Looking for something like a volumetric MaxDiff

Question

Recently we worked on our first volumetric conjoint.  The project isn't completed yet, but it is going well so far.  It was a simple conjoint with just the product items as one attribute and price as the other.

During the set-up for that project, we had some conversations about how price wasn't really likely to change and whether we could simplify the model by leaving it out.  Of course, that wasn't possible since we can't have just one attribute in a conjoint.  But it seemed to me that there must be a way to gather and use data in this manner.

I have looked around and tried to find an existing methodology that does this, but have been surprised that I can't find something.   I feel I've probably missed something obvious, but I'm hoping someone could point me in the right direction.

To summarize the research problem, imagine wanting to do a volumetric choice-based exercise when you have a "product" attribute that has maybe 45 levels.  Ultimately you want to be able to run a conjoint-style stimulator to predict market share given a particular sub-set of the 45 products actually on the shelf.

Answer 1

Andrew, I think I'd set this up as something like a volumetric MBC experiment.  The 45 product items appear in choice sets or not according to an experimental design and the respondent task is to choose some quantity (0 or more units) of each of the available product items.  This isn't built into our MBC software but it's a custom analysis that could be run from such a design.

Comments

Thank you.  Menu-Based Conjoint is one area that I haven't used before, and the similarity between thinking of the shelf of products vs. the menu wasn't something I had connected.

I assume your answer is telling me that I can do MBC with just one attribute (the 45 product items).   That's great.

Too bad that your software doesn't manage this kind of analysis directly.  Are you aware of any articles or papers that talk about how this analysis is done?

---

Andrew, I think of this as being more a study with 45 binary (on/off) attributes.  Back in the day we'd have called this an availability cross-effects experiment and the classic reference here is Anderson, D.A. and J.B. Wiley, 1992, “Efficient Choice Set Designs for Estimating Cross-Effects Models,” Marketing Letters.  The volumetric aspect adds a bit of challenge but I don't think it's insurmountable.