The Meta ads daily check that catches leaks early

The signature leak: Advantage+ Placements and the Audience Network drain

Here is the leak that catches almost everyone at least once. When you build an ad set, Meta turns on Advantage+ Placements by default. It then spreads delivery across the surfaces it controls (Facebook, Instagram, Messenger, Threads, and the Audience Network) and shifts your budget toward whichever surface produces your chosen result most cheaply. That sounds smart, and for some accounts it is. The problem is that "cheapest" and "most valuable" are not the same thing.

The Audience Network is a network of third-party mobile apps and games, and one of its formats is rewarded video, where someone opts in to watch your ad to earn an in-game reward like coins or a power-up. Those viewers are not interested in you. They are interested in the reward. But because they sit through the video to get it, this traffic generates very cheap impressions, cheap ThruPlays, and a wall of green that looks like a winning campaign. It almost never produces landing page views, leads, or purchases.

This is not theoretical. One real account had roughly 97 percent of its spend draining into Audience Network rewarded video, with about 87 percent piled into a single country it never meant to target, all while the top-line numbers looked healthy. The money was gone before anyone thought to look one level deeper. That is exactly the kind of thing the daily check exists to catch.

The 7-minute daily check, step by step

Run these steps in order. The order matters, because each one assumes the one before it is clean.

Step 0, funnel integrity. Before you trust a single number, confirm that what is actually running matches what you think is running. Check that your live ad sets point at the audiences you intended and that your active ads carry the creative you intended. If you retarget video viewers, confirm the audience is built from videos you are still running. A funnel can be perfectly green and still be wired to the wrong audience or a paused creative. Verify the wiring first.

Step 1, account health. Scan for anything not delivering normally: ads in review, disapprovals, ad sets with delivery errors, payment or policy flags. Read Meta's own issue message rather than guessing. A disapproved ad quietly starves the variations next to it.

Step 2, yesterday versus targets. Pull yesterday's spend, CPM, ThruPlay cost, reach, and landing page view rate, and compare each against the target you set for this campaign, not against a number you found online. You are looking for drift, not perfection.

Step 2b, the placement and geo audit. This is the most important step, so do not skip it. Pull a performance breakdown by publisher platform and platform position, and a separate breakdown by country. Any spend on publisher_platform equal to audience_network should usually be zero unless you deliberately chose it. Watch the platform_position values for surfaces like rewarded_video, marketplace, search, and right-hand column. On the country breakdown, watch for one geo eating far more or far less of the budget than its fair share. This single view is what surfaces the leak from the first section.

Step 3, audience growth. If you run a retargeting or nurture stage, confirm the feeder audience is actually growing and is still sourced from creative you are running. A retargeting pool that stops growing is a leak upstream.

Step 4, downstream signals. Leave the ad platform and check the real world: site traffic from the campaign, signups or leads in your own database, and conversion events firing in your events tooling. Meta saying a conversion happened and your own system agreeing are two different facts. Reconcile them.

Step 5, the report. Write three or four lines: spend and key costs, placement health (clean or leaking and where), country share, audience trend, and one verdict (nominal, watch, or red flag) with the fix. Keeping this short is what makes you actually do it every day.

How to fix and lock down the placement leak

If the audit shows budget on the Audience Network or other junk surfaces and you did not want it there, switch that ad set from Advantage+ Placements to manual placements. Then deselect the surfaces that do not serve your goal. For most direct-response campaigns that means excluding the Audience Network, and commonly Messenger, in-stream video, search results, right-hand column, and Marketplace, so your budget stays on Facebook and Instagram feed, reels, and stories where genuine intent lives.

Now the part people miss. Placement settings are fragile. If you later edit that ad set's targeting through the API or a tool and leave the placement fields unset, Meta can silently fall back to Advantage+ Placements and the leak comes straight back. The rule is simple: after any edit to an ad set, re-open it and confirm the placements are still exactly what you set. Verifying placements post-edit is a permanent habit, not a one-time cleanup.

A quick note on retargeting audiences, since they tie into Step 0 and Step 3. When you build a video-view custom audience, Meta lets you choose from defined watch thresholds: a 3-second view, a ThruPlay, and 25, 50, 75, or 95 percent completion. The highest completion threshold Meta offers for a custom audience is 95 percent, so that is your safest high-intent option (100 percent exists as a reporting metric, not as a custom-audience threshold). Whatever you pick, make sure the audience is built from the videos you are currently running, or your retargeting stage is feeding on an empty or stale pool.

Read the funnel, not just the dashboard

Every metric makes more sense as a stage in a pipeline: impressions, then reach, then 3-second and 25, 50, 75, and 95 percent video views, then landing page views, then add-to-cart or lead, then purchase or subscribe. A leak at any stage starves every stage after it, so the skill is finding where the drop-off is unusually steep.

Two drop-offs tell clear stories. A weak hook shows up as people falling away between the 3-second view and the 25 percent view; they started the video and bailed, which is a creative problem, not a budget problem. A broken or mismatched landing page shows up as clicks and views that never become landing page views; the click happened but the page failed to load, was slow, or did not match the ad's promise.

Use benchmarks as loose context, never as universal law. They swing hard by niche, geo, objective, and season. As rough orientation for many feed and reels campaigns: CPM often lands somewhere in the low tens of dollars; a genuine cost per ThruPlay is usually a few cents, and a suspiciously low sub-cent figure is a classic rewarded-video tell; a landing page view rate above roughly 50 percent is often healthy while below roughly 25 percent points to a broken page or the wrong audience; and frequency climbing past about 4 per week on a feed or reels placement risks fatigue. Your own 7-day and since-launch numbers are far more trustworthy than any figure from a blog, including this one.

Catch creative fatigue before it costs you, and let an agent watch the rest

Creative does not fail all at once; it wears out. The early signals are mechanical: CPM trending up more than about 20 percent week over week, or frequency creeping past roughly 4 as the same people see the ad again and again. When you see that pattern, refresh the hook. Raising or lowering the budget does not fix a tired creative; it just pays more for the same fatigue. Rotate in a new opening and let the funnel reset.

The honest catch with a daily check is that it only works if you actually run it every single day, in order, without skipping Step 2b on the busy mornings, which are exactly the mornings a leak slips in. That is the part worth automating. An agent like AdsBud reads your Meta and Google Search accounts on a daily pass, scores every campaign against today's number, your 7-day average, and your since-launch baseline, and flags the leaks this guide describes (the Audience Network drain, a geo imbalance, a placement reset after an edit, a fatiguing creative) before they compound. It stays read-only by default, proposes the specific fix for your one-click approval, and you talk to it in plain language. The routine above is still yours to understand and own. An agent just makes sure no one ever forgets to run it.