Advantage+ Placements: when it helps and when it leaks

What Advantage+ Placements actually does

When Advantage+ Placements is on, you hand Meta one decision: where to show your ad. Meta then optimizes for the cheapest path to your objective, which in practice usually means the cheapest impression that still moves your chosen metric. Its inventory is not all equal. The Facebook and Instagram feeds, Reels, and Stories are where high-intent people scroll. The Audience Network is a separate pool of third-party apps and mobile games, and it is almost always among the cheapest inventory in the auction.

The reason cheap is dangerous is a placement type on the Audience Network called rewarded video. These are the ads inside mobile games where a player is offered an in-game reward (extra lives, coins, a continue) in exchange for watching. Because the player has to sit through the ad to get the reward, the view is close to guaranteed and the cost per view looks fantastic. The intent behind it is usually close to zero. Almost nobody installs your SaaS because a mobile game made them watch your ad to unlock a level.

Meta defaults to Advantage+ Placements because, on average and at scale, letting the system find efficient inventory can beat hand-tuning for a lot of advertisers. The default is not a trap. It is a reasonable starting point that becomes a problem in specific situations, which is what the next two sections are about.

When it genuinely helps

Automatic placements tend to work best when you are optimizing for a real conversion event (purchase, subscribe, lead, install) rather than a cheap proxy like ThruPlay or impressions. When your objective is a true conversion, Meta has a quality signal to chase, so cheap-but-worthless inventory tends to get starved automatically because it does not produce the event. The optimization is pulling in the same direction you are.

It also tends to help when your targeting is genuinely broad and your budget is large enough for Meta to gather meaningful signal across many placements. Broad audiences plus a healthy conversion volume give the system room to find pockets of cheap, high-quality delivery you might not have picked manually. If you are spending enough that each placement accumulates real conversion data, letting Meta allocate is often better than guessing.

The short version: conversion-optimized, broad, and well-funded campaigns are where Advantage+ Placements tends to earn its default. If that describes your campaign, you can leave it on, but watch your placement breakdown anyway. Trust, then verify.

When it leaks (and the signature failure)

The leak tends to show up when your objective is a cheap proxy metric rather than a conversion, most commonly video views, ThruPlay, awareness, or reach. With no conversion event to chase, Meta optimizes for the cheapest impression that hits your proxy, and rewarded video on the Audience Network often wins that race. Budget can pour into incentivized views that produce no landing-page views and no conversions, while your dashboard shows a low CPM and a low cost per ThruPlay that look like success.

Small budgets can make it worse. With limited spend, the system has less conversion signal to learn from, so it can lean even harder on the cheapest inventory to hit your proxy metric. This is the under-served reality for many solo operators and lean teams, the exact segment where the default can quietly work against you.

This is not theoretical. On one real account, roughly 97% of spend had drifted to Audience Network rewarded video, with about 87% concentrated in a single wrong country, before anyone caught it. Every surface metric looked green. The funnel produced almost nothing. Cheap is not the same as good, and a green-looking dashboard is not evidence that your funnel works.

How to detect a leak in your own reports

You do not need a tool to catch this. In Ads Manager, open your campaign and pull two breakdowns. First, break the performance report down by placement, which in the underlying data maps to the publisher_platform and platform_position dimensions. Look at where spend actually went. Any spend on publisher_platform = audience_network should usually be near zero unless you deliberately chose it. Watch specifically for platform_position values like rewarded_video, instream_video, marketplace, search, and right_hand_column, which are rarely where buying intent lives.

Second, break the same report down by country. One country taking far more or far less than its fair share (say, one geo eating 80% of spend when you targeted four roughly equally) is a delivery imbalance worth understanding before you spend another dollar.

A few rough sanity checks, and these are ranges that vary a lot by niche, geo, objective, and season, so treat them as smell tests, not laws. For many feed and Reels campaigns, CPM often lands somewhere in the low tens of dollars, though it can be much higher or lower. A cost per ThruPlay of a few cents can be genuine, but a suspiciously sub-cent cost per ThruPlay is a classic rewarded-video tell. A landing-page-view rate above roughly 50% is often healthy, while below about 25% can suggest a broken landing page or the wrong audience. Frequency climbing past about 4 per week on a feed or Reels placement can risk ad fatigue. None of these are universal truths. They are the questions to ask, not the answers.

How to take manual control without breaking delivery

The fix is to switch the ad set from Advantage+ Placements to manual placements and keep your budget on the surfaces that carry intent: Facebook and Instagram feed, Reels, and Stories. Exclude the Audience Network, and for most direct-response campaigns also consider excluding Messenger, in-stream video, search, right-hand-column, and Marketplace. You are not crippling reach. You are refusing to pay for incentivized views and junk inventory.

There is one gotcha that bites people who edit ad sets later. If you patch an ad set's targeting through a tool or the API and omit the publisher_platforms field, Meta can quietly re-enable Advantage+ Placements, and the leak can come back. The rule is simple: always re-check your placements after any edit, not just the edit where you set them. Treat a placement audit as part of every change, not a one-time setup step.

Build a short daily habit, roughly seven minutes. Confirm your funnel wiring (the live audience and creative actually match what you think is running). Check account health for disapprovals or delivery issues. Compare yesterday's numbers to your targets. Run the placement and geo audit above, which is the single most important step. Check audience growth and downstream signals (site traffic, signups, conversion events firing). Then write yourself a two-line status. Keep in mind the funnel pipeline runs impressions to reach to 3-second and 25/50/75/100 percent video views to landing-page views to add-to-cart or lead to purchase or subscribe, and a leak at any stage starves the next. A weak hook tends to show up as people dropping between the 3-second and 25 percent video views. A broken landing tends to show up as clicks that never become landing-page views. And remember that budget tweaks do not fix a creative problem: when CPM trends up more than roughly 20% week over week or frequency pushes past about 4, the creative may be tiring and the fix is usually a fresh hook.

One practical note if you script video-view custom audiences off this funnel through the API: the event names Meta accepts are strict. The reliable building blocks are a 3-second watch event plus video_view_25_percent, video_view_50_percent, video_view_75_percent, and video_view_100_percent. Meta does surface a 95% watched metric in reporting, but a 95% rule is unreliable to build against programmatically, so anchor a deep-completion audience on the 100 percent event instead.

This is the kind of watch that benefits from being automated. An agent like AdsBud reads your account on a daily check, scoring every campaign against today's number, your 7-day average, and your since-launch baseline. It can flag exactly this Audience Network drain, explain it in plain language, and propose the placement fix for your one-click approval (read-only by default, with one-click rollback). You stay the decision-maker. The agent just helps make sure a leak does not run for three weeks before anyone notices.