measurement.the CPM went up. your campaign didn't.

Adjust measured global CPM up 32% since 2024, with the US reaching US$20.54, an all-time high. In the UK and Ireland, cost per app install rose 80%. These are market figures: they describe the price of inventory, not the quality of the media you bought with it.

The same move shows up inside Meta. In its Q2 2026 results the company reported average price per ad up 12% year over year, with impressions up 14%. Independent trackers measuring real accounts recorded average CPM up 20%, from USD 11.82 to USD 14.19.

And the average hides the part that matters. That global 12% ranges from +1% in Asia-Pacific to +21% in Rest of World, with US and Canada at +20%. Almost all the media we run sits in Rest of World, the region that got most expensive and the one furthest from the average everyone quotes. Buy in the expensive region and the global average understates your problem. Buy in the cheap one and it overstates it.

A year-over-year comparison assumes both years were measured the same way. In digital media that assumption almost never holds. In January 2026 Meta removed the 7-day view and 28-day view windows from Ads Manager; the current default is 7-day click plus 1-day view. Comparing 2025 against 2026 without flagging that is comparing two different definitions of a conversion and calling the gap a trend.

There is a quieter trap. Since 6 August 2026, three Insights API breakdowns, device, time of day and frequency, only return data for accounts that explicitly requested the permission. On an account without it, the query doesn't fail: it returns cleanly, with an empty table. Empty is not zero, and a conclusion written on top of an empty table is a conclusion you made up.

Both changes land inside the same chart as the performance they are supposed to describe, and neither announces itself there. Anyone putting two periods side by side owes the reader one line on whether the ruler stayed the same, before a single percentage is discussed.

Picture a campaign paying 20% more per thousand impressions this year while serving the same message to the same audience at the same frequency. Cost per outcome rises. The report turns red. Nothing the agency controls has changed. And yet the exact same number is also consistent with a campaign that genuinely decayed. Two hypotheses are fighting over one figure, and the only way to settle it is to look at something other than that figure.

Real decay leaves fingerprints inventory inflation never leaves: 7-day frequency above 3 in prospecting, CTR down more than 20% over fourteen days against week one, CPA climbing more than 20% above its own baseline, spend share falling because the system stopped feeding that creative. If those signals are there, it's fatigue. If they're absent and CPM rose, it's the market.

Next comes the unit of analysis. Campaign-level CPM is an average that buries wide variation between creatives: two ads in the same set can be paying very different prices. Reading at the ad level keeps you from concluding that the campaign got more expensive when what actually happened is that the system shifted volume into pricier inventory.

Then the source. The average price per ad Meta publishes in its quarterly results is realised revenue per ad delivered, averaged globally. It is not what a specific account pays in a specific market. External benchmarks are context; the CPM measured in your own account is the one that decides.

Framing inventory inflation as campaign decay damages a relationship through excess humility. You apologise for something that never happened, you change what was working, and you leave behind the belief that the work got worse. The cost isn't one month, it's the lens applied to the next twelve.

The reverse is worse. Using a rising CPM as the explanation for everything turns a market fact into an alibi. The first time a client cross-checks it against another source, and they will because the figure is public, no future read gets credit, including the accurate ones.

The way out is identical in both cases: state the full comparison. How much inventory cost rose in that market, how much cost per outcome rose, and the gap between the two. That gap is the only part of the number anyone can claim as the work, for better or worse.

The note explains the problem; the services explain the work.