Google rewrites Merchant Center performance reports on August 24, and restates July's numbers behind them

On August 11 Google published a changelog entry in the Merchant Center help center. It announces that performance reports change on August 24, lists four modifications, and then, in a single line near the end, clarifies something the coverage skipped: two of those changes are applied backwards, to data that already exists, going back to July 1, 2026. A number that left that platform in July and travelled into a client report, a dashboard or a bidding rule will stop matching what the platform shows after the 24th.
Four changes, and only two look backwards
Of the four announced changes, only the two that touch YouTube restate history; the other two add new information going forward and leave everything already reported alone.
The first splits YouTube affiliate traffic, meaning creators who earn a commission by featuring products, out of the Organic bucket where it had been mixed in, and gives it a value of its own. The second aligns the definitions of organic clicks and impressions for YouTube traffic with the standards YouTube already reports against on its own. Both are, at bottom, the same decision made twice: what used to be counted together is now counted separately, and under a different rule.
The third runs the other way, and it deserves to be said without hedging, because it is a real improvement. Product-level reporting on the Ads side expands to all channels and formats, including every network inside Performance Max and Video, App and Demand Gen campaigns. That cut did not cover this ground before, and without it the most basic question an e-commerce account has, which products are driving the result on each surface, was answered by approximation. The fourth adds a Network dimension for segmenting performance, similar to the one already in Google Ads.
The limit on the restatement is written into the document itself: "Historical data will only be retroactively updated for the separation of YouTube affiliate results and updated Youtube organic traffic definitions." The Ads reporting expansion rewrites nothing; it adds. What moves backwards is organic.
This already happened with GA4, and the lesson only half landed
A platform redefining what a traffic source means and then applying the new definition to history that already exists is not new this year.
In June, GA4 unified social sources into a Source Group and did it retroactively, with the same second-order effect: time series that change shape without anyone touching a campaign. The pattern repeats because the platform's logic is defensible. Counting two different things together is a measurement defect, and fixing it only going forward leaves an artificial step in the series on exactly the day of the change. Restating the past is, technically, the cleaner call.
The cost lands on the other side of the counter. Whoever reports outward does not have a time series, they have documents already sent. A July report carrying an Organic figure is a fixed object, dated, that somebody filed. When the platform restates, that document does not update: it gets contradicted. And because the restatement does not arrive with an in-product notice saying what changed and by how much, the conversation that follows is not about methodology, it is about why the number that was sent is not the number on screen.
What breaks is traceability, not the metric
The operational damage is not that the number changes, it is that it changes after being reported and without leaving a trace of the previous value.
The only available defence comes before the fact: export the July performance reports before the 24th, save the file with its date, and treat it as the version that was used to report. After the 24th that value cannot be reconstructed from the platform. With the export saved, the difference is explained in one line and everyone moves on; without it, there is no way to show that the number reported at the time was the right one.
The same applies to any automated rule that uses this history as its reference. An alert calibrated against July's organic volume will read a drop that never happened, triggered by a change in definition.
Who this actually changes something for, and who it does not
This matters to anyone reporting Merchant Center organic traffic outward, especially where creator content is pushing products, because that is where the YouTube affiliate split genuinely moves the number.
If an account has no YouTube affiliate traffic and does not depend on YouTube organic, the effect may be close to zero. That is worth saying plainly and without drama: Google published no figure for how far organic falls, gave no range, and did not say what share of accounts is affected. Headlines that assume a steep drop are adding something the source does not say. The sensible way to treat this is as a cheap risk to cover rather than an emergency: the export costs a few minutes, and the scenario where it is needed is the one scenario where it cannot be improvised.
The surface is worth bounding too. What is announced are Merchant Center performance reports. The product-level reporting expansion is described inside that announcement, and the document carries no equivalent statement about what changes in the Google Ads interface.
The two things the changelog does not answer
There are two declared gaps and neither can be closed with the available source.
The first is magnitude, already covered: there is no number. The second is the date for the Network dimension, which the document leaves at "later", with no month and no quarter. And there is a third that is not a gap in the announcement but in practice: it is unclear whether the restatement will be flagged in any way inside the interface, or whether the old values will simply be replaced in silence. Based on how the GA4 case went, the second is more likely, and that is precisely why exporting beforehand stops being an optional precaution.
It is the same underlying move that shows up when Shopping ad descriptions start writing themselves or when the companies that verify delivery quality change hands: the pieces that were treated as fixed stop being fixed, and the only layer left under your own control is the record. Periodic exports are worth treating with the same seriousness as a data source feeding a mix model, because a history the platform can rewrite is not a history, it is a query.
Sources
- Merchant Center announcement, August 11, 2026
- Merchant Center, detail of the performance reporting changes
Built with AI. Reviewed by humans. If any data point ages badly, blame the ecosystem, not us.