Conjoint Analysis
Conjoint Analysis (Choice-Based and Adaptive Variants) · Also known as: CBC conjoint, choice-based conjoint, adaptive conjoint analysis, full-profile conjoint, Birleşik Analiz (Conjoint Analysis — CBC, ACA)
Conjoint analysis is a preference-measurement technique that decomposes overall product evaluations into the separate utility values — called part-worths — that respondents assign to each attribute level. Formalised by Green and Srinivasan in their seminal 1978 Journal of Consumer Research paper, the method has become the dominant tool in marketing research and product design for quantifying what buyers truly trade off when they choose between options.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use conjoint analysis when you need to measure how much each feature of a product or service contributes to consumer preference, and when direct questioning about importance would be unreliable due to social desirability or inability to introspect. The method suits cross-sectional survey data collected from at least 100 respondents. Key requirements are: profiles must be constructed using a D-optimal or orthogonal array design to allow clean estimation; CBC assumes the Independence of Irrelevant Alternatives (IIA) property at the aggregate level; ACA is appropriate when the number of attributes is large (more than six) and a traditional full-profile design would become cognitively overwhelming. The method is not suited to situations where the number of realistic attribute combinations is so large that no valid fractional design can be constructed, or where respondents cannot be expected to evaluate trade-offs consistently.
Strengths & limitations
- Recovers implicit trade-off weights without directly asking respondents to state importance, reducing social desirability bias.
- Produces individual-level utility estimates via Hierarchical Bayes, enabling market segmentation by preference profile.
- Supports market simulation: any combination of attributes can be scored and choice shares forecast for competing product concepts.
- Applicable to both continuous and categorical attributes, including price, brand, and feature levels.
- Requires careful experimental design (D-optimal or orthogonal array); poor design yields confounded, uninterpretable estimates.
- CBC assumes IIA — that the odds of choosing one option over another are unaffected by adding or removing a third option — which may not hold in real markets.
- Results are only as meaningful as the attribute list chosen; omitting an important attribute causes its effect to be absorbed into others.
- Large attribute sets inflate cognitive burden for respondents; ACA or hierarchical designs are needed beyond roughly six attributes.
Frequently asked
What is the difference between full-profile, CBC, and ACA conjoint?
Full-profile conjoint presents each respondent with complete product descriptions rated on a scale; it works well with up to six attributes and uses OLS for estimation. CBC replaces ratings with forced choices among small sets of full profiles, mirroring real purchase decisions and using multinomial logit for analysis. ACA (Adaptive Conjoint Analysis) customises the question sequence for each respondent based on earlier answers, making it feasible to handle ten or more attributes at the cost of requiring specialised software.
How many respondents do I need?
For aggregate-level OLS estimation in full-profile designs, around 50–100 respondents is often cited as a minimum. For CBC with Hierarchical Bayes, a minimum of 100 is recommended, with 200 or more providing stable individual-level estimates. Precision also depends on the number of attributes, levels, and choice tasks per respondent.
What is Hierarchical Bayes and why is it preferred for CBC?
Hierarchical Bayes (HB) is a Bayesian estimation framework that borrows information across respondents to stabilise individual-level part-worth estimates. Because each respondent answers only a subset of possible choice tasks, their data alone are insufficient for reliable estimation; HB pools information from the full sample through a population-level prior, producing individual utilities that are both data-driven and regularised. This enables market segmentation and individual-level prediction that aggregate MNL cannot provide.
How is Relative Attribute Importance interpreted?
Relative Attribute Importance (RAI) expresses, as a percentage, how much of the total utility variation in the study is attributable to each attribute. An RAI of 40% for price means price drives 40% of the preference differences between profiles — more than any other single attribute. RAI values sum to 100% across all attributes in the design, making them directly comparable.
Sources
- Green, P.E. & Srinivasan, V. (1978). Conjoint analysis in consumer research: Issues and outlook. Journal of Consumer Research, 5(2), 103–123. DOI: 10.1086/208721 ↗
- Orme, B.K. (2020). Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research (3rd ed.). Research Publishers. link ↗
How to cite this page
ScholarGate. (2026, June 1). Conjoint Analysis (Choice-Based and Adaptive Variants). ScholarGate. https://scholargate.app/en/experimental-design/conjoint-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Discrete Choice SimulationSimulation↔ compare
- Fractional Factorial DesignExperimental design↔ compare
- Full Factorial DesignExperimental design↔ compare
- Multinomial LogitEconometrics↔ compare
- Randomized Controlled TrialExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare