We would like to conduct a MaxDiff study with a total of 17 items, with the goal of obtaining an importance ranking of these items, both at the aggregate level and at the respondent level.
Data collection will be done via telephone interviews, which limits the number of items per task. Therefore, we want to keep the design to 3 items per screen. 17 items per respondent may be too many, so we are considering showing each respondent only a random subset of 12 items.
We are considering implementing this using Express MaxDiff  where not all items are shown to each respondent.
My questions are:
Is Express MaxDiff an appropriate approach in this case, even though it is typically used for larger item sets (e.g., >40 items)?
With this setup, can we still obtain respondent-level output?