Many poor leads usually means the promise was too broad
The vaguer the creative's promise, the more mixed the clicks. "Better skin" attracts everybody; "breakouts that have lasted three months or more" attracts one kind of person. Fewer leads, but each one closer to somebody you can actually serve.
Check message consistency before targeting
Do the creative, the landing and the first reply say the same thing? When they do not, customers arrive expecting A, meet B, and leave. That loss looks like bad targeting and is really an inconsistent message.
Filter with one qualifying question
Put one question in the first exchange that has to be answered, to separate a real need from an idle tap. Make it specific, easy to answer, and genuinely useful to the recommendation — otherwise it is just another barrier.
Measure by source, so you can switch off the group that is failing
Group leads by creative, placement and audience, and measure each group through to the order. Poor overall quality is usually one or two groups dragging the average; switching them off works faster than optimising everything.