Re-seeding my lookalike from buyers, not leads, moved ROAS 40%

Re-seeding my lookalike from buyers, not leads, moved ROAS 40%

My lookalike audience was built on the wrong people and I didn't question it for a month.

— The setup: e-com, $48 product, Facebook lookalikes built off a 'leads' event (email signups), $150/day.
— The move: I rebuilt the 1% lookalike off the 'purchase' event instead — fewer seed users, but actual buyers.
— The numbers: the leads-seeded lookalike pulled cheap clicks but weak buyers: 1.6x ROAS. The purchase-seeded lookalike had higher CPMs but the people actually bought: 2.4x ROAS. Same offer, same creative, same budget. Over $2,000 each: leads-seed +18%, purchase-seed +58%. The purchase seed only had 1,100 buyers in it — I'd avoided it thinking the seed was 'too small.'

— The lesson: a lookalike is only as good as its seed's intent. 1,100 real buyers beat 40,000 email-only leads every time. Seed quality outranks seed size.

What I'd do differently: Seed from the deepest-intent event I have, always, even if the audience is small. A month on the leads-seed cost me roughly $1,200 in ROAS gap. Numbers illustrative; the buyer-seed edge is consistent.