to a price you can defend.
