Classification of conjoint analysis
We are often asked what types of conjoint analysis exist and which ones we offer on Conjoint.ly. Here is an opinionated classification of conjoint analysis that helps you understand what some experts are talking about.
By response type
Rating-based conjoint: Respondents are asked to rate the product alternatives they are shown. This can be on a scale of 0 to 100. Respondents may be required to allocate scores so that the scores sum to a certain number (e.g., all scores in each question must add up to 100).
Best-worst conjoint: Respondents are asked to indicate which option is best and which option is the worst among three or more alternatives in each question.
Ranking-based conjoint: Respondents are asked to rank alternatives from best to worst. It is similar to best-worst scaling, but respondents also need to allocate rankings to the intermediate alternatives.
Choice-based conjoint (CBC): Respondents are asked to choose which option they will buy or otherwise choose. This is the most theoretically sound, practical, and popular method of conjoint analysis. Conjoint.ly proudly offers only CBC because other response types are known to be inferior for practical market research.
By questioning approach
Standard conjoint: In standard conjoint, the questionnaires are developed before they are sent to participants. They are carefully designed by using sophisticated algorithms to ensure best quality analytics, including segmentation analysis. Conjoint.ly offers standard conjoint.
Adaptive conjoint (ACA; when it is choice-based, it is also called ACBC): In adaptive conjoint, the questionnaire is constructed during the interview. It “adapts” to participants’ responses to optimise a certain parameter (such as minimising confidence intervals for certain partworth utilities). It is a more complicated technique, which may generate problems with certain types of analysis, such as in segmentation. It does have certain valid applications (such as reduction of the required sample size in very complex studies) – please reach out to us if you require assistance with this.
By type of design
(generic or unlabelled design): Generic conjoint allows you to understand which features and price levels drive customers' choice. It is used for studying either:
- a single brand, or
- commoditised products, where product characteristics do not vary substantially by brand (in industries such as agriculture, pharmaceuticals, furniture, bottled water).
- Brand-specific (alternative-specific conjoint, alternative-specific design, ASD, or labelled design): Brand-specific conjoint helps you understand what drive customers' choice differentially for different brands. It is suitable for studies that include a variety of brands, where potential product characteristics vary across brands. Example applications: FMCG, telecommunications, and home appliances. Unlike in generic conjoint, when you set up the study, under “Basic settings”, you need to specify “Applicability of levels across brands”.
Conjoint.ly offers both types of designs. The features available for the two types of conjoint currently differ:
|Partworth utilities||Yes||Yes, differentiated by brand|
|Brand performance||Yes (if brand is included as an attribute)||Yes (separate report)|
|Ranked list of product concepts||Yes||Yes|
|Preference share simulation||Yes||Yes|
|Marginal willingness to pay||Yes||No|
|Segmentation and profiling of segments||Yes||Yes|
|Excel export of raw data||Yes, including individual coefficients and model matrices||Yes, including individual coefficients and model matrices|
By whether all attributes are shown in every question
Full profile: In full-profile studies, all attributes are shown in every choice set. It is recommended that the number of attributes is limited to about six because it is hard for respondents to digest more information.
Partial profile: In partial-profile studies, only a subset of attributes is shown in each choice set. For example, the study may include 12 attributes, but only 6 will be shown in each question. This is a useful technique when you need to select different features for your product.
Two-stage conjoint (also called dual response): In two-stage studies, respondents are first given a choice of products they would buy, and after making a selection they are asked if they would consider buying this alternative at all. This questioning technique leads to a slightly more realistic estimate of respondents’ preference not to buy an item. On Conjoint.ly, dual response option is enabled when you specify that you are investigating a new product that customers are not used to buying.
Availability design (cross-effects): If there are reasons to believe that the presence of one brand affects people’s preferences for other brands, then there is an “availability effect”. Special considerations are required in the design of the questionnaire and analysis of the data. In practice, only in rare cases researchers observe availability effects.
Interaction effects: If there are reasons to believe that preference for a product is not simply the sum of preferences for its levels, but rather that certain levels have different likeability when combined with each other, then there may be an “interaction effect”. For example, furniture buyers may prefer lower prices and more exotic types of wood, but when a very low price is combined with an exotic wood material, they may in fact dislike the product (for example, because the combination is perceived as too “cheap” or incongruent). However, research suggests that in the vast majority of studies, interaction effects are minimal.
Our team at Conjoint.ly can help you with any type of customised conjoint analysis, even if it is not offered as part of our online tool. If you require a two-stage, partial profile, or any other type conjoint, please do get in touch with us.