Conjoint Types

As mentioned in the previous chapter, conjoint studies come in many flavors. The number of attributes or even the type of attributes (discrete vs. continuous) between each conjoint can change. Some conjoints have prohibitions (e.g., not showing Blue and Ferrari as levels in one profile), and sometimes the sheer number of attributes / levels is just too much to show.

All this is to say that the study type influences the Experimental Design (plan of the experiment) algorithm you used to craft the tasks shown to respondents. However, the majority (if not all) of the Choice-Base Conjoint studies can be categorized into one of three buckets:

• Full-Profile
• Partial-Profile
• Alternative Specific Design

In this Google Sheet example you can find a few examples of conjoint cards (and model specification) for all three major conjoint types. It it best accompanied by the video review below.

The video has 3 sections, one for each of the Conjoint Design Types.

In the Full-Profile design, all attributes are displayed on each conjoint card and are common for every concept present in the design. There may be certain prohibitions (like mentioned above, do not display Ferrari and blue in single concept), but usually the number of prohibitions in a full-profile design is small.

The Partial-Profile designs are useful when there are a lot of (like 9 or more) attributes. They are used to reduce cognitive burden, and mitigate the fact that respondents usually use Lexicographic-Scoping (they tend to take decision based on the levels of only one or two attributes in the whole set).

A Partial-Profile only shows a subset of attributes in each profile. They typically work best when the number of levels within each attribute is low (around 2-4). You can find an example in this Google Sheets example of Partial Profile Cards.

This type of design comes in handy when there are a lot of prohibitions, or even better if the alternatives have completely distinct attributes/levels. A classic example is pricing studies, where there are only two attributes; (1) Brand (or Product) and (2) Price.

Since respondents usually have different perceptions of price for each of the brands/products, we define them as separate attributes in the model specification. It happens quite often that all the brand/products have different price ranges that we want to test, so instead of using just one price attribute with multiple prohibitions, we set price attribute separately for each of the brands. You can find a graphic representation of this (and another example of an Alternative Specific Design) in this Google Sheet example.

Usually, we use only one None alternative (regardless if it is a Single Response None, or Dual Response None). However, in some specific cases, there could be more None alternatives. The simplest example of when we have more than one None alternative is: I would not buy at all and I would buy something else. More sophisticated (yet somewhat unrealistic) examples are shown in this Google Sheet.