How Meta ad spend distributes across individual creatives

The spend-per-ad distribution on Meta is heavily right-skewed. A small fraction of creatives capture the majority of spend — which is the mechanical reality behind why winners are rare and why portfolio allocation matters.

Spend per ad on Meta is not distributed evenly. It is heavily right-skewed: most ads receive very little spend, a middle band receives modest spend over 28+ days, and a small tail of ads captures budget that dwarfs the rest.

This is the underlying distribution that produces every pattern elsewhere in this report. The winner threshold, the portfolio breakdown, the concentration of spend on outliers — all of it follows from the shape of this curve.

What the distribution looks like

The full dataset — 578,750 creatives, 6,015 accounts, $1.29 billion in realized Meta spend — produces a spend-per-ad distribution with these qualitative features:

A large mass near zero. About half of all creatives receive little to no spend. Meta's auction pushes budget away from ads that don't earn it, and this shows up in the distribution as a heavy pile at the low end.
A middle band of mid-range spenders. Ads that reach 28+ days of spend but never become winners. At portfolio level, these make up 38–46% of creatives by tier.
A thin tail of high-spend winners. Roughly 5% of all creatives capture the majority of spend. These are the ads that clear the 10× threshold.

Why the distribution isn't published at bin level

This chart is described qualitatively rather than with a full percentile breakdown. The decision is intentional: publishing detailed bin or tail data on the spend-per-ad distribution could, combined with other dataset metadata, support reconstruction of individual high-spend creative behavior. For this reason, aggregate totals are published, but individual percentile points are not.

What the shape implies for strategy

Three practical implications follow from a right-skewed distribution:

Mean spend per ad is misleading. A small number of winners pull the average up. Planning against median spend per ad — not average — is more realistic. The median creative in this dataset spends considerably less than the mean.
Most ads won't spend. Budget follows signal. An account launching 20 ads in a month shouldn't expect spend to distribute evenly across 20 ads; it should expect 2–3 ads to dominate, another 6–8 to run moderately, and the remainder to be turned off within weeks.
Winner concentration is mechanical. The reason 55% of spend goes to winners (at portfolio level) and 64% at Enterprise isn't preference or discipline alone. It's the auction selecting for outliers and allocating toward them.

Spend concentration — the portfolio-level share of spend to winners, mid-range, and losers, and how the split shifts by advertiser size.
Winners are rare — why ~5% is the average hit rate, and what that tells you about the system.
The 10× benchmark — where the winner threshold sits on this distribution.

Frequently Asked Questions

Is spend per ad on Meta a normal distribution?

No. The distribution is strongly right-skewed — most ads receive little or no spend, and a small tail of ads captures disproportionate budget. This is the mechanical reality of Meta's auction: budget concentrates behind the creatives that earn it.

Why isn't the spend-per-ad distribution published at bin level?

Bin-level percentile data for spend per ad isn't published in this edition. Exposing fine-grained distribution tails could, in combination with other dataset metadata, support identification of individual high-spend creatives or accounts. The distribution is described qualitatively; aggregate totals ($1.29B across 578,750 creatives) are published.

What does right-skewness mean for how I should plan my creative portfolio?

It means most of your spend will concentrate on a small number of ads. Planning around average spend per ad is misleading — averages are pulled up by the tail winners. The useful framing is median-based: how does a typical creative perform, and how many winners are you likely to surface given your testing volume?