Sorry, when I said it's not significant I meant it is but it doesn't trump the process issues. That doesn't mean I don't get constant perf 'feedback' :P especially by folks who don't understand the process issues and don't worry about the (significant) cost to address those. The goal is 'instantaneous' so anything over zero is undesirable. Necessary of course, but there's a strong recurring desire to be slimmer. I'm surprised bitwise expressions don't involve indexes(1) and was hoping SQLite could work more optimization magic Adding indexes didn't help with some experimentation. I'd need to do a deeper redesign-think to see what options might exist, other than exploding INTEGER bitfield columns into multiple INTEGER boolean columns. But I'm really hoping I don't have to go there... (1) we have a fixed system w/o adhoc queries so for perf reasons we aim for 100% index coverage