Conjoint Analysis Consumer Preferences Conjoint Model

Best Tips for Sound Conjoint Analysis Design


Developing a new product can seem like an overwhelming task: Brand, color, price, and a host of other features all impact whether consumers choose one product over another. Therefore, predicting which features are most influential to consumers is key to success in a competitive landscape. However, product features something that should be left to chance or intuition. In situations where there are many attributes and combinations of attributes to consider, one methodology stands above the rest: conjoint analysis.


What is a Conjoint? 


Conjoint analysis (or simply conjoint) is a tool to understand the desirability or preference of various features (or levels) and categories of features (or attributes) that make up a given product or service. When compared with other methods, a conjoint provides a cost- and effort-efficient way to predict consumer preference of a potential product by evaluating the impact that changes in individual attributes and levels have on the desirability of different product profiles. 


Depending on the specific product, the attributes used within a conjoint can be anything from name or color to RAM size or annual percentage rate. Including some product attributes allows for increased versatility and utility in conjoints.


From the standpoint of a participant in the survey, a conjoint begins with a brief description of what to expect over the course of the conjoint module. In most cases, this will be a series of three to five consecutive screens asking respondents to review two or more product or service profiles populated by the conjoint algorithm from a predetermined set of attributes and levels. On each screen they will be asked to choose their preferred configuration of features from the ones displayed. This could be for anything from different loadouts for a computer or car to different rates and benefits for a credit card or health insurance plan.


When and Why to Use a Conjoint? 


Conjoints are powerful analytical tools for predicting how consumers will behave and have versatility and many benefits. However, that doesn’t mean they are always the correct fit for the job. 


Here are some things to consider when deciding whether a conjoint is the right tool for your survey.


Life Cycle Development Stage

The first thing to consider when deciding whether a conjoint is right for you is where you are in the product development lifecycle. While robust, conjoints cannot test an infinite number of features or categories of features. Therefore, conjoints are best utilized once you have already decided on a finite number of product attributes and a finite number of levels to test. Usually, the standard choice-based conjoint analysis will have between three to six product attributes with three to six levels within those attributes. 


If you don’t know which categories of features and features are most important to your market, or have a lot more than six attributes and levels, it might be too early for a conjoint. Instead, you might benefit more from in-depth, qualitative interviews or a more exploratory market landscape survey containing a MaxDiff. This is because we want to be as specific as possible with the attributes and levels we include in a conjoint.


Product Features

The next thing to consider is whether the product you are looking to develop is impacted by multiple features being judged as a whole. Conjoints are best suited for products (such as computers, appliances, etc.) or services (such as credit cards, streaming subscriptions, etc.) that have many characteristics that can reasonably be considered impactful on consumer choice.


For example, if you are primarily interested in how consumers prefer or rank a list of features in one category or which categories overall are most important to consumers, a conjoint is probably not the best tool. In these cases, a MaxDiff or other simple rank-choice survey tool will likely serve your needs better. 


In situations where the category you are most interested in evaluating its impact is price, and you don’t plan on offering many different product configurations, you will likely find a Van Westendorp, Gabor Granger, or simple pricing study more useful. 


Market Share Limitations

One final thing to consider is that while versatile, conjoint analysis is primarily designed to understand consumer preference. This means that while it cannot determine the final market share, it can predict the theoretical market share that any new product or service will be able to capture. 


Setting up a Conjoint


Step 1: Determine the Levels and Attributes

The success of a conjoint analysis depends on many factors. One key component every conjoint needs to get right is the initial input levels and attributes. After all, for a conjoint to be practically useful, it should simulate a product or service that can potentially be presented to consumers or other target audiences. Additionally, a single conjoint analysis has a finite number of levels and attributes that it can model.


In some projects, the input levels and attributes are easily determined by external factors. For example, a computer company might only have a fixed number of GPUs or hard drives it offers or that work within certain technical configurations. A bank offering a new credit card might only have a small range of rates that make financial sense to include. 


However, many novel product and service offerings will have more possible features than could usefully be included in a single conjoint analysis. In these cases, it is often necessary to begin with a pilot study to narrow the elements that will be included in the final survey. 


Pilot studies can take different forms depending on the specifics of each project. For example, in-depth interviews are useful for projects where well informed respondents can shed light on the larger population. Other discrete choice exercises, such as MaxDiffs, are common for projects where individual respondents are unable to shed light on other respondents. Regardless of the form the pilot study takes, the goal is to reduce the number of inputs that will go into the final conjoint analysis.


Step 2: Choosing the right Conjoint

The most important things to consider when deciding on the appropriate conjoint methodology to use is which type best fits the product you are developing and which most accurately resembles the decision-making process your consumers will go through when buying the product or service. For example, if you plan to offer only pre-configured phone plans, then a choice-based conjoint is going to work better for you. If consumers will be able to choose which elements they want in their phone plan a la carte, then a menu-based conjoint might suit your needs better. 


Below are a list of the basic types of conjoints along with some of their pros and cons:


Conjoint study example

Mobile Phone Manufacturer A is interested in bringing a new phone to market and needs to understand which features are most important to consumers when they purchase a new phone. Through previous qualitative consumer research they discovered that their phone’s four most influential attributes are storage size, camera quality, color, and length of warranty. To best understand how these attributes and their levels will affect consumer preference they decide to conduct a choice-based conjoint analysis.


Their first step in designing this conjoint will be determining the different levels they want to test for each attribute. This is often called an attribute map. For example, for the attribute storage size, Mobile Phone Manufacturer A may choose to include 100 gigabytes, 250 gigabytes, and 500 gigabytes as the levels. This step is then repeated for each attribute. It is important that the levels are specific and well differentiated to get the best results. A level with a range of 250-500 gigabytes likely won’t generate a clear, actionable response. Each product attribute does not need to have the same number of levels.


Below you can see an attribute map based around this phone example:


Attribute Map:


Storage size


Camera quality 

Warranty length

100 Gigabytes


5 Megapixels

6 Months

250 Gigabytes


10 Megapixels

1 Year

500 Gigabytes


15 Megapixels

2 Years

1 Terabyte


20 Megapixels



The next step for Mobile Phone Manufacturer A is to create a balanced design for the conjoint. This means designing the conjoint exercise so respondents see each individual level an equal number of times over the course of the module. A well-balanced conjoint will usually show between 8-15 choice tasks to each respondent. This number ensures that the levels are shown the optimal number of times without fatiguing the respondents. The number of choice tasks each respondent will need to complete in order to generate good data is determined by the number of attributes and levels, as well as the total sample size. 


Below you can see an example of both the attribute template and two filled-in profiles that a respondent might be asked to select from:


Conjoint Template


Phone #1


[Storage Size]

[Camera Quality]

[Warranty Length]


Conjoint Question Example with Filled In Template:


Please select which of the following phones you would be most likely to purchase based on the profiles below.


Phone #1


250 Gigabytes

5 Megapixels

1 Year


Phone #2


500 Gigabytes

20 Megapixels

3 Years


Step 3: Fielding your conjoint and interpreting the results

Once you’ve settled on your attributes and levels and decided the appropriate variation of conjoint to use, it’s time to field your research study. Like all survey projects, which platform is appropriate for hosting your survey and which panel vendor is suitable to provide a sample relies heavily on your research, budget, and level of research experience. For example, while some sample providers specialize in reaching specific target populations, such as healthcare professionals, others are more adept at reaching out to the general public or specific consumer types. 


After data collection is complete and each attribute and level is shown to the full sample a predetermined amount of times, researchers and data scientists will analyze the relative importance of each attribute and level in determining whether a specific configuration was preferred by respondents. With the information gained from this, product designers, marketers, and other professionals will know which product attributes, levels, and configurations are most preferred, allowing them to take their next steps with confidence.




Regardless of whether you are designing a new phone or rolling out a new messaging campaign, a well designed conjoint can act as the gold standard and help you make consumer-driven, strategic market decisions.


Does your organization need help identifying your consumers’ product preferences? Gradient would be happy to help you with your conjoint analysis. 

Jordan Boeder

Written by Jordan Boeder

Jordan received his master’s degree and Ph.D. in Developmental Psychology from Claremont Graduate University. After receiving his Ph.D., he worked as a Post-Doctoral researcher at the University of Zurich where he honed his skills in Bayesian data analysis. Jordan uses his years of teaching experience to help distill complex research findings into simple insights.

Related articles: