Thinking · Newsletter
How many creatives do you actually need for AppLovin?
AppLovin's own June 2026 data says top campaigns run 182 active creatives on average, and that campaigns using every available creative beat manually curated sets by 7 to 23 percent. Here is what producing that volume actually takes.
This piece is now part of a longer guide: AppLovin creative: the complete guide.
Most brands arriving on AppLovin bring ten creatives and a Meta mindset. The channel is built on a different assumption, and the gap between those two numbers is the single biggest predictor of whether the account works.
The number, and where it comes from
In June 2026 AppLovin released performance data on curated versus uncurated creative sets. Two findings matter.
Campaigns that used every available creative outperformed manually curated sets by 7 to 23 percent. Not the best creatives. All of them.
Those campaigns averaged 182 active creatives, ranging from 65 to 340. The curated control groups ran between 5 and 10 ads.
| Curated control | Full-library campaigns | |
|---|---|---|
| Active creatives | 5 to 10 | 182 average, 65 to 340 |
| Relative performance | baseline | +7% to +23% |
If you are running twelve ads on AppLovin, you are not running a small version of a working strategy. You are running a different strategy, and the data says it is the worse one.
Why the channel behaves this way
Meta's algorithm has been trained for a decade to find the winner inside a small set and push spend into it. Consolidation is rewarded. That is why Meta teams learn to kill underperformers quickly.
AppLovin's does something closer to the opposite. It is an auction over a very large, heterogeneous in-app inventory pool, reaching 1.4 billion daily in-app users. Different creatives win in different pockets of that inventory, and an ad that looks mediocre on an aggregate report is often the only thing working in a specific placement.
AppLovin's own framing, from that June release: "Many ads play a crucial supporting role in driving conversions, even if they aren't the last clicked AppLovin ad that gets attribution credit."
Killing that ad because its attributed ROAS is unremarkable removes a piece of machinery that was earning its place.
What the scaling curve looks like
The clearest public example: a clothing brand went from 20 creatives at $60,000 a week to more than 350 creatives at $223,000 a week over four weeks, with ROAS held constant.
Read that carefully. Spend went up 3.7x. ROAS did not degrade. The variable that moved was creative supply.
That is unusual and it is the whole argument for treating creative as a supply-chain problem on this channel. On most channels, scaling spend degrades efficiency because you exhaust the good audience. Here the constraint was how much creative existed to feed the auction.
Other published outcomes from the same period: MaryRuth's exceeded incremental ROAS targets by 49%, Portland Leather generated over 70,000 incremental purchases, PROOF now attributes more than 25% of total sales to the channel, and The Foldie recorded a 17% lift in new-customer ROAS.
The part nobody costs properly
Here is where most brands stall, and it is not budget. It is production.
182 creatives is not 182 shoots. It is somewhere between six and fifteen concepts, each built as a system that can produce twelve to thirty finished variants without breaking the idea.
That distinction is the difference between an achievable quarter and an impossible one. A team that thinks in finished films will quote you a year. A team that thinks in concept systems will define, before anything is shot, what is allowed to vary:
- The opening. Five openings against one body is five assets from one shoot.
- The performer. The same script delivered by three people is three assets, and it tells you something about your audience that the first two do not.
- The problem frame. The same product solving a different stated problem.
- Length. Note that top-performing videos in AppLovin's data averaged 37.4 seconds, which is longer than most Meta-native teams expect and changes what you shoot for.
- The end card. AppLovin added built-in end-card generation to Ads Manager in June 2026, so this axis got cheaper.
Six concepts, each yielding thirty variants, is 180 assets. That is the arithmetic, and it is a production planning problem rather than a creative one.
The mistake that wastes the volume
Producing 182 creatives that all point at the same product page.
Volume buys specificity. Each of those ads is free to speak to one motivation rather than the average of all of them. Then every one of them lands on a destination built for the average, and the specificity you paid for evaporates on arrival.
We read brand libraries for a living, and this is close to universal. Ads are specific. Pages are generic. The gap between what the ad promised and what the page opens with is where the money goes, and it is invisible in any report that stops at the click.
If you are about to scale creative volume, the cheapest thing you can do first is check where it all points.
What to do this quarter
- Count your active creatives on AppLovin. If it is under 30, supply is your constraint, not targeting.
- Stop curating down to winners. The data says the full set outperforms your shortlist.
- Convert your next brief from a film into a concept system, and write down what is allowed to vary.
- Before producing, list every destination your current ads point at. Count how many are product pages.
Questions people ask
Is 182 a target? No. It is the average of what worked in AppLovin's dataset. The useful reading is directional: the number is an order of magnitude above what most DTC brands run, and the range starts at 65.
Does this mean quality does not matter? It means singularity does not work. The brands in that data are not shipping 182 bad ads, they are shipping 182 legitimate variations of a smaller number of good ideas.
Can we get there by repurposing Meta assets? Partly, and it is the most common starting point. It is also why so many AppLovin accounts underperform: repurposed assets are not built for the placement, the length or the audience, and they hit a ceiling that looks like a channel problem.
What does it cost? That depends entirely on whether you are producing films or systems, which is why the same volume target quotes wildly differently from different suppliers. Ask any production partner how many finished assets one shoot day yields. The answer tells you which of the two they are.
Sources: AppLovin creative-set performance data, June 2026, as reported by Common Thread Co. Audience scale and skew via published AppLovin figures.