Winning ads have a shelf life measured in weeks.

The account came to us running well into six figures a month in a crowded, gift-driven category with an average order value just under $47. On paper, the job was to scale spend. In reality, the enemy wasn’t the budget — it was creative fatigue.

And fatigue is sneaky. It doesn’t show up in ROAS until after you’ve already paid for it. By the time the return dips in the dashboard, the money is spent and the winner is dead. Scaling this account meant solving fatigue before it reached the reports — not reacting to it after.

We treated creative like the targeting — because it is.

Meta’s algorithm finds the buyer; the creative decides whether that buyer converts. So we built a system where creative was the primary lever, structure was the delivery mechanism, and fatigue was monitored as a leading indicator rather than a lagging one.

Phase 1 — Test creative like it’s the targeting

• Six testing angles: problem-to-solution, demonstration, social-proof stack, founder/origin, occasion & gifting, and comparison — so we always had distinct messages in market, not variations of the same one.
• UGC format matrix: we ran creator formats against each other to find which packaging carried each message furthest.
• Fixed copy framework with hook rotation: holding the body steady and rotating hooks let us isolate exactly what moved performance.

Phase 2 — Kill fatigue before it costs you

• Early-warning monitoring: we tracked cost per checkout initiated and three-day CTR decay to catch fatigue days before it hit ROAS.
• An escalation ladder for dead ads: swap the opening frame → try a new creator → try a new format → retire. In that order, every time.
• The insight that paid for itself: a dead ad usually has a live message inside it. Roughly 40% of underperforming ads recovered with a new opening frame alone — no new concept required.

$213,748.01 spent. $1,039,367.59 generated.

In March 2026 alone, across 22,222 purchases, the account delivered a 4.86x blended ROAS at a $46.77 average order value and a $9.62 cost per purchase — while spend scaled well past where fatigue would normally have broken it.