Large-Scale Conjoint
Large-Scale conjoint is about making conjoint analysis look and feel like much larger e-commerce systems, where products have a large number of attributes and where there is textual noise and variation. To do this means updating a number of elements of conjoint analysis from what is shown to the way people choose, to the analysis required.
On LinkedIn we have been running a discussion about Large Scale Conjoint analysis. In the real-world people look at hundreds and thousands of products and posts on systems such as Amazon, or booking.com or Instagram, picking and choosing what to look at and ultimately what to buy.
Traditional conjoint, on the other hand, tends to be small and relatively simple, but because of the similarity of levels between products it also feels repetitive and complex, despite only having a small number of attributes.
The idea of Large Scale conjoint is to make a conjoint system that is more natural and more real and that improves participant experience by borrowing ideas from e-commerce, and techniques like hot-cold scales (thumbs up, thumbs down ratings) and selection and rejection elements.
The more difficult area is analysis and the attached document is our discussion with Claude about the possibilities (and code) for creating an updated Hierarchical Bayes approach to handle Large Scale Conjoint.
- Large Scale Conjoint - a discussion with Claude
- 📎 (view)...
