Bandit MaxDiff is especially valuable if you want to identify the top few items for the population from a list of 60, 200 or even 2000 items. It also can be useful for studies involving relatively few items, such as 15 items where you need good individual-level precision for all the items (via HB estimation) but where you want to increase the precision for the few best items.
If the main purpose of your study is to identify and measure the top few items for the population (or a segment of the population) from a list of eighty or more items, Bandit MaxDiff can save you 50% to 80% on your data collection costs. Depending your study, 200 to 500 Bandit MaxDiff respondents can give you the same precision for the top few items as 1000 standard MaxDiff respondents.
Because Bandit MaxDiff needs to keep track of previous respondents’ choices to build choice tasks effectively for the next respondents (by accumulating respondents' preferences in a database on the server), Bandit MaxDiff cannot be done via paper-or-pencil. And, Bandit MaxDiff will not work well for distributed data collection such as offline CAPI.