The 4x Rule: Why Your Best Ad Gets 80% of Spend Explained

Quick Answer

The 4x rule says that within any mature ad set, the best-performing ad will receive roughly 4x more spend than the worst-performing ad, and capture 70-90% of total ad set spend. This isn't a Meta policy — it's an emergent property of the intra-ad-set bandit, which converges on the highest-eAR ad once exploration concludes.

The Mechanism Explained

Inside an ad set, the bandit picks ads using Bid × eAR + Quality + ExplorationBonus. As impressions accumulate, the exploration bonus decays towards zero, leaving the bandit in pure exploitation. At that point, the ad with the highest baseline score wins every subsequent impression — not just slightly more, but essentially all of them.

The reason this looks like 4x rather than 100x in practice is twofold:

The bandit still occasionally explores — even mature ads get a small ongoing exploration share to detect performance drift
Different impressions favour different ads — an ad optimised for video viewers might lose to a still-image ad on a different placement

Combined, these effects mean the spend ratio between top and bottom is rarely 100:1 — it's typically 4-6:1, with the absolute spend share being 70-90% on the winner.

The deeper insight: the 4x rule means adding more variants doesn't get you more variant testing. The bandit will identify a winner and run it. The only way to genuinely test alternatives is to:

Use Dynamic Creative with rotation forced via Multiple Optimisation Goals
Run separate ad sets so each variant has its own bandit
Use A/B testing in Experiments tool with explicit traffic split

Each of these approaches trades efficiency for testing power. The bandit is efficient because it doesn't waste impressions on losers — but that's also why you can't extract reliable test data from it.

Practical Implication

Treat ad variants in a single ad set as search, not as experiment. The ad set's job is to find the winner, not to compare alternatives fairly. If you need fair comparison, run an experiment.

Real Numbers

Median top-ad spend share in mature ad sets: 78%
Median 2nd-place ad spend share: 14%
Median spend ratio top vs bottom in 5-ad ad sets: 4.3x

FAQs

Q: Can I force more even rotation?
Only via Dynamic Creative or Experiments.

Q: Does pausing the winner distribute to others fairly?
No — the next-best ad takes most of the freed-up spend.

Q: How many ads should I have if one always wins?
Enough that the bandit has options if the winner declines — typically 4-6.

Q: Does the 4x rule apply to Advantage+ Shopping?
Yes, but at campaign level rather than ad set level.

Q: Will adding more losing ads hurt my best ad?
Marginally — extra exploration spend gets pulled from the winner.

Pix-Vu

If one ad is going to capture 80% of spend, that ad needs to be your best possible creative. Pix-Vu helps you generate the variants where the winner is genuinely strong rather than the least-bad option — at https://pix-vu.com.