One market served ads at a fraction of the rate every other market did

This finding has been rewritten twice, and both earlier versions claimed more than the data carried. What follows is what is left after removing them, which is a clear anomaly and no explanation.

What was measured

Ad impressions divided by page views, by country, over one window. Where readers load pages and ads fill, the ratio lands somewhere near the number of ad slots on the page. Every market on this property sits between 0.75 and 2.42. One does not.

Why this particular comparison holds

It is one instrument, one metric pair, one window and one ad configuration. The markup is identical on every page a reader in any country receives. The only thing varying across the rows is the market, which is what makes the gap worth reporting even with no explanation attached to it.

What it does not establish

Not the cause. At least three things produce this same signature and nothing measured here separates them.

• Unfilled inventory. An ad request can return no ad, and our own operating notes for this property record unfilled slots as common on it.
• Automated traffic. Requests that fetch the document and never run what comes after it. Suggestive here only because the three lowest ratios are cloud regions, which is a pattern and not a result.
• Ad blocking. Adoption differs sharply by market, and a blocked slot and an unfilled one look identical in this ratio.

A comparison we are not making

Search Console recorded 72 clicks from this market over the same window against 11,638 reported page views. An earlier version of this page led with that as a factor of 162 between two instruments. It is not a discrepancy and the framing has been withdrawn.

The two numbers count different populations. Ad network page views include every acquisition channel, direct, referral, social and assistant traffic among them, and count several views within a single visit. Search Console counts clicks arriving from Google Search, once each. A site can legitimately report both figures with nothing disagreeing anywhere, so the ratio between them means nothing on its own.

It is recorded here rather than deleted because it is an easy comparison to reach for, it looks compelling, and we published it.

What would actually settle it

Three things, in order, and none of them has been run.

1. 01Ad request coverage. Requests against matched requests for that market. This measures fill directly, and it is the one that either removes the unfilled explanation or confirms it.
2. 02Traffic validation at the log. User agents, request patterns and whether the page assets were ever fetched. Analytics cannot answer this, because a client that never runs the script never appears in it at all.
3. 03Engagement time, as support only. We previously called this decisive and it is not. It measures focus for sessions that were recorded at all, brief genuine visits also produce very little of it, and ordinary engagement would not establish that inventory went unfilled.

Why it is worth reporting unresolved

Because the number is load bearing. Page views for this market were the largest non-domestic block on the property and were being read as an audience worth expanding toward. Whatever the cause, a market serving ads at a tenth of the rate of every comparable market is not a market anyone should size a plan against until the reason is known.

How to check your own

Pull ad impressions and page views by country for one month and divide, then rank ascending. A market an order of magnitude below your others is worth investigating, not discarding, and the ratio tells you where to look rather than what is there.

Then resist the comparison in the section above. Before setting two numbers against each other, check that they count the same population, over the same window, at the same granularity. Ours failed on all three and still produced a figure that read well.