Meta Says CPMs Rose 12% - A Major Tracker Says They Fell 13%. Here's Why You Shouldn't Trust Either Number Blind

Pull up three CPM benchmarks today and you might get three different answers about whether your ad costs are heading up or down.

That is not you doing something wrong.

It is what happens when three different sources measure three different slices of the same auction, then publish the result as if it were one number.

In its July 29 second quarter earnings release, Meta reported that average price per ad rose 12 percent year over year, the second straight quarter at that rate.

Nine days later, an independent benchmark tracker built on roughly $3 billion in ad spend put the median Facebook CPM at $16.47 for July, down almost 13 percent from a year earlier and the lowest point in its 13-month series.

The same tracker had shown CPMs up 16.5 percent year over year in June. Direction flipped inside a single month.

Both numbers are accurate. They are measuring different things.

If you benchmark your own account against either one blind, you are comparing yourself to a moving target that even the people tracking it cannot agree on.

What Meta’s earnings actually said

Meta’s average price per ad increased 12% year over year in the second quarter of 2026.

Ad impressions delivered rose 14% over the same window, and revenue climbed 28 percent to $60.8 billion. That price figure is a blended, company-wide average across every surface, objective, and country Meta sells ads in.

It moves with the mix of what advertisers bought as much as with what any single ad actually cost. A 12% lift in that number says nothing about what happened to your CPM in your account, your vertical, or your geography.

What the independent tracker found

Superads’ benchmark tracker measures a large advertiser sample directly, rather than deriving price from Meta’s revenue math, and its July print told a different story.

Median Facebook CPM landed at $16.47, the deepest trough in the tracker’s 13-month window and a 12.7 percent decline from July 2025.

One month earlier, the same series showed CPMs up 16.5 percent year over year. That is a full sign flip in four weeks, on a benchmark built to be stable.

The divergence gets sharper below the headline number.

Over the same 13 months, ecommerce CPMs fell roughly 13 percent while media-vertical CPMs climbed nearly 38 percent. A single “Facebook CPM” headline is really an average of markets moving in opposite directions, flattened into one figure.

Why both numbers can be true at once

Meta’s figure is a global average pulled up by ad performance and weighted by where impressions actually landed.

The company’s own filing points to impression growth on lower-monetizing surfaces as one force holding the blended price down even as price per ad climbs elsewhere.

An independent tracker samples a specific set of advertiser accounts, objectives, and geographies, and none of that sample is guaranteed to match Meta’s global mix, let alone your account.

This is the same disconnect that shows up whenever Meta, Google, and TikTok ROAS never quite agree: each platform and each vendor is answering a slightly different question, and none of them is answering the one you actually need answered.

What this looked like inside real accounts

The gap between “the market” and “your account” showed up hard in late July.

Some advertisers reported a single account’s CPM falling from roughly EUR 60 to EUR 7 in a matter of days, while click-through rate collapsed from 3.5% to 0.85% over the same stretch.

That is cheap impressions buying worthless traffic, not a discount worth celebrating.

It happened the same week Meta rolled out a new ranking system, one of several July changes that never show up in Ads Manager but still move your numbers.

Cost per acquisition reportedly doubled to nearly five times its prior level for some accounts inside a single week. Neither headline benchmark captures any of that, and it is the only number that actually determines your budget.

What to actually do about it

Stop benchmarking your account against a single external CPM figure, no matter how reputable the source. Use your own trailing 90-day average as the baseline instead.

One industry rule of thumb holds that a good CPM is one that sits within 10 to 15% of your own rolling 30-day average, not one that matches a published number pulled from a different set of accounts entirely.

The same logic applies to marketing efficiency ratio and every other planning metric: the outside number sets context, your own trend line sets the decision.

If you do reference an outside number, name the source and the methodology every time you cite it, so future you can tell what actually changed versus which tracker you happened to read that week.

And treat any “industry CPM is up or down X percent” headline as a question worth digging into, not a conclusion worth acting on.

It is the same lesson AdBeacon’s clients learn with platform-reported ROAS versus actual ROAS: a number pulled from someone else’s aggregate is a starting point for a question, never a verdict on your own account.

Your own click-based, first-party data is the one baseline that cannot disagree with itself from one press release to the next. If you want to see what independent measurement actually looks like against your own account, not someone else’s blended average, book a live AdBeacon demo.

—-

FAQ

Why do Meta’s earnings and third-party CPM trackers disagree?

Meta reports a blended global average across every surface and country it sells ads on, weighted by impression volume. Independent trackers sample a narrower set of advertiser accounts and objectives. Different inputs produce different numbers, even when both are measured correctly.

Are Facebook CPMs going up or down in 2026?

It depends which measure you use. Meta’s Q2 2026 earnings showed average price per ad up 12 percent year over year. A major independent tracker showed median Facebook CPM down 12.7 percent over roughly the same window, after showing CPMs up 16.5 percent just one month earlier.

What is a good CPM benchmark to use?

Your own trailing 90-day average, not a published industry figure. A CPM within 10 to 15 percent of your own rolling baseline is generally healthy. A number pulled from someone else’s account mix tells you little about your own.

Why did some advertisers see CPMs crash in late July 2026?

Community reports tied a sharp CPM drop, paired with a collapse in click-through rate, to a new Meta ranking system that rolled out on July 29. Cheaper impressions did not translate to better traffic for every account.

How often should I update my CPM baseline?

Recalculate your trailing average on a rolling basis, at least monthly, so seasonal and platform-driven shifts get folded into your baseline instead of being compared against a stale number.