How much control to give up with Advantage+ app campaigns
Advantage+ app campaigns hand targeting and placement to Meta. A practical look at which levers you still hold, which you lose, and how to test the trade before moving budget.
Advantage+ app campaigns, Meta's name for what used to be Automated App Ads, are built on a simple bargain. You supply the creative, the countries and the goal, and Meta's system decides who sees what and where. For many app advertisers that is a fair deal. For others it removes the very controls they use to protect efficiency. The useful question is not whether automation works in general, but which of your current levers are doing real work.
List the controls you actually keep
Meta's developer documentation describes the structure plainly: one campaign, one ad set and multiple ads, in contrast to the many ad sets of a manual app campaign. The levers left to you are budget, country targeting with an optional language setting, the creative itself, the optimization event, the bidding approach and, for iOS, the SKAdNetwork attribution setup. Audience controls are limited to country and locale, and the documentation states that operating system targeting is not supported.
Notice what that does not include. There are no interest stacks, lookalike audiences or placement toggles to tune. If your current manual structure leans on any of those to hold your cost per install down, you are giving them up in exchange for whatever the automation finds on its own.
Check that your optimization event can feed the system
Automation can only learn from the events you give it. In an August 2022 update to the product, Meta added support for 7-day click attribution, better asset pairing and stability, split testing, and more detailed reporting by region and by ad. Those are useful, but none of them fixes a thin event signal.
If you optimize for installs, the system will find install-prone people, who are not always the people who subscribe or buy. If you optimize for a deeper event, you need enough of those events for the system to learn from. Meta's own help center says an ad set is labeled "learning limited" when it is unlikely to get roughly 50 optimization events in the week after its last significant edit. Our rule of thumb is to choose the deepest event you can reach that volume on, and step up the funnel one level if you cannot.
Put the creative to work because it is now your main lever
With audience and placement taken away, creative carries most of the weight. That changes how you brief it. Instead of one polished video, supply distinct concepts: different problems solved, different formats, different opening seconds. Meta combines assets and text, so variety within the pool matters more than small variations of one idea.
Keep a naming convention that tags each asset by concept. Because reporting is limited compared with a manual build, your own tagging is how you learn what the system is favoring.
Run it as a split test, not a migration
Do not pause your manual campaigns and move the budget across. Run Advantage+ app campaigns beside them, with a fixed share of spend and a fixed window. A reasonable starting point is 20 to 30 percent of your app install budget, which is large enough to exit the learning phase and small enough that a poor result is tolerable. Suppose your manual setup spends $50,000 a month; a $10,000 to $15,000 trial leaves most of your delivery on the structure you already understand.
Meta's split testing can divide the audience so the two structures do not compete against each other in the same auctions. Use it, because overlapping campaigns bidding for the same people will contaminate the comparison. Keep creative and the optimization event identical across both sides, otherwise you are testing the assets, not the structure.
Decision rules to set in advance
- Judge on a downstream metric. Compare cost per trial start or purchase, not only cost per install. Cheap installs from a broad system are common.
- Allow a full learning period. Avoid edits for the first week or so. Significant edits restart the clock.
- Check country mix. If automated delivery drifts toward your cheapest countries, split by country instead of running one pooled campaign.
- Keep manual where you have a thesis. If a specific audience or placement consistently wins for you, protect that structure and use automation to cover the rest.
Where to start
Start with the smallest scope that can give a clean answer: one country, one platform and one optimization event. If it matches or beats the manual setup on your downstream metric after a few weeks, expand country by country. If it falls short, you have learned that your manual levers were earning their keep, which is a perfectly good result.