How the Choice-Based Conjoint experimental design works

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 choice set. Generally speaking:

• Fewer choice sets, and fewer profiles per set, are better for participant experience. This minimizes decision fatigue.
• An attribute level needs to appear frequently enough to calculate or infer a comparison of it against other levels.
• If an attribute level appears too few or too many times, the experimental design of the Choice-Based Conjoint question is considered imbalanced. Therefore, the data could be unreliable.

To minimize participant fatigue and ensure that the data being collected is thorough and reliable, Alida's conjoint solution:

• Calculates a minimum number of required responses required to report on a Choice-Based Conjoint solution.
• Randomizes profile generation and choice set generation to ensure participants see a variety of attribute levels.

As you edit attributes and levels and make your selections for Number of Cards per Set and Number of Sets Shown to Each Respondent, the information panel at the bottom of the page updates to display the minimum number of responses you need to collect for a valid Choice-Based Conjoint Report. If you do not meet the minimum required number of responses, the Choice-Based Conjoint Report cannot be generated.

The minimum number of required responses for a Choice-Based Conjoint question is determined by the following formula:

Minimum number of required responses = (1000 * largest number of levels across all attributes) / (number of rendered choice sets per participant * number of profiles per choice set)

If you show individual participants more profiles and choice sets, the minimum number of required responses decreases. Conversely, if you show individual participants fewer profiles and choice sets, you need to collect more responses.

The trade-off between minimizing participant fatigue and minimizing the number of required responses is a fine balance. The goal is to maximize coverage across all levels and ensure sufficient data has been collected for comparisons. At the same time, you want to avoid mentally overloading participants. Participants may be more overwhelmed and provide less reliable responses as they proceed through lots of choice sets, each one featuring many profiles.

The profiles are generated by randomizing the levels to display for each attribute. If exclusion sets are specified, the randomized profiles will adhere to the exclusion set rules.

Randomization helps avoid duplicate profiles being presented in the same choice set. Randomization also guarantees the collection of data across all levels to make you have the number of responses required for the Choice-Based Conjoint analysis. What's more, randomization helps avoid any possible design bias and unforeseen interactions between attributes when the attributes are treated as statistically independent (or orthogonal).

The choice sets are generated by randomizing the profiles displayed in a choice set. Randomization helps avoid duplicate choice sets being presented to the same participant. It also ensures uniform distribution of choice sets across participants.