Ads Manager gives you a real view of creative performance. Sort by cost per result and you can see which ads are working, kill the ones that are not, and brief more of what won. For an account running a handful of creatives, that is most of the job.
It stops being enough at exactly the point where creative starts to matter most: when you are trying to scale, and the way you scale is volume. Dozens of concepts a month, hundreds of ads in rotation, several sources producing them. At that point the per-ad view does not break so much as stop answering the questions you now have.
Page Contents
Where the per-ad view runs out
The questions that decide a creative programme at volume are not questions about any single ad:
- Which ideas are carrying the account, as opposed to which executions?
- Is the mix of what we are making shifting toward what works, or away from it?
- Where should the next batch come from, and what should it be about?
- Are we concentrated on one thing that is quietly fatiguing?
Every one of those is a property of the set of creatives. Ads Manager reports on ads. It has no idea what a creative is about, who made it, or what family it belongs to, so it cannot group by any of that. You can scroll a leaderboard of two hundred ads and still not know whether your account is built on three ideas or thirty.
At small volume you hold the answers in your head. At scale you cannot, and the leaderboard gives you no way to reconstruct them.
The view that answers them
Here is five months of one account’s spend, cut three ways, one cut per section below.
Same total spend in every view. Not one of these three views is available in Ads Manager, because Ads Manager does not know who made a creative, what idea it carries, or what format it is. Those are things you have to tell it.
The logic underneath is simple. On a modern platform, spend follows performance: the delivery system gives budget to the creatives it expects to convert. So where the money went is a readout of what the algorithm learned about your creative. Group that readout by the attributes that matter and you are reading the algorithm’s verdict at the level of strategy, not the level of a single ad.
The first cut: who made it
The left panel splits spend by source: agency, in-house team, an individual designer, an AI agent. Most teams skip it because it feels political. It answers the most practical question in a volume programme: where should the next creative come from? A source whose share keeps growing is carrying the account and should get more briefs. A source that is well-funded and shrinking is producing volume that does not earn spend.
Read it across months. One month tells you who won a batch. Five tells you whose work holds up.
The second cut: what it says
The middle panel splits spend by messaging theme, the idea underneath each creative. This is the cut that turns a pile of ads into a strategy. Individual ads fade; a theme that keeps earning spend across successive batches is an insight about the audience, and it outlives any single execution. When a theme’s share collapses, that is fatigue at the level of the idea, and no amount of re-editing the ad fixes it. The next batch needs a different theme.
Watch for concentration. One theme taking most of the spend means two things at once: you have found something, and you are exposed to it.
The third cut: what form it takes
The right panel splits spend by format: video, static, UGC. It is the cheapest cut to act on, because format is a production decision you control directly. If static is quietly gaining share while every brief still says video, the briefs are wrong. If UGC’s share is small and flat, the question is whether that is supply or performance, and crossing this view with the theme view usually tells you.
What has to be true before you can build this
None of those panels can be drawn unless every creative was tagged consistently when it was made. That is the unglamorous truth of creative analysis at volume: it is only as good as the tagging.
- Tag on the attributes you will cut by. Source, theme, format at minimum. Hook type, angle, audience, emotion and proof type if you want to know why a theme works, not just that it does.
- Tag everything, not only winners. The comparison is what makes the cut useful.
- Tag before launch. Retroactive tagging is where consistency dies.
At any real volume this is not something a person does well by hand. Tagging hundreds of ads on nine attributes is tedious, tedium produces inconsistency, and inconsistent tags make the whole view meaningless. This is why we use BruteForce AI: it tags every creative on the same attributes automatically, so the distribution views can be built across the entire creative set rather than the few someone got round to labelling.
Read it on the right metric
The three cuts can be drawn on any metric. Draw them on the deepest one your attribution supports: cost per trial, cost per paying user, ROAS. Spend distribution tells you where the algorithm put the money; reading it against a value metric tells you whether that was the right call. At scale, a theme that is winning spend and losing on value is the single most expensive thing in the account, and it is invisible on a per-ad leaderboard.
Structure the test, or the view reads noise
The distribution assumes each creative could earn spend on its own merits. Two things break that. Several creatives in one ad set are fine for a Bayesian test where the lopsided allocation is the signal, but wrong if you then compare them as though they had equal exposure; decide which you are running before launch. And below roughly two to three conversions a day per test, the distribution reflects variance rather than a verdict. How to structure the test itself is its own subject: how to structure creative tests for Meta ads.
What a report built this way tells you
- Which themes and formats are earning spend, and whether that is shifting. The pattern is the asset, not any single ad.
- What was tested and killed, and why. No kill list means half the work is unreported.
- What the next batch is built on, and what would change the plan. A report that produces no decision is a status update.
If a creative report only tells you which ads won, it is an Ads Manager export with a logo on it, and it will not carry you past the volume where the leaderboard stopped being enough.
This post is part of our guide to how to choose a mobile UA agency.
Frequently asked questions
How do you analyze ad creative performance? At small volume, the Ads Manager per-ad view is most of the job. Once you scale with creative volume, read spend distribution over time by the attributes that drive decisions: who made the creative, what messaging theme it carries, and what format it takes. Because spend follows performance, the distribution is a readout of what the algorithm learned at the level of ideas rather than ads.
Is Ads Manager enough to analyze creative performance? For a handful of creatives, yes. It stops being enough when you scale with volume, because the questions that matter then, which ideas carry the account and where the next batch should come from, are properties of the creative set, and Ads Manager reports only on individual ads.
Why look at spend distribution instead of a leaderboard? Spend follows performance on modern platforms, so where the budget went is the algorithm’s verdict. A leaderboard tells you how individual ads did; the distribution tells you which ideas and formats the account is being built on, and whether that is changing.
What do I need in place to analyze creative this way? Consistent tagging of every creative before launch, on the attributes you will cut by. At volume that means automated tagging, because manual tagging goes inconsistent and inconsistent tags make the analysis meaningless.
What metric should the analysis be read on? The deepest your attribution supports: cost per trial, cost per paying user, ROAS. A theme winning spend but losing on value is the most expensive thing in a scaled account, and the per-ad view cannot show it.
How often should creative performance be analyzed? Weekly for cut and brief decisions, and across several months for strategy. A week tells you who won a batch; a quarter tells you whose ideas hold up.
