As many of you know, we’ve been working hard on a Software-as-a-Service (SaaS), streamlined version of CBC. This will be an attractive, extremely easy to use version of CBC that is initially targeted for university classroom settings. We plan to have a beta version for early adopters before the end of the year.
This new CBC software involves new algorithms to make it faster for students to work in the collaborative environment of web servers: a rapid experimental designer (that’s actually just as good or better than SSI Web’s experimental designs for CBC) as well as an extremely rapid short-cut Bayesian utility estimation routine (that’s nearly as good as the full HB).
New to this application of CBC is the inclusion of attribute level rating questions, where respondents first rate attribute levels for brands, colors, styles, etc. (those attributes that are non-ordered) using a 4-point scale. Those relative ratings are used as customized utility constraints at the individual level. We also use them to avoid dominated concepts within the subsequent CBC choice tasks for each individual (meaning that product concepts that are logically inferior to another product concept will not appear in the same choice tasks).
We just completed a major comparison study this week for the SaaS-CBC vs. SSI Web-CBC, finding that the SaaS-CBC does just as well as the standard SSI Web version of CBC. Read the white paper to see the results (described in Appendix B of the paper).
We'll be discussing these developments further and showing off pieces of the software we've already completed at the upcoming Sawtooth Software Conference 2013 (in Dana Point, California, October 16).