How I cut a client's cost-per-lead by 64% — without touching the targeting.

Everyone reaches for the audience settings first. That's the wrong lever. The thing that actually moved cost-per-lead from $45 to $16 was boring, unglamorous, and sitting in plain sight: the creative.

When I took over the account, the campaigns weren't broken in the way people assume. The targeting was fine. The budget was fine. What was missing was volume and honesty in the creative — the same three videos had been running for months, and the algorithm had nothing new to learn from.

The real bottleneck

Meta's delivery is a learning system. Starve it of fresh creative and it optimises itself into a corner — showing the same tired ad to the same fatigued audience at a rising price. The fix isn't a new audience; it's a new thing to say.

The winning ad is found, not designed. Your job is to make finding it cheap.

So I rebuilt the creative pipeline around throughput. I directed the video myself, cut variations in-house, and used AI to spin more angles per week than a single shoot could ever produce. Then I let cost-per-result decide which ones lived.

What actually changed

• More creative shipped per week than the previous quarter combined
• Every ad judged on CPL, not on whether anyone “liked” it
• Conversions API wired to real bookings, so Meta optimised against money

Fourteen days later, cost-per-lead had more than halved. None of it came from a clever audience. It came from treating creative as the variable it actually is.