I could use a few general pointers on the best way to handle a large, complex (as in, topically dense) ACBC with an alternative specific design. The study has a total of 13 attributes (including conditional attributes). Many of these attributes have 5-6 levels, so the number of possible combinations is quite large. Four of these attributes are price, with two price attributes shown for each product configuration (conditional to the alt specific primary attribute). The price variables are in discrete levels, instead of a summed continuous.
When I test the design, "as is", the design report is blazing red (no surprise), with most levels shown a minimum of 1 time. I know I could boost the number of screeners and tasks, but I'm already worried about the demands placed on respondents.
Assuming I can't get the client to reduce the number of attributes and levels, what is the next-best option?
Constructed attributes and/or constructed levels based on questions prior to the ACBC?
Increasing the sample size and making do with levels with a minimum of 1 exposure (Is there such a thing as a sparse ACBC?). We currently have n=200 budgeted, which is way lower than I feel comfortable with. I could push for more sample. (This is for a niche application)
Other options?
If #1 is the way to go, is there a paper that describes how this is done? I looked in the ACBC technical papers and didn't find one.
Thanks!