Conjoint analysis: learn what customers value through real trade-offs

Conjoint analysis measures how people make trade-offs between complete product options. Instead of rating every feature as important, respondents choose among profiles with different features, service levels, and prices. Those choices reveal relative preference.

What is conjoint analysis?

Conjoint analysis is a family of survey-based methods for estimating how parts of an offer contribute to preference. The survey presents whole product profiles. Each profile combines one level from every attribute, such as storage, support, contract length, and price.

Respondents choose or rank the profiles. A choice model works backward from those decisions to estimate the relative value of each level. These estimates are commonly called part-worth utilities.

The method is useful because products are not bought one feature at a time. A stronger support package may come with a higher price. More storage may require an annual contract. Conjoint makes those trade-offs explicit and repeatable.

Why choices reveal more than importance ratings

If you ask whether speed, reliability, support, and affordability are important, most people can honestly say yes to all four. The result does not tell you what should change when the team cannot maximize everything.

Compare that with a choice between two plans:

Neither plan wins on every dimension. Across a balanced series of tasks, the pattern of choices shows how this audience traded storage and support against commitment and price. That is much closer to the product decision the team needs to make.

Four conjoint terms worth knowing

Utility values do not have an independent natural unit. Their zero point and scale depend on the model. Compare them within the study according to the chosen coding and analysis; do not present “12 utility points” as a universal amount of customer value.

Common types of conjoint analysis

Choice-based conjoint

Choice-based conjoint (CBC) shows a small set of profiles and asks the respondent to choose one, often with a “none” option. It resembles a buying decision and is widely used for product, packaging, and pricing studies.

Adaptive conjoint

Adaptive approaches change later questions based on earlier answers. They can focus attention on relevant trade-offs when the attribute space is large, but the adaptive logic and analysis require care.

“Conjoint” is not one survey template. The form should follow the decision, the number of attributes, the respondent burden, and the model you can defend.

How to design a conjoint study

1. Name the decision. Are you choosing a package, pricing a feature, or testing a competitive offer? One study should have one primary decision.
2. Choose attributes people can trade. Include dimensions that can realistically vary and that the audience understands.
3. Write realistic levels. Cover the decision space without adding impossible or obviously dominant offers.
4. Add price only when it belongs in the question. Price is needed for willingness-to-pay work, but including it does not repair a vague product concept.
5. Define prohibited combinations. If two levels cannot coexist, keep the design from generating that profile.
6. Consider a “none” option. Use it when refusing all offers is a realistic choice. The option can improve realism but may also need separate modeling attention.
7. Build an efficient experimental design. Levels need enough variation and balance to estimate their effects without showing every possible profile.
8. Pilot the tasks. Check comprehension, task difficulty, completion time, and whether respondents use simplifying shortcuts.

Sample-size planning belongs to the design. Consider the model, attributes, levels, tasks per respondent, expected choice shares, desired uncertainty, and segment analysis. A convenient round number is not a substitute for that plan.

How to interpret conjoint results

Part-worth utilities

These estimates show the direction and relative strength of preference for levels. Read the pattern and uncertainty, not just whether one value is positive. A level can look weak because of the alternatives chosen for the study, the audience, or the coding convention.

Relative attribute importance

Importance is often calculated from the utility range within each attribute. It depends on the levels included. If you test a very wide price range and a narrow storage range, price may appear more important partly because the study gave it more variation.

Willingness to pay

When price is modelled appropriately, the utility change from a feature can be compared with the utility change from price. This creates an estimated willingness to pay. The estimate can become unstable when the price effect is weak or nonlinear, so report the model and uncertainty.

Choice simulations

Utilities can be combined into hypothetical product scenarios. A simulator estimates preference share among the offers entered; it does not automatically forecast market share. Awareness, availability, competition, switching costs, and real budgets remain outside the survey unless the research accounts for them.

When is conjoint analysis the right method?

Use conjoint when:

• the decision contains several attributes with real trade-offs;
• you can describe realistic levels before asking respondents to choose;
• you need evidence for packaging, configuration, or price-feature trade-offs; and
• the audience understands the category well enough to evaluate the profiles.

Choose a simpler or different method when:

• you only need a relative list—consider MaxDiff;
• you need to learn how presence and absence affect satisfaction—use Kano analysis;
• you are discovering a broad acceptable price range—start with Van Westendorp; or
• you do not yet understand the problem—begin with interviews or open feedback instead of forcing premature attributes.

Common conjoint mistakes

• Too much to compare. Long, dense profiles make people ignore attributes or fall back on one simple rule.
• Vague or overlapping levels. If respondents cannot see the difference, the model cannot recover it cleanly.
• Dominant profiles. An option that is better on every attribute teaches little about trade-offs.
• The wrong audience. The model faithfully estimates the preferences of whoever answered, even if they are not buyers.
• Overstated precision. A rank without uncertainty, holdout checks, or model diagnostics can look more certain than it is.
• Confusing preference with behavior. Validate survey estimates against actual choices after the offer reaches the market.

How to run a conjoint study with SenseFolks

UserChoice supports conjoint surveys for product trade-off research. Define the research design before configuring the component so the tasks, audience, and analysis all answer the same question.

1. Add the website where qualified respondents will see the study.
2. Create a UserChoice survey.
3. Define concrete attributes, realistic levels, and any invalid combinations.
4. Pilot the choice tasks with people from the intended audience.
5. Embed the survey at a relevant evaluation moment.
6. Review utilities and scenario outputs with the study assumptions beside them.
7. Record the product decision and validate it with observed behavior.

Read the UserChoice reference for configuration details and the embedding guide for installation.

Conjoint analysis questions, answered