When to Use Conjoint Analysis: Is It Right for Your Study?

Andrzej Szymkowiak · Last reviewed June 2026

You are choosing between a full conjoint study and a few rating questions, and you are not sure which one your research question actually needs. Conjoint analysis is the right method when your research question is specifically about trade-offs between product attributes — how much price, quality, or brand each contribute to preference. When your question is about a single attribute or overall satisfaction, a simpler rating scale answers it faster and with less respondent burden. This guide helps you decide, and shows the key design choices if conjoint fits.

Conjoint analysis is a powerful method for measuring trade-offs. But it is not always the right choice. Before you design a conjoint study, you need to answer a more basic question: does your research problem actually require trade-off data? This guide helps you decide — and if conjoint is the right fit, it walks you through the key design decisions. For the broader study design process that precedes this decision, see from research question to questionnaire →.

What conjoint measures — and what it does not

Conjoint analysis asks respondents to evaluate product profiles that vary on several attributes at once. By seeing how ratings (or choices) change as attributes change, you estimate the relative importance of each attribute and the part-worth utility of each level within that attribute.

Conjoint answers:

• Which product attributes matter most to respondents?
• How much does a price increase reduce preference, compared with a quality improvement?
• What combination of attributes produces the highest overall preference score?

Conjoint does not answer:

• Whether respondents like the product category at all. A simple attitude rating does this.
• How often respondents buy. A frequency question does this.
• Why respondents prefer one attribute over another. Conjoint reveals that they do, not the reason.

If your research question is "how does brand compare with price and delivery speed in driving preference?", conjoint is a good fit. If your question is "does our target segment find this product appealing?", a simple rating scale or attitude survey is faster and easier to analyse.

When is a simple rating question enough instead of conjoint?

Use a simple rating item when:

• You have one attribute or feature to evaluate.
• You want an overall satisfaction or quality score.
• You are in the exploratory phase, before committing to a full conjoint design.
• Your available sample is small (conjoint requires more respondents to produce stable part-worth estimates).
• Your respondents are likely to tire easily — for example, clinical populations or older adults.

Reserve conjoint for the cases where the trade-off structure is the actual research question. For guidance on designing and wording those simpler rating items, see how to write good survey questions →.

How do you choose conjoint attributes and levels?

The quality of a conjoint study depends almost entirely on how you define the attributes and their levels.

Attributes are the features of the product that vary across profiles. Common examples in consumer research: price, brand, quality tier, delivery time, packaging type.

Levels are the values each attribute can take. For price: €10, €20, €30. For delivery time: next day, three days, one week.

A common mistake is including too many attributes because everything seems important. Run a simple importance-rating question in a pilot first to narrow the list.

Ratings-based vs choice-based conjoint

There are two main variants, and they are not interchangeable.

Ratings-based conjoint asks respondents to rate each profile on a scale — for example, 0 to 100 likelihood to buy, or a 1–7 preference scale. It produces continuous data, is straightforward to analyse, and puts less cognitive demand on respondents than making repeated binary choices.

Choice-based conjoint (CBC) presents respondents with a set of profiles and asks them to choose one — or none. CBC is considered more realistic because it mirrors an actual purchase decision. Hierarchical Bayes (HB) estimation is the standard analysis method for CBC, as it recovers individual-level part-worths from a manageable number of choice tasks per respondent.

What Explonia supports: both. You can design the attribute structure, generate a balanced, D-efficient set of profiles or choice tasks, collect responses, and estimate part-worths and attribute importance within the same platform — ratings-based conjoint (utilities from regression) and choice-based conjoint (CBC) with hierarchical Bayes (HB) for individual-level part-worths, plus a market simulator and post-hoc segmentation.

For most academic consumer studies, ratings-based conjoint is simpler to administer and answers the trade-off question well. Choose CBC/HB when you need choice shares, a market simulator, or individual-level part-worths for segmentation.

Respondent burden and number of profiles

A ratings-based conjoint study typically presents 12 to 25 profiles per respondent. With 3 attributes and 3 levels each, you have 27 possible profiles (3³). A fractional factorial design reduces this to a manageable subset — usually 9 to 18 — while still allowing estimation of all main effects.

Most respondents can rate up to 15 to 20 profiles reliably before fatigue sets in. If your design produces more than that, consider a between-subjects design: different respondents see different subsets of profiles, and estimates are pooled across subsets. For sample size planning specific to conjoint studies, see how many participants do I need →. To recruit your target sample from an online panel, see how to recruit participants on Prolific →.

What you can conclude

Part-worth estimates from a conjoint study tell you about relative preference within the attribute levels you included. They do not:

• Predict actual market share without a separate simulation model.
• Generalise to attributes or levels you did not include in the design.
• Tell you how people will actually behave. Conjoint is a designed experiment, so part-worths do reflect the causal effect of the attribute levels on stated preference within the task — but stated preference in a survey is not the same as real purchasing, and the estimates say nothing about drivers outside the attributes you varied.

State these limits clearly in your methods and discussion. A sentence such as "Part-worth utilities reflect relative preference within the evaluated attribute space and do not constitute a market simulation" is honest and easy to defend.

Design your study with the conjoint designer

Explonia's conjoint designer (/tools/conjoint-designer/) guides you through entering attributes and levels, and builds a balanced, D-efficient design for ratings, choice-based (CBC) or ranking conjoint, then previews the survey as a respondent will see it. You can browse live conjoint examples →. Once data is collected, part-worth analysis, attribute importance, and — for CBC/HB — the market simulator and segmentation run in the same platform, see how the conjoint analysis works →.

Common mistakes to avoid

• Including more than 6 attributes, which creates fatigue and makes estimates less stable.
• Using extreme or unrealistic levels — a price of €0 or €5,000 for a grocery item distorts part-worths.
• Choosing conjoint for a study where a simple importance-rating question would answer the research question.
• Treating ratings-based and choice-based conjoint as equivalent — the data structure, analysis method, and outputs differ.
• Claiming market-share predictions from part-worth utilities without a proper simulation model.

Frequently asked questions

References

Green, P. E., & Srinivasan, V. (1978). Conjoint analysis in consumer research: Issues and outlook. Journal of Consumer Research, 5(2), 103–123.

Louviere, J. J., Hensher, D. A., & Swait, J. D. (2000). Stated choice methods: Analysis and applications. Cambridge University Press.

Orme, B. K. (2010). Getting started with conjoint analysis (2nd ed.). Research Publishers.

How Explonia does this for you

Explonia supports ratings-based, choice-based (CBC/HB) and ranking conjoint from design to part-worth estimation: define attributes and levels, generate a balanced, D-efficient profile set, collect responses, and read part-worths and attribute importance without leaving the platform. CBC/HB adds a market simulator and post-hoc segmentation. The conjoint designer is at /tools/conjoint-designer/. Rigor, made runnable.

Planning a conjoint study? Design your attribute structure and analyse part-worths in the same tool where you run your other analyses. Start free →