Yes, that's exactly what I'm talking about. And what I'm thinking.

When a user selects options 1,4,7,9, and 15 on Monday, either:
a) store it as 5 separate insertions of update_id - option pairs
b) first join them into a delimited string and do one insert

Then on retrieval, either:
a) select all rows with Monday's update ID (which means a while loop)
b) select a single row and split the retrieved string

I expect the first reaction to be "let the DB do its work; it's undoubtedly smarter and faster", but I suspect that the counter-intutitive ins in fact true: that the extra DBI layer, with its preparation and the retrieval loop, is going in fact ot be slower than a pure join or split, especially when it gets more complicate with multiple questions to store and therefore multiple queries to do.