The Ad Placement That Was Quietly Burning a Third of Our Budget

In three days, one Meta ad placement — Facebook's Reels overlay — took a third of a live campaign's budget and produced real engagement almost never: 1 genuinely engaged visit out of 119 landing page views. The campaign's headline numbers — impressions, click-through rate, cost-per-click — looked completely normal the entire time. The only way we caught it was breaking performance down by placement instead of trusting the top-line dashboard. That's the actual point of this post.

What the dashboard showed vs. what was actually happening

We recently increased the daily budget on a live Florida/Texas campaign to move it out of Meta's learning phase faster. A few days later, Google Analytics showed something that didn't match: 100% bounce rate and 0 seconds average engagement time on both landing pages, with 28 active users and 109 events logged that day. Meanwhile, the campaign's own numbers inside Meta Ads Manager looked healthy — click-through rate between 1.5% and 2%, cost-per-click between $0.25 and $0.40. Nothing at the account level suggested a problem. If we'd only checked the summary view, we'd have moved on.

Breaking it down by placement found the real problem

Meta Ads Manager lets you break any campaign down by exact placement — Facebook Feed, Instagram Stories, Reels overlay, and so on — instead of just the platform-level total. Here's what three days of real data looked like once we did that:

Source: Meta Ads Manager, Andflow campaign data, July 18–20, 2026. "Real engaged visit" is Meta's own engaged_page_view conversion event, which requires meaningful time on the page — not just a page load.

Reels overlay alone accounted for roughly a third of total campaign spend across both markets, more than any other single placement, while returning almost no real engagement. Feed did not perform meaningfully better on this measure, on a small sample — the one placement that clearly stood out was Instagram Stories, at a fraction of the cost per real visit.

Why the algorithm drifted there

This wasn't random. Two things pushed spend toward the weakest placement. First, a significant budget increase resets Meta's learning phase, and during that reset the algorithm explores more broadly to find delivery volume fast. Second, this ad set is optimized for "Landing Page Views" — an event that only requires the page to start loading, with no requirement that the person meant to click. A Reels overlay ad interrupts someone mid-swipe; a reflexive tap to dismiss it can register as a landing page view without the person ever choosing to engage. That's a cheap way to hit the optimization goal on paper, and exactly the kind of goal a learning-phase reset will chase.

Two systems, same answer

What made this worth acting on immediately, rather than shrugging off as one noisy day, is that two independent systems agreed: Meta's own pixel showed almost no engaged visits from Reels overlay, and Google Analytics independently showed 100% bounce and zero engagement time on the same pages, over the same period. Neither number alone would have been conclusive. Together, they weren't a coincidence.

What we changed

We restricted both ad sets (Florida and Texas) to Facebook Feed and Instagram Feed/Stories only, manually excluding Reels overlay, Facebook Stories, in-stream video, and Marketplace — placements where someone is either mid-swipe or mid-video, with no real moment to evaluate an ad before moving on. This will reset the learning phase again for a day or two, which is expected, not a new problem. We're checking the same placement breakdown again in a few days to confirm the fix holds.

The habit this is really about

None of this shows up if you only look at account-level numbers. Click-through rate and cost-per-click tell you whether people clicked and how cheaply — they don't tell you whether the placement behind that click gave anyone a real chance to see the offer. A campaign can look completely healthy at the top level while a third of its budget quietly goes to inventory that was never going to work. The only way to catch that is checking placement-level breakdowns on a regular basis, and always right after a budget or targeting change, since that's exactly when Meta's delivery re-explores and can drift toward whatever's cheapest to fill volume with.

This kind of check is part of what actively managing a live ad account looks like, week to week — not something that gets set up once and left alone.

Curious whether one of your own campaigns has a placement quietly eating the budget? Book a free ads audit — we'll look at the real breakdown, not just the dashboard.