Stop guessing. See which ads to kill
and which to scale.
Every ad you run is a dot. Plot CPA against total spend and a little Poisson math draws two lines. Cross the red one and you're 90% sure it'll never hit target, so cut it. Cross the green one and it's a proven winner, so pour budget in. Everything in between just needs more spend before you can call it. No login, no email, and your numbers never leave your browser.
JSSlice of the Pie Marketing. We do this for a living.
Your action list
Sorted by priority. “Spend to decision” estimates how much more budget an ad needs, at its current pace, before the model can call it a winner or a loser.
| Verdict | Ad | Spend | CPA | Results | Δ vs target | Spend to decision |
|---|---|---|---|---|---|---|
| Load demo data or upload a CSV to begin. | ||||||
How the math actually works, and why it beats gut feel
Every result (purchase, lead, install) is a rare, countable event, so the number of conversions an ad produces over a given spend follows a Poisson distribution. If your target CPA is T and an ad has spent S, then a hypothetical ad that exactly hits target would be expected to produce μ = S / T conversions. The spread around that expectation is √μ, which shrinks (relative to μ) as spend grows. That's why a bad early CPA means little, but a bad CPA at high spend is damning.
The scale line (green)
CPA_scale(S) = T / (1 + z·√(T/S)). Beat this and the ad is performing better than a target-CPA ad would (1 − confidence) of the time by luck alone. It's a real winner. Scale it.
The kill line (red)
CPA_kill(S) = T / (1 − z·√(T/S)). Worse than this and a genuine target-CPA ad would almost never do this badly. You're confident it won't recover. Cut it. The line only exists once S > T·z²; below that there simply isn't enough spend to be sure, so those ads sit in the “Wait” band.
The z value
z is the one-sided normal quantile for your confidence: 80% is 0.84, 90% is 1.28, 95% is 1.64. Higher confidence means you demand more spend before acting, so you get fewer false kills but slower decisions.
This is a decision aid, not a guarantee. It assumes conversions are roughly Poisson and that past performance says something about the near future. That holds up for stable creative and audiences, less so right after a big change.
Want us to run this across your whole account, and act on it?
This tool tells you what to kill and scale. We're the ones who actually do it. Book a free 20-minute call and we'll pull your Meta account apart live, show you the spend that's leaking, and map where your next winner is hiding. No pitch deck, no obligation.
Grab a free audit callJSBuilt by Slice of the Pie Marketing · Everything runs locally. Your CSV and every calculation stay in your browser and are never uploaded.