We recommend you compose your attributes and levels list so that each level of each attribute can freely combine with each level of each other attribute without leading to absolutely absurd or obviously impossible combinations.
Respondents are often able to imagine many possibilities that we as researchers (or especially managers at the client company) may think are quite unusual or even impossible. Even though it may seem counter-intuitive, it is usually better to show respondents such unusual combinations because it leads to more precise estimation of preference scores for each attribute level in your study. Of course, during analysis with the what-if simulator, you can avoid specifying products with these unrealistic combinations.
However, if there still is a particular combination of levels that is absolutely absurd or utterly impossible, despite your best efforts to create an attribute list where all levels can freely combine, then you may have to specify a Prohibition.
Warning: prohibitions lead to lower precision of your preference scores, so we recommend you avoid them whenever possible. If you try to specify too many prohibitions, the software will warn you and not allow you to proceed. Specifying prohibitions also may lead Discover-CBC to increase its recommendation regarding the Number of CBC Questions to ask.
Two-Way Prohibitions
Prohibitions can be between two attributes, for example: Prohibit Brand A from occurring with Purple.
To specify a prohibition, click Add Prohibition. Click the Attribute and Level that should be prohibited with another Attribute and Level.
Three-Way Prohibitions
Sometimes, prohibitions can be between three attributes, for example: Prohibit Brand A with shape Round and color Purple. To add a third attribute to the prohibition, click Add Attribute.
With this example, the 3-way combination of {Brand A, Round, Purple} is not allowed. Three-way prohibitions are actually less damaging statistically than two-way prohibitions.
Up to N-way prohibitions are allowed, where N is the total number of attributes in your study.