I'm looking for literature or advice on how to proceed and what pitfalls to consider in the following situation:
Client product has over 20 attributes (2-5 levels each) which can be grouped into composite attributes like "service", "user-interface", "performance", etc.
The plan is to use MaxDiff to get preference order across entire range of attribute levels. Then a CBC shows product concepts consisting of different levels of composite attributes "service", "performance", etc. where each level is a combination of levels from the corresponding (sub-)attributes.
The idea is supposed to work similarly to "Dual Conjoint" (http://www.sawtoothsoftware.com/download/techpap/acaorcbc.pdf --> I've Pinell referenced there but am unsure about the exact implementation steps) where granular utilities are found via ACA (or here MaxDiff) and then used/scaled) to explain performance/utility of composite attribute (abstract).