Do you need one bank at a time ?

400 items -(identifying criteria) that's peanuts. Grab and keep in memory

How many questions ? 500 ? 1000 ? Consider grabbing all of them at the start and putting them in a dictionary ( maybe one per bank, depoending on the context - or one dictionary containing dictionaries of banks ?)

Even a dictionary as a static property of a class might be considered

Why would you need the "statistical 'heavy-lifting' " in the back end ? .Net is fast enough

>As it works in VFP, the main bank is broken up into smaller banks, so I grab all the identifying criteria at the start and hold that in a temp table (around 400 items) in memory, once the next question is chosen - I just pull that question from the SQL DB. Most of the statistical 'heavy-lifting' is done on the VFP side, which I'll need to convert to the back end - if I can.
>
>>>
>>>Can't - it's an adaptive test, I don't know what the next question is until they answer the last one.
>>
>>Any way to make some educated guesses ? Fetch all the possible next questions - discard the ones you don't need
>>
>>How many questions are there possible in one test ? Bring them all in
>>
>>
>>Keep a pool of questions in a dictionary. If it is in, take it from there, if not fetch it