Create a Choice-Based Conjoint question

A Choice-Based Conjoint question presents participants with different sets of product profiles, and asks participants to choose the most appealing product profile from each set.

Note:

• Participants cannot skip choice sets when answering Choice-Based Conjoint questions. Incomplete answers are excluded from reporting.
• Choice-Based Conjoint questions can be lengthy and fatiguing for participants. Therefore, only one Choice-Based Conjoint question per survey is recommended.
• After you publish the survey, you can still make changes to question text, attribute and level text, and images. However, you cannot add or remove attributes, levels, or exclusion sets.
• This feature is available at an additional cost. For more information, please contact your Account Representative.

Sample modern view:

Sample classic desktop view:

Sample classic mobile view:

1. From the Survey Builder Toolbox, under Questions, do one of the following:
   • Drag the Choice-Based Conjoint question to the Table of Contents.
   • Click the Choice-Based Conjoint question.
   
2. In the Table of Contents, click the Choice-Based Conjoint question.
   Result: The Edit Pane appears.
3. Optional: Edit the Question name.

   Question names:

   • Should be unique.
   • Can contain symbols and all alphanumeric characters.
   • Are not displayed to participants.
4. In the Question text field, enter the question text.
   Note: The recommended character limit is 2500.
5. Add your attributes and levels.
   For both attributes and levels, 40 characters is the maximum. You do not want the profile to have too much text. Longer text can be overwhelming for participants as they evaluate multiple profiles at the same time. It dissuades them from reading carefully and making an informed selection.
   1. Type the Attribute Text.
   2. Type the text for the Levels.
   3. To add an image to a level, click Add Image.
      

      Browse for the image, type the alternative text, and click Add. Images are automatically resized to 90 pixels by 130 pixels in responding.

   4. To delete an attribute, click Remove.

      This function only appears when there are 2 or more attributes.

   5. To delete a level, click the corresponding delete button.
      

      This function only appears when there are more than 2 levels.

   6. Click Add Level to add more levels.
   7. Click Add Attribute to add more attributes.

   There is a fine balance to how many attributes and levels a Choice-Based Conjoint question can have. You do not want to overwhelm participants with too many attributes, levels, and profiles to evaluate. At the same time, the question needs to be information-rich enough to gather meaningful results, and ensure the comparisons being made are valid.

   Generally, you should follow these sizing recommendations:

   Attributes	2 is the minimum. 6 is the maximum. 3-4 is recommended.
   Levels	2 is the minimum. 6 is the maximum. 3-5 is recommended.

6. Optional: If necessary, add an exclusion set.

   An exclusion set represents a combination of levels that will never be shown to participants.

   Ideally, you should avoid using exclusion sets whenever possible. Comparisons are at the root of conjoint analysis, and having enough comparisons across levels to build a rich and thorough data set is crucial. Therefore, each exclusion set introduces a small data gap. The more exclusion sets you have, the more your experimental design and your response coverage across levels are affected.

   Because of this, having a lot of exclusion sets can interfere with the accuracy of the calculations in reporting. What's more, it reduces the flexibility and accuracy of the conjoint market simulator. The conjoint market simulator uses Hierarchical Bayes (HB) analysis to borrow strength across multiple levels and hundreds of profiles, choice sets, and participants to fill in the data gaps and make predictions about preference share. If there are a lot of exclusions, the ability of the HB analysis iterative algorithm to borrow strength, fill in the data gaps, and make educated guesses is limited.

   If you find yourself adding complex exclusion sets, or exceeding the exclusion set maximum, consider redesigning the attributes and levels in your question so you rely less on exclusion sets to achieve the desired result.

   Exclusion sets

   5 exclusion sets is the maximum. 2-6 rules per exclusion set.

   Note: If you make changes to attributes and levels as you are authoring, remember to review your exclusion sets as well.
   1. Click Add Exclusion Set.
   2. Set the Attribute and Level values.
   3. Click Add exclusion to this set to add a new level.
   4. Click Remove to delete a level.
   5. Repeat steps a to d to add more exclusion sets and rules.
7. Select the Number of Cards per Set:
   • 2
   • 3
   • 4
   This option controls the number of profiles or concepts shown to participants in each choice set.
8. Set the Number of Sets Shown to Each Respondent.
   This option controls the number of choice sets shown to participants in the Choice-Based Conjoint question. You can set a number between 1 and 30. A number between 8 and 10 is recommended to avoid participant fatigue.

As you make your selections, the information panel at the bottom of the page updates to display the minimum number of responses you need to collect for reporting. If you do not meet the minimum required number of responses, the Choice-Based Conjoint Report cannot be generated.

9. Select or clear Include a "None of these" option as the last choice.

   Generally, a None option is useful if you want to test for product adoption and how new concepts will perform in the market. Including a None option will give you valuable information about how acceptable concepts are as a whole. If none of the concepts are optimal or acceptable (particularly if this is a new product category), that is good information to have.

   If your goal is research and development and learning about the best package you can assemble, then not including a None option is recommended. This prompts participants to think more carefully about the trade-offs they must make between levels and profiles, which will give you better part-worth data.