Conjoint Analysis for Surveys
Build a ratings-based or choice-based (CBC/HB) conjoint study, collect data and read part-worth utilities and attribute importance — end to end.
What it is
Conjoint analysis works out how people trade off the features of a product or offer. Instead of asking which features matter — where everyone says everything matters — you show respondents whole profiles that combine several attributes at different levels, and they react to them. There are two main families. In ratings-based (or rankings-based) conjoint, respondents rate or rank profiles, and utilities are estimated with regression. In choice-based conjoint (CBC), respondents pick one profile from a set, and utilities are estimated with a discrete-choice model (typically multinomial logit, often with hierarchical Bayes for individual-level estimates); CBC mirrors real purchase decisions and is the market-research standard. Either way, conjoint estimates a part-worth utility for each attribute level and an importance score for each attribute, telling you what really drives choice.
When to use it
- You want to know how customers trade off features, brand and price.
- You are testing product concepts or feature bundles before launch.
- You need attribute importance, not just stated preference.
- You want to simulate how preference shifts when you change a feature level.
When not to use it
- You only need to compare two or three finished concepts overall — a simple between-groups test may be enough.
- Your attributes cannot be combined into realistic profiles.
Assumptions
- Attributes and levels are clear, realistic and roughly independent.
- Profiles are balanced so each level appears a fair number of times.
- Respondents evaluate complete profiles, not single features in isolation.
- Enough profiles per respondent to estimate the utilities, without causing fatigue.
Thresholds & cutoffs
| Criterion | Rule of thumb | Source |
|---|---|---|
| Attribute count | Keep to about 4–6 attributes to limit respondent fatigue | Orme (2010), practical guidance |
| Profiles per respondent | Enough to estimate all part-worths; balance against fatigue | Orme (2010), practical guidance |
| Importance scores | Sum to 100% across attributes — read relative, not absolute | Green & Srinivasan (1978) |
How to report it in APA
A ratings-based conjoint analysis estimated part-worth utilities for each attribute level. Price was the most important attribute (relative importance = XX%), followed by brand (XX%) and feature set (XX%). Utilities were highest for [level] and lowest for [level], indicating that respondents most preferred [interpretation].
How Explonia runs it
Explonia helps you build a ratings-based, choice-based (CBC) or ranking conjoint study, collect responses, and estimate part-worth utilities and attribute importance — all in one place. You define attributes and levels, Explonia generates a balanced, D-efficient set of profiles, and after data collection it reports the utilities with a clear importance breakdown and drafts an APA 'Article description' you can edit. Choice-based conjoint uses multinomial logit or hierarchical Bayes (HB) for individual-level part-worths, with a market simulator, post-hoc segmentation, and respondent quality flags (straightlining, speeding) built in. See live conjoint examples in the gallery.
Frequently asked questions
What is a part-worth utility?
A part-worth utility is the value respondents place on a specific attribute level. Higher utilities mean a more preferred level. Comparing utilities within an attribute shows which level wins.
How many attributes should a conjoint study have?
About 4–6 attributes is a practical range. Too many attributes overwhelm respondents and add noise, which hurts the quality of the estimated utilities.
What is the difference between ratings-based and choice-based conjoint?
In ratings-based conjoint, respondents rate or rank profiles and utilities are estimated by regression. In choice-based conjoint (CBC), respondents choose one profile from a set and utilities come from a discrete-choice model (multinomial logit, often with hierarchical Bayes). CBC mimics real purchase decisions and is the market-research standard; ratings-based conjoint is simpler and well suited to academic and concept-testing work.
Does Explonia support choice-based conjoint?
Yes. Explonia runs choice-based conjoint (CBC) with multinomial logit and hierarchical Bayes (HB) end to end, alongside ratings-based and ranking conjoint, with a market simulator and post-hoc segmentation.
How do I read attribute importance?
Importance scores sum to 100% across attributes and show the relative weight each one carries in driving preference. Read them in comparison with each other, not as absolute percentages.
Last updated: July 2026
Run Conjoint on your own data
Collect or upload your data, run the analysis in your browser, and get a draft APA paragraph you can edit. Free to start. No credit card.
Start free