Conjoint Value Analysis (CVA) is a legacy conjoint analysis software based on the original ratings-based conjoint approach from the 1970s. It is not often used today (~2% of total conjoint analysis projects as reported by Sawtooth Software users), but has some advantages for certain research situations. CVA can display either one or two products at a time. It may be useful for both product design and pricing research, when the number of attributes is about six or fewer. CVA includes an excellent designer for traditional conjoint analysis that some users employ for general design of experiments or for specific applications such as MBC.
Before choosing the CVA system for your project, we recommend you discuss your research needs with a Sawtooth Software content expert to determine that it would be appropriate.
The CVA System adds traditional full-profile conjoint analysis to the suite of Sawtooth Software conjoint products. CVA is useful when measuring interactions is not a priority and when sample sizes cannot justify the use of CBC. CVA questionnaires may be administered via web-based, CAPI (computers not connected to the web) or paper-and-pencil surveys.
The CVA system provides the tools to design either single-concept (ratings-based or ranking) or pairwise comparison questionnaires. Since full-profile conjoint studies ask respondents to consider all features at the same time, CVA is appropriate for measuring a limited set of attributes (generally about six or fewer).
CVA's Designer helps you create the conjoint interview. The CVA software advises you regarding how many conjoint questions you should ask to obtain enough information to estimate respondent preferences for the attribute levels (utilities). Once you specify the number of conjoint questions to use, CVA's computer-search routine finds a set of questions (product profiles) with high relative D-efficiency. D-efficient designs lead to efficient estimation of utilities.
The base CVA package provides two options for calculating part worths. For ratings-based designs, ordinary least squares (OLS) is appropriate, though the included HB estimation option is preferred. For ranking tasks, monotone regression is provided.