A dimension we had never requested was worth 184,133 impressions a month

Analytics reporting has a shape. There is a default view, most people use it, and the tools built on top of the interface tend to request whatever the default view requests. Anything outside that shape is not hidden. It is simply never asked for.

What was sitting there

Search Console carries a dimension describing how a result appeared on the results page. It is available in the interface, documented, and free. On this property it had never once been requested.

Requesting it returned a surface that had never appeared in any analysis of this site: 184,133 impressions in twenty eight days at an average position of 8.37, producing 255 clicks. Google was automatically translating pages and serving them to readers who had searched in another language.

Two findings, and they point opposite ways

The first is that the site had an international audience nobody had decided to serve. Readers in other languages were being shown these pages at scale, on a property built entirely for one domestic market, with no translated content and no international targeting of any kind.

The second is that it converted at 0.14 percent, the same rate as the structurally zero-click clusters elsewhere on this site. A machine-translated title competing against results natively written in the reader language loses, and it loses at page one.

So this is not a discovery of demand waiting to be captured. It is a discovery of a real surface that performs badly, which is a different and less exciting thing, and it is what the number says.

Why there is no percentage here

The obvious next sentence is what share of the site that represents. We have published two different answers to that and both were wrong, so the third answer is that we are not going to publish one.

This number is a property-level slice by appearance type. It carries no URL breakdown at all, so there is no way to tell whether it includes the fragment-row impressions that our page-level denominator excludes. Three plausible denominators give 8.96, 8.85 and 8.73 percent, and choosing between them was itself the mistake: a share needs a numerator and a denominator counted over the same population, and we cannot demonstrate that for any pairing here.

The counts are exact and they are the finding. 184,133 impressions, 255 clicks, 0.14 percent, average position 8.37.

A trap this finding also set for us

Search appearance is a separate DIMENSION over the same impressions, not a separate slice of them. These 184,133 impressions are already inside the page-level clusters reported elsewhere on this site.

An earlier version of finding 01 added this row to two page clusters and reported the sum as a combined total. That was double counting, and it is an easy mistake because both figures come out of the same tool and look like the same kind of number. They are not. Adding a dimension to a page group is not addition.

Why it matters even though it converted badly

It had been invisible for the entire life of the property. Every summary, every report and every decision about where the audience was had been made without it. The problem was not that the number was bad. The problem was that nobody knew the number existed, and there is no way to know how many other dimensions are in that state without going and asking for them.

This is the general case. A reporting default is a decision someone else made about what matters, inherited silently, and then hardened by every tool that reads the same default.

How to check your own

Open Search Console and request the search appearance dimension explicitly. Then do the same for the dimensions you have never exported: country, device, search type, date at day granularity rather than aggregated. Each one takes a minute. On this property the first one produced a surface nobody had ever seen.