Conjoint analysis is a market research method that measures how much customers value individual product attributes by asking them to choose between, rank or rate complete product profiles that vary systematically. Instead of asking directly which feature matters most, it infers preferences from trade-offs, which mirrors the way people actually decide when they buy.
In marketing, conjoint analysis supports pricing, product design and market simulation. Its output, a set of part-worth utilities, shows the relative appeal of every attribute level and allows analysts to estimate willingness to pay and predict share of preference for products that do not yet exist.
How conjoint analysis derives preferences
Conjoint analysis decomposes a product into attributes and levels, such as price (49, 69 or 89 dollars), storage (64 or 128 GB) and battery life (10 or 20 hours). An experimental design combines these levels into profiles, and respondents evaluate many such profiles across a series of tasks.
A statistical model then estimates a part-worth utility for each level, so that the sum of part-worths predicts which profile a respondent prefers. The model typically used for choice data is the multinomial logit model, often estimated with hierarchical Bayes to obtain individual-level utilities from a limited number of answers per person.
The table below lists the core components of a conjoint study and the role each one plays.
| Component | Description | Typical range |
|---|---|---|
| Attributes | Product dimensions being tested | 4 to 8 per study |
| Levels | Specific values of each attribute | 2 to 5 per attribute |
| Choice tasks | Screens shown to each respondent | 8 to 15 per respondent |
| Concepts per task | Profiles compared on each screen | 3 to 5, often plus a none option |
| Sample size | Respondents in the target market | Commonly 300 to 1,000 |
From the part-worths, analysts calculate attribute importance, the range of utilities within an attribute as a share of all ranges. Importance depends heavily on the levels chosen: widening a price range makes price look artificially more important, a frequent source of misinterpretation.
Types of conjoint analysis
Several variants of conjoint analysis exist, and they differ mainly in the task respondents perform and in how many attributes they can handle. Choice-based conjoint is the most widely used today because choosing among options resembles real purchasing.
| Variant | Respondent task | Strengths | Limitations |
|---|---|---|---|
| Choice-based conjoint (CBC) | Choose one profile per screen | Realistic; supports a none option | Less information per task |
| Adaptive choice-based conjoint (ACBC) | Screening, then tailored choices | Handles more attributes | Longer, more complex survey |
| Ratings-based conjoint | Rate each profile on a scale | Simple to analyze | Scale bias; weaker realism |
| Menu-based conjoint | Build a bundle from options | Fits configurable products | Large samples and complex models |
| MaxDiff (related method) | Pick best and worst items | Ranks many single features | No full product trade-offs |
MaxDiff is often grouped with conjoint, but it scales individual items rather than whole profiles. It works well for prioritizing a long feature list before a conjoint study narrows the attributes to the decisive few.
Conjoint analysis compared with other research methods
Conjoint analysis is distinct because it measures preferences indirectly through trade-offs. Direct survey questions tend to produce the answer that everything is important, while conjoint forces respondents to give up one benefit to gain another.
| Method | Question it answers | Output |
|---|---|---|
| Conjoint analysis | Which feature combination and price win? | Utilities, importance, simulated share |
| Direct rating survey | How important is each feature? | Stated importance scores |
| Focus group | Why do customers feel this way? | Qualitative insight |
| Van Westendorp pricing | What price range seems acceptable? | Acceptable price band |
| A/B price test | Which live price converts better? | Observed conversion rates |
Teams often combine methods: qualitative research first to discover which attributes matter and how customers describe them, then conjoint to quantify trade-offs, and finally in-market testing to validate the chosen configuration.
Designing and using a conjoint study
Good conjoint results start with careful attribute selection. Attributes should be independent, meaningful to buyers and described in customer language; levels should be realistic and span the range the business actually considers.
- Keep attributes mutually independent, because overlapping attributes such as speed and performance distort utilities.
- Avoid impossible combinations through prohibitions, but use them sparingly, since they reduce design efficiency.
- Include a none option when the goal is forecasting demand, not only relative preference.
- Test the survey with a pilot sample to catch confusing wording and excessive task fatigue.
The most valuable deliverable is usually a market simulator, a tool that converts utilities into share of preference for any scenario defined by the analyst. Product managers use it to compare configurations, estimate the effect of a price change and model competitor responses.
| Business question | Conjoint output used | Typical decision |
|---|---|---|
| Which price to set | Price part-worths and simulated share | Price point within a tested range |
| Which features to build | Attribute importance and utilities | Roadmap prioritization |
| How to tier a product | Simulated shares of bundle scenarios | Good, better, best packaging |
| How to respond to a rival | Competitive scenario simulation | Feature or price counter-move |
Limitations and common mistakes
Conjoint analysis measures stated choices in a survey, not actual purchases, so its predictions reflect preference rather than guaranteed market share. Simulators ignore awareness, distribution and promotion, all of which shape real sales.
Common errors include too many attributes, which push respondents toward simplifying shortcuts, and unrealistic price levels, which inflate or deflate price sensitivity. Willingness-to-pay figures derived from utilities are especially sensitive to model choices and should be treated as directional estimates, not precise prices.
Finally, conjoint assumes that a product’s value is roughly the sum of its attribute utilities. Emotional brand appeal or strong interactions between features can violate that assumption, so experienced analysts test interaction effects where theory suggests them and validate results with holdout tasks.
The simple explanation
Conjoint analysis works out what people really care about by watching their choices. Instead of asking whether they want a cheaper price or a longer battery, a survey shows several complete products and asks which one they would buy.
After many such choices, patterns appear. If people keep picking the longer battery even when it costs more, battery life clearly matters to them. The result is a score for every feature and price, which helps a company build the version people will most likely choose.
How it is used in practice
A software company planning a new project management subscription is unsure how to price three plans. Its research team runs a choice-based conjoint survey with 600 small-business managers, testing five attributes: monthly price (8, 12, 16 or 20 dollars per user), storage, number of integrations, reporting depth and support level.
Each respondent completes twelve choice tasks with three plans and a none option. Hierarchical Bayes estimation shows that advanced reporting adds more value than extra storage, and that demand drops sharply above 16 dollars. Using the simulator, the team designs a middle tier at 14 dollars with advanced reporting, which captures the largest simulated share of preference against two competing products, and later confirms the result with a limited live price test.