5. Our advice

The number of attributes and modalities to be considered in a conjoint analysis must remain limited to avoid having to carry out an overly complex experimental design. Even if adapted experimental designs can be used to take several dozen attributes into account, their use requires expertise. Experimental protocols for collecting preferences, based on choices rather than ratings, are also very effective for dealing with large-scale problems. In general, conjoint analysis does not take into account the effects of interactions between attributes, although it is possible to do so with a suitable experimental design.

To detect trends, conjoint analysis is generally carried out on large customer panels, with several hundred subjects. It is well suited to the functional/utilitarian dimensions of products. It lends itself poorly to dealing with the subjective, symbolic/emotional...