Where Meta budgets actually leak

Meta will happily spread your budget across every surface it owns — Facebook feed, Instagram, Reels, Stories, Marketplace, Messenger, the Audience Network, search results. The pitch is that the algorithm finds buyers wherever they are. The reality, in account after account, is that a meaningful slice of spend settles into placements that have never produced a single sale.

Start with the 30-day placement breakdown

For any campaign with real spend, I pull performance by placement over the last 30 days. The pattern repeats: one or two placements carry every purchase, while a long tail of placements each takes a small bite of budget and returns nothing. Individually the bites look ignorable. Added up, I routinely find a quarter to a third of spend producing zero purchases.

The Audience Network is the classic offender — especially rewarded video, where people tap ads to earn game credits. The click-through rates look spectacular. They're junk: reward chasers, not buyers. If a placement shows sky-high CTR and zero purchases, you've found it.

Cut the zeros — the highest-confidence move in Meta

Excluding placements with zero sales is the rare optimization with no downside case: there are no sales to lose. The budget flows back to the placements that convert. I quantify it in dollars per month before touching anything, because “cut Audience Network” lands differently when it reads “recover $150/month currently producing nothing.”

Then be honest about the ceiling

Placement surgery buys you 10–15% efficiency. It does not rescue a campaign whose core placement is underwater.

If Facebook feed — usually the volume placement — is itself returning less than breakeven, no amount of trimming the edges makes the campaign profitable. At that point the problem is the creative or the offer, and the honest recommendation is a creative test, not another month of knob-turning. An agency that only ever proposes settings changes is billing you for motion, not progress.

One caveat that belongs in every Meta report

If part of your sales close by phone or off-platform, Meta's tracked numbers understate your true return — sometimes badly. That changes what absolute ROAS means, but it doesn't change the relative picture: a placement with zero tracked sales next to one with thirty is still a loser. State the caveat, then act on the comparisons that hold.

← All insights