Excluding purchasers from your Meta campaigns is table stakes. You already do it. But across every account I've audited, that single layer leaves three gaps that only become expensive at real spend levels — and one of them nobody talks about.
The underrated one: lookalike seed bleed. When you build a LAL from your purchaser seed, Meta doesn't automatically exclude that seed audience from the lookalike ad set. The result is that existing customers show up inside your prospecting delivery. I've measured this at 8-15% of LAL impressions depending on seed size and LAL percentage. At ₩20M+ monthly spend, that's not a rounding error — it's a dedicated line item of wasted budget serving ads to people who already bought from you.
The second gap is less obvious: high-LTV retained buyers in the 31-180 day window. Most accounts only exclude 0-30 day purchasers. When I first added a 2+ order / 180-day retained customer exclusion to prospecting campaigns, I expected ROAS to hold flat. Instead it went up — because Meta had been burning budget trying to re-acquire people who were going to buy again on their own anyway. Removing them tightened the prospecting pool and pushed the algorithm toward genuinely cold users. The algorithm was essentially being trained on self-converting buyers and calling it a win.
The third layer is engaged non-converters who've aged past their conversion window. Someone who hit your PDP three times, added to cart, and still hasn't converted after 21+ days is a different kind of user than a fresh visitor. Keeping them in prospecting inflates CPM and sends low-converting signals upstream. They either need a separate retargeting treatment or a hard exclusion from upper-funnel campaigns — not another prospecting impression.
Stack all four layers — recent purchasers (pixel + uploaded list both), retained high-LTV buyers, aged non-converters, and LAL seed exclusion — and the addressable prospecting pool shrinks 15-25%. That's intentional. Smaller, cleaner pool beats a bloated one with noise baked in.
I wrote up the full breakdown — including the AEM deduplication lag problem that makes single-layer exclusion porous at high daily spend, and the exact implementation steps for each layer — over on themedilog.
Top comments (0)