Conjoint Analysis Explained: Methods, Examples, Tools

Overview: Conjoint analysis measures the importance of product attributes by quantifying how people make choices. In a conjoint study, product concepts are evaluated, and preferred options are selected. These choices force trade-offs between attributes. The resulting data is modeled and converted into utility scores that quantify each attribute’s relative importance. Conjoint analysis is commonly used in product development and pricing research.

Getting Started: Create a Conjoint survey using the template below, customize attributes, concepts, and choice sets. Analyze results, identify top preferences, and explore segments. The tool offers survey panels for concept testing. This guide explains how Conjoint works, the different methods, how to design a proper study, and provides a walk-through of the results analysis.

Below is an interactive example of how the Conjoint analysis tool displays a study. A retailer launching a new ice cream product must choose the right flavor, size, and price. Each option becomes an attribute with defined levels, and respondents select their preferred combinations. The resulting choices reveal the value customers place on each feature and help identify the optimal product offering.  

Conjoint analysis is most valuable in two primary scenarios: marketing research and pricing analysis. It helps organizations understand how people value individual features of a product or service, which price points are most acceptable, and which attribute combinations create the strongest overall appeal.

In the example above, without conjoint analysis, pricing would need to be evaluated separately from other attributes such as size or flavor. This disconnect makes it difficult to assess how price interacts with features in real purchase decisions, reducing the reliability of the resulting data. Conjoint analysis addresses this by presenting realistic trade-offs between product options, allowing the relative importance of attributes such as price, size, flavor, and packaging to be quantified.

At SurveyKing, conjoint analysis is often part of financial modeling services where preference data is translated into revenue, pricing, and profitability insights to guide go-to-market decisions. This allows teams to evaluate trade-offs before launch, rather than reacting to performance after the fact.

Conjoint analysis is commonly used in marketing research to identify which features of a product or service are most appealing to a target audience. Studies may be conducted on existing products to improve messaging, refine positioning, or identify feature gaps that limit adoption. Conjoint analysis is also frequently used during early-stage research to evaluate product feasibility before development.

Most conjoint studies include demographic questions, such as gender and age, allowing results to be segmented by customer group. This segmentation often reveals meaningful differences in preferences that can be used to sharpen marketing strategy.

Conjoint analysis is excellent for pricing research because it forces respondents to make trade-offs. Rather than asking whether a price is “acceptable,” customers must choose between competing options, mirroring real purchase decisions.

In the ice cream example on this page, the highest utility was associated with a $5 price and a medium size. Without a conjoint study, it may have seemed logical to assume the large size should command the highest price. By modeling trade-offs, conjoint analysis revealed the optimal size-price combination based on actual customer preference.

Conjoint analysis is not well-suited for early-stage or poorly defined concepts where meaningful trade-offs do not yet exist. When products are brand new, features may be uncertain, and pricing ranges too wide for respondents to evaluate realistic combinations.

In these situations, simpler methods are often more appropriate:

• Use MaxDiff to identify which features or attributes matter most before designing trade-offs.
• Use Van Westendorp to estimate an acceptable price range before testing detailed pricing and feature bundles.

Conjoint analysis refers to a family of survey designs used to elicit trade-offs between product attributes. Types or methods of conjoint analysis primarily reflect differences in survey format, how concepts are displayed, and how participants interact with them, rather than differences in the underlying mathematics. Regardless of format, preference estimates are derived from regression-based models that infer part-worth utilities from observed choices.

Choice-based conjoint is the most common and widely used form of conjoint analysis. Respondents are shown a set of product concepts and asked to choose the option they find most appealing. Each set contains a subset of concepts drawn from the whole design, allowing preferences to be inferred from observed trade-offs. Because respondents select complete product profiles, this approach closely simulates real purchase behavior and supports reliable estimation of attribute importance and price sensitivity.

Best–worst conjoint, sometimes referred to as MaxDiff conjoint, asks respondents to identify the most and least appealing concepts within each set. This approach is practical when attributes may generate both strong positive and negative reactions. For example, certain features may increase appeal while others create clear deterrents. By forcing explicit extremes, best–worst designs can sharpen preference signals in these situations.

Adaptive conjoint builds on choice-based conjoint by tailoring subsequent choice sets based on earlier responses. Rather than presenting purely random combinations, the survey adapts to focus on regions of the design space most relevant to the respondent. This can improve engagement and efficiency, particularly for complex products with many attributes.

A full-profile conjoint presents respondents with many concepts at once and asks them to rate each on the likelihood of purchase. This method predates modern choice-based approaches and is generally considered outdated. Rating large numbers of concepts increases cognitive load, introduces scale bias, and often results in lower-quality data.

Ranking conjoint asks respondents to evaluate multiple concepts within a set using a rating scale or by allocating points across options. This approach is sometimes referred to as continuous sum conjoint. While more flexible than traditional full-profile methods, it still relies on stated preferences rather than forced trade-offs and is less commonly used than choice-based designs for pricing and product decisions.

Menu-based conjoint allows respondents to build a product by selecting multiple options, such as choosing a base model and adding features at incremental prices. While this can mirror real configuration behavior, it requires more complex survey logic and modeling. In practice, many modern studies achieve similar insight using adaptive or choice-based conjoint designs with follow-up questions, making menu-based approaches less common outside of custom implementations.

Often, preliminary data is collected before running a conjoint study. An initial survey may include a MaxDiff question to identify which features matter most, or a Van Westendorp question to establish acceptable price ranges. This early research helps narrow attributes and levels, reducing the number of concepts shown, lowering respondent fatigue, and improving data reliability.

Before launching the conjoint study, any custom data used for segmentation should also be reviewed and prepared. For example, if results will be analyzed by region or customer type, that data must be accurate, complete, and available at the time of survey design. When internal data is unavailable, segmentation questions can be included directly in the survey.

SurveyKing currently supports choice-based conjoint. To create a conjoint survey, open your survey and add a conjoint question using the Conjoint element. Define your attributes and levels, then choose how many sets and concepts to display and adjust any additional options as needed.

SurveyKing provides several options to customize a conjoint study:

• “None” choice: Adds an option in each set labeled “None,” allowing respondents to choose not to purchase any option. This reflects real-world decision-making and is enabled by default. Disable this option only when respondents must select an option, such as for certain government or mandatory services.
• Reset choices: Allows respondents to clear their selections and restart the conjoint question from the first set. This option should be used sparingly, as it can encourage second-guessing and reduce data quality.
• Require question: Requires respondents to complete all sets in the conjoint exercise before proceeding to the next page.

A well-designed conjoint question balances statistical power with respondent effort. As a general guideline, a practical study includes approximately 5 attributes, 4 concepts per set, and 5 to 10 sets.

• Concepts per set: Each set should typically display 4 concepts. Showing more than four increases cognitive load and makes it harder for respondents to make meaningful trade-offs. The total number of possible concepts is calculated by multiplying the number of levels across all attributes. For example, 4 flavors × 3 sizes × 2 prices results in 24 total concepts. Ideally, the total number of concepts should remain under 50, as larger designs become harder to analyze reliably.
• Number of sets: To avoid survey fatigue, respondents should generally see no more than 10 sets. In most cases, 3 to 5 sets provide a strong balance between data quality and completion rates.

Most choice-based conjoint studies require around 100 responses per segment for regression-based utility estimates to stabilize, yielding interpretable coefficients and attribute importance without excessive variance; larger samples improve precision, while smaller samples may be less reliable.

Conjoint analysis is an advanced research technique that uses a variety of unique terminology. To help you get a complete understanding, here is a list of commonly used conjoint terminology:

The high-level product features that respondents will evaluate are called attributes. Attributes are the first column in the above example question. That example has the following features: flavor, size, and price. If you studied a new car offering, you might have features such as color, make, model, MPG, and tire type. There is a limit of 20 attributes on the SurveyKing platform.

The items listed within an attribute are called levels. In the example, the "Flavor" attribute has levels of "Chocolate," "Vanilla," "Cookie Dough," and "Strawberry." When you create the conjoint survey, you define an attribute and the levels that go with each attribute. There is a limit of 15 levels on the SurveyKing platform.

Combining all your attributes and levels, which creates a hypothetical product, is called a concept. In the above example, concepts are the columns that respondents choose. Concepts are sometimes referred to as "cards" in statistical software. There is a limit of 7 concepts on the SurveyKing platform.

Also referred to as a task, a set contains multiple concepts or product offerings. Respondents will choose one concept per set and then be shown a new set of concepts. There is a limit of 20 sets on the SurveyKing platform.

This term is the most crucial in conjoint analysis. It defines how a respondent values each attribute level. When all the utilities for all respondents are analyzed, a researcher can determine an overall product value. Utilities are the output of a regression equation.

Utilities have no scale compared to other conjoint projects you run. They only matter in the context of the current question you are looking at.

Sometimes utilities are called "part-worths" or "part-worth utilities." We use the term "utility."

Conjoint analysis estimates preferences using regression-based models applied to observed choice data. Because conjoint attributes are categorical (e.g., flavor or size) rather than continuous, discrete-choice models such as logistic regression are used to estimate preference strength.

The output of these models is a set of coefficients commonly referred to as utilities, also known as part-worth utilities. Utilities do not represent a physical unit of measure. Instead, they reflect relative preference strength, with higher values indicating stronger preference.

For example, if respondents consistently choose product concepts containing “Cookie Dough” and rarely choose those containing “Vanilla,” the estimated utility for Cookie Dough will be higher than that of Vanilla. These utilities are then used to calculate attribute importance, showing how much each attribute influences overall choice.

To illustrate this, the ice cream example above was analyzed using a small sample of responses. The results below show the estimated utilities for each attribute level, along with the relative importance of each attribute.

A key benefit of conjoint analysis is the ability to evaluate how complete product concepts are likely to perform in the market. By combining attribute utilities, each possible concept can be ranked from highest to lowest overall preference. This makes it easy to identify both the top-performing configurations and weaker options. The table below shows an example of how concepts are ranked based on total utility.

Conjoint results are often more informative when analyzed by segment, such as gender, region, or customer type. By including segmentation questions before the conjoint exercise, preference estimates can be compared across groups to uncover differences that may be hidden in overall results, including:

• Differences in preferred features or attribute levels
• Variation in price sensitivity by segment
• Changes in attribute importance across audiences
• Differences in top-ranked product concepts

The utilities shown in the table are the output of the regression model used in conjoint analysis. Utilities represent relative preference strength, not an absolute unit of measurement. Higher utility values indicate a stronger preference than other levels within the same attribute.

Attribute importance is derived from the range of utilities within each attribute. Attributes with larger utility differences have a greater influence on overall choice. To illustrate, the highest utility level for each attribute is summed:

• Flavor: 14.11
• Size: 4.03
• Price: 5.06
• Total: 23.02

Each attribute’s importance is then calculated as its share of this total. For example, flavor importance is calculated as 14.11 / 23.02 ≈ 61%, indicating that flavor was the strongest driver of choice in this example.

By combining utilities across attributes, complete product concepts can be scored and ranked. This makes it possible to identify the strongest-performing configurations as well as weaker options, translating individual trade-offs into clear insight about what customers value most.

SurveyKing uses ChoiceModelR, an R-based statistical package, to compute conjoint utilities. ChoiceModelR estimates attribute-level coefficients via logistic regression with maximum likelihood. Because respondents evaluate only a limited number of concepts, utilities are first estimated at the respondent level and then averaged across respondents to produce stable aggregate results.

The regression output consists of coefficients commonly referred to as utilities (part-worth utilities). These values represent relative preference strength for each attribute level. ChoiceModelR uses effects coding, which constrains the utilities within each attribute to sum to zero. As a result, some attribute levels will have negative utility values. To improve interpretability, SurveyKing shifts utilities by a constant so the lowest level is set to zero. This transformation does not change relative preferences or rankings; it only affects presentation. In the ice cream sample output, the flavor utilities were shifted by 5.43 before being displayed.

The following parameters control model estimation within ChoiceModelR. All attributes, including price, are coded as categorical variables. While price can be modeled as continuous, categorical coding provides consistency across attributes and avoids imposing linearity assumptions on price sensitivity.

• Data Matrix – See this Excel file, which is the input for the ice cream example. The first row of each card set contains the card number selected (column G). In this example, the first card selected was 4, indicating the “None” option was chosen. When “None” is selected, the card number is recorded as the highest concept index + 1. This is the input required for ChoiceModelR. Other programs use a similar structure, except column G contains a “1” if the card is selected or “0” if not selected. In those cases, an additional row is added for the None option.
• R – The total number of iterations of the Markov Chain Monte Carlo (MCMC) chain to be performed. Default value: 4000.
• Use – The number of iterations used for parameter estimation. Default value: 2000.
• Keep – The thinning parameter defining the number of random draws to retain. Default value: 5.
• wgt – The choice-set weight parameter, with possible values from 1 to 10. This parameter only needs to be specified when estimating a model using a share-dependent variable. Default value: 1.
• xcoding – Specifies how each attribute is coded. All attributes are coded as categorical (value 0). While prices could be modeled as continuous, all variables are treated as categorical for consistency and ease of calculation.

SurveyKing includes Conjoint as a native question type in the standard survey editor, with no additional licenses required to get started. For teams that need Conjoint without enterprise research overhead, SurveyKing is a practical Qualtrics alternative focused on ease of use.

Conjoint analysis measures how people value different parts of a product and is ideal for launching new products and understanding market preferences. For example, when testing a new ice cream, it shows whether flavor, size, or price matters most and which combinations are most valued.

Use an online tool like SurveyKing. Upload your attributes (such as product color) and levels (the available colors), set up the choices, and collect responses through a web link or survey panels. Results are calculated automatically, showing which options and combinations people prefer most.

SurveyKing offers product research surveys starting at $19 per month, including Conjoint, MaxDiff, Gabor-Granger, and Van Westendorp, with results calculated automatically. Survey panels are available for validation, with pricing based on the number of respondents and targeting.