Google to Shift Ad Impression Counting to Begin-to-Render in 2027

Full analysis

Google is changing how it counts an ad impression. Starting February 17, 2027, an impression only counts once the ad has actually loaded and started to paint on the screen. Today, a lot of impressions get counted at the moment the ad is requested or delivered, which is earlier. Move the counting line later, and every publisher's reported inventory number gets smaller on the same day, industry-wide.

How hard is this to undo? Easy to undo on paper, hard to undo in practice. Google can slip or reverse the date with a blog post. But once buyers rebaseline CPM benchmarks and contracts against the new counts, the market resets around the lower number and doesn't go back.

What's actually being decided: Not really a decision for you. It's Google deciding, on Google's timeline, with Google's definition, what a "real" impression is. Your decision is how fast you get your own counting and forecasting in line before the discrepancies show up in demand-partner reconciliation.

What sets the deadline: February 17, 2027, is the hard cutover. The soft deadline is earlier: you want a year of parallel measurement before that, so the clock effectively starts now.

The Market Analyst. Shrink nominal impression supply across the whole open web on one day and, all else equal, CPMs go up, because the same demand chases fewer counted impressions. Good optics for anyone selling on a CPM. Less good if you sell on total volume. The cleaner beneficiary is Google's own inventory, which sits inside Chrome's rendering plumbing and can measure a render event more precisely than a publisher on a third-party ad server can. For an outside reader: Google just moved the finish line and happens to own the most accurate stopwatch. DoubleVerify and IAS don't get killed here, but they now certify against a render signal Google defines and controls, which is a weaker position than setting the standard yourself.

The Skeptic. Slow down. This lands in 2027, and Google's history with announced deprecations is a graveyard of slipped dates. A three-year runway is how you buy goodwill headlines without creating urgency. For this to actually reset the market, buy-side and sell-side counting logic has to move in sync, vendors have to agree on what "begin to render" means, and app and connected-TV inventory (where the render signal is genuinely messy) needs a clean path. None of that is done. And the buyers who move real money already transact on viewable CPMs, so the compression mostly hits publishers who were padding supply anyway. In plain terms: the people this scares are the people who most need scaring.

The Operator. The date starts your clock, not Google's. First thing that breaks: forecasting. Every direct guarantee and programmatic deal quoted against today's impression counts throws discrepancies the moment the switch flips, and your demand partners will notice before your finance team does. Second thing that breaks: the board deck. Any slide showing "inventory growth" needs rebaselining, or you're explaining a drop that isn't a real drop. Run both methods side by side for a full 12 months, not 6, so you can show the delta is definitional and your actual ad delivery didn't change. Audit your counting against Google's spec now, while a year of parallel data is still possible to collect.

The CFO. The revenue doesn't move. The reported units do. That is the whole trap. Ad delivery, fill, and dollars can be flat while the impression line drops, and a CFO who doesn't understand the methodology change reads that as demand softening and starts cutting. So the cost here isn't measurement engineering, it's the internal story. Budget the parallel-measurement work as cheap insurance against a much more expensive misread upstairs. And watch the RPM math: if impressions fall and revenue holds, revenue-per-thousand-impressions jumps, which flatters you. Don't let anyone bank that as performance. It's arithmetic.

Where the council splits.

The Market Analyst and the Skeptic disagree on whether this resets the market at all. One says the compression mechanically lifts CPMs and hands Google a definitional advantage worth reacting to. The other says the real money already buys on viewability, so the reset mostly reprices supply that was inflated to begin with, and the 2027 date may not even hold.

The Operator and the Skeptic disagree on urgency. The Operator wants parallel measurement running now. The Skeptic says a three-year Google runway rewards the people who wait. Both can be right: the smart move is cheap enough to do even if the date slips.

The CFO reframes the whole thing. The others argue about impression counts. The CFO points out the counts don't matter to the P&L. The risk is a flat business that looks like a shrinking one to someone reading the wrong line.

What this actually hinges on. Two things. First, whether the render-event definition lands cleanly across app and connected-TV inventory, or stays a browser-first standard that leaves everything else fragmented. Second, whether your own finance and board narrative can absorb a lower impression number without treating it as lost demand. The first you can't control. The second you can, and it's most of the risk.

Which way it leans: get the parallel measurement running and rebaseline the internal story early. It's cheap, it de-risks the expensive misread, and it costs you almost nothing if Google slips the date. Don't fight the optics of a lower count. Fight the misreading of it.

What to verify before committing real effort: whether Google's render signal is available and consistent on your non-Google ad server, and whether your major demand partners have committed to counting on the same basis by the cutover. If they haven't, your reconciliation breaks regardless of what you do.

Prediction: Google will slip the February 17, 2027 Begin-to-Render cutover, either by pushing the date or by softening it to a non-enforced default, before that date arrives.

Confidence: Medium. Google's own track record on announced deprecations, plus unresolved app/CTV render signals, points that way.

Why: Google announced this in September 2026 with a runway stretching to February 2027, and a three-year-style lead time is how Google buys goodwill without committing to a hard operational date. The same company spent years announcing and then repeatedly slipping third-party cookie deprecation, which shows the pattern: a clean-sounding standard collides with messy real-world inventory (here, app and connected-TV render signals that are genuinely hard to instrument) and the date moves. The opposite outcome, a clean on-time cutover across the whole ecosystem with buy-side and sell-side counting in sync, would require a coordination Google has never once pulled off on a comparable measurement change. The safer-looking bet is that they ship on time; the pattern says otherwise.

Revisit by 2027-03-13: We're right if, by mid-March 2027, Google has publicly pushed the Begin-to-Render date, carved out app or CTV inventory, or shipped it as an unenforced reporting option rather than the counted standard. We're wrong if Begin-to-Render is live and counting all standard display impressions on the announced February 17, 2027 basis with no material carve-out.

Watch for this before then: whether Google publishes a render-event spec that DoubleVerify and IAS can certify against on non-Google ad servers. No shared spec by late 2026, no clean 2027 cutover.