A page with fifty five times fewer impressions produced more clicks

Ask which page on a site is the most important and most people reach for the one with the most impressions. It is the obvious answer, reporting tools sort by it, and on this property it is wrong.

Two pages

The calculator has 55 times fewer impressions and returns 27 percent more clicks. On a dashboard sorted by impressions it sits far down the page. It is the second largest source of search traffic on the site.

Two rankings, and they barely agree

This is not one odd pair. Rank every page by impressions, rank every page by clicks, and take the top ten of each. Only three pages appear in both lists.

Both distributions are concentrated, which is normal. They are concentrated onto different pages, which is not what the reporting implies.

Ten pages hold 80.7 percent of everything the site is shown for and produce a third of its traffic. Meanwhile half of all clicks come from 11 pages and 80 percent from 30, and those are mostly not the same 30.

What this does to a reporting habit

Impression-sorted lists are the default in every interface, so the pages that get attention, internal links, refreshes and rewrites are selected by the metric least connected to traffic. The work goes where the volume is, and on this site the volume is where the clicks cannot be.

The inverse error costs more. A page returning 9 percent on 8,521 impressions is not a small page. It is a page whose demand ceiling has not been found, and it is invisible on the sort order everyone uses.

The dead half, and why pruning it is not the answer

198 of 382 canonical URLs, 52 percent of them, earned zero clicks in the window. The standard prescription is to prune, consolidate or noindex them.

Those 198 carry 0.6 percent of impressions. Removing every one of them would remove 0.6 percent of reach and zero clicks. It is not the problem, it is not the fix, and it would take weeks of work that feels productive throughout.

An export of pages is not a list of pages

The first version of this finding reported 255 of 440 pages dead on 2.1 percent of impressions. Both figures were wrong, and correcting them took two passes.

First: 46 of those 440 rows are fragments, anchors within a page, which Search Console reports as their own rows. Every one showed zero clicks here. That is expected rather than informative, because a click resolves to the canonical page and not to the anchor - though 46 rows in one window is an observation, not a proof that such a row can never register a click.

Second: the remaining 394 rows are still not 394 pages. The same URL appears with and without a trailing slash, and with and without a www prefix. One of the largest pages on this site is split across three rows that way. Merging them leaves 382 canonical URLs, and that is still URL identity rather than verified page identity - reconciling it properly would need the canonical tags themselves.

The honest numbers are 198 of 382 on 0.6 percent of impressions. The conclusion never moved. What moved was our confidence that we were counting what we thought we were counting.

How to check your own

Export your pages. Remove every row containing a #, merge the www and trailing-slash variants, and only then sort twice, once by impressions and once by clicks, and compare the top ten of each list. The overlap is the number worth knowing. Ours was three.

Then look at whatever sits high on the click list and low on the impression list. Those pages are converting well on demand you have not gone after yet, and no impression-sorted report will ever put them in front of you.