CBC Training at SKIM Conference
On Wednesday 26 and Thursday 27 May SKIM and Sawtooth Software hosted two days of training sessions on Discrete Choice Modeling (DCM) in Germany (Köln). My colleague Mike (from R&D) and I were fully immersed in the ins and outs of Choice-based conjoint (CBC), Adaptive CBC and MaxDiff.
The first session I attended was by David Lyon, who is an authority on survey-based pricing research. In his very interesting session he showed us all methods that can be used in pricing research, and with a critical view discussed for each one of them, detailing both the pros and the cons.
The second session was about Maximum Difference Scaling (MaxDiff), a methodology widely used here at InSites Consulting. Aaron Hill from Sawtooth Software clearly outlined for which research problems MaxDiff can be used. At the end of his session he showed two new exciting add-ons to MaxDiff: TURF and the MaxDiff simulator.
In the last session on Wednesday David Lyon talked about weighting the data: principles, perils and pitfalls. In a lot of market research studies data are weighted, but the main lesson of the session was that weighting is not without implications as it can create more variance in your sample and make your sample less accurate. A very interesting and eye-opening session!
On Thursday both sessions were on Choice Based Conjoint (CBC). In the morning Richard Neggers from SKIM showed three extended case studies. As the session was very interactive it gave a very good opportunity to enhance my CBC knowledge. In the afternoon Sawtooth guru Bryan Orme and David Lyon talked about advanced CBC examples. It was amazing to see how complex CBC designs can be. Bryan showed us some power tricks which will definitely be useful in our next CBC projects.
Mike joined me in the training sessions about choice-based conjoint. It was truly an interesting and informative experience. Below you can find Mike’s thoughts on his sessions.
The sessions I will focus on were both run by Bryan Orme, a true “guru” of CBC methodology. He has been working on CBC modeling for more than 15 years, and is very well known in the world of conjoint analysis. The first presentation focused on a new type of CBC methodology: Adaptive CBC (ACBC for short). This is a new type of CBC application that can adapt to the choices consumers make at the beginning of an exercise. It looks to be a very interesting technique, and one that can be more interesting for respondents, as well as being less reliant on the assumption that consumers rationally consider every aspect of a given product when deciding whether to purchase it.
Another very interesting session I saw was also run by Bryan Orme. This session focused on the background of conjoint analysis and hierarchical Bayes modeling. In this session, we learned a number of tips and tricks to take CBC analyses to the next level of complexity. Topics included the use of covariates, probing, testing and interpreting interactions, and the use of constraints when estimating CBC models.
All in all, it was a very informative event and all of the information I learned will come in handy when working on upcoming CBC analyses!