Finding Pages That Are Seen but Not Chosen

What Is the Problem?

A seen-but-unchosen page damages on three faces, and none shows in a standard report.

The report mistakes it for success

A high-impression page fattens the “visibility is growing” line in the monthly report; the fattened line gets applauded, and applauded problems go unfixed. When the visibility row swells while the sales row stands still, management moves from puzzlement to distrust of the whole channel — yet the culprit is not the metric but the way it is read. The thirteen-fold scissor in our portfolio is manufactured daily in exactly this spot.

The opportunity raises no alarm, so it waits

A page losing traffic triggers an alert and jumps the queue; a page displayed and ignored never alarms — it never rose, so it cannot fall. Yet the engine has already judged that page trustworthy, showing it to users again and again; the only missing piece is a reason to click, which makes this the cheapest repair in all of search. Earning a rank takes months of trust-building; converting an earned rank into clicks is often one sentence’s work.

It dissolves into the average

A site averaging two percent click-through hides eight-percent champions and 0.2-percent ghosts in the same figure. The average has one job — period-to-period comparison; for page decisions you always descend to the distribution. Averages soothe management; inventories direct work.

Why Does It Happen?

These pages arise along three typical roads.

The page ranks for the wrong question

The content explains a topic broadly, and the engine files it under a generic query — a building-materials site surfacing for the bare word “timber”. Query and page intent never intersect; the user reads the mismatch in a glance and moves on. The engine’s placement is no error: a page that does not position itself sharply gets the widest possible reading. The diagnostic ruler is the intent-match test.

The shopfront is dimmer than the neighbours’

Query right, rank fine — but the title is dry and the description machine-picked. On the same screen a rival has stated a number, named a difference, answered the question; the finger goes there. Content wins the rank, the shopfront wins the click — two separate races, and the second hands out medals without consulting the first. Here the page stays; the label gets rewritten.

The answer lives on the screen itself

For definitions, dates and conversions, the results page answers directly; boxes and summaries finish the errand. In this group a low click rate is not a defect but the query’s nature — these pages’ assignment is citation and recall, not clicks, and their report card is written in that currency. Telling this group apart is what keeps repair effort from being wasted.

How Is It Done?

A four-step sieve in Search Console’s Performance report; total time, half an hour.

Steps 1–2: fix the period, set the floor

Select the last three months and open the click-rate and average-position columns. Then, on the Pages tab, set an impression floor — a meaningful base by the site’s volume, from a few hundred to a few thousand — and drop everything under it: the rate of a rarely-shown page is noise, not statistics. Google’s help documentation is the base reference for the report’s fields.

Step 3: cross the two filters

In what remains, two conditions are searched together: position 10 or better, and a click rate clearly below the site’s own average. The intersection is the on-the-shelf-but-not-selling inventory. Then each page is opened on its Queries tab: which words display it tells you the repair’s address.

Step 4: split into three labels, remove the untouchables

The inventory divides three ways: intent mismatch (wrong query), weak shopfront (wrong label), answer-on-screen (clickless query). The third group leaves the list and gets reported separately — effort spent there never returns. Brand-query pages are also untouchable: their modest rate is usually sitelinks sharing the clicks, a feature, not a disease.

How Long, at What Cost?

This tip’s budget fits a lunch break; its return spreads across months.

Half an hour of sieving, one hour of labelling

The filters take minutes; the real time goes into reading query lists and assigning the three labels. On the panel where we watch eleven properties, this job runs monthly and each round yields three to ten work orders — a round that yields none is itself a signal: shopfront work on that property is done.

Cost zero, precondition one

The tool is free; the one precondition is a verified property with accumulated data. Without the data, the measurement period is waited out first — trading weeks for years is this trade’s standard bargain.

The return compounds

Every repair from this inventory activates a page already being displayed; unlike new content, it builds no trust from zero. Pound for pound it is the fastest-returning job in search — and the practical antidote to unclicked first place.

The Common Mistake

The inventory gets built — then misused in three ways.

Watching the rate, forgetting the volume

Sorted by click rate alone, ten-impression pages float to the top. Priority comes from rate multiplied by impressions: the page bleeding the most clicks is repaired first.

Writing one prescription for all three labels

The three labels take three medicines: intent mismatch wants content and targeting change, a weak shopfront wants label writing, an answer-on-screen query wants nothing at all. Unlabelled repair rubs the same ointment on all three and fails on two.

Building the inventory once and shelving it

The query mix is a living thing; this month’s clean page can enter the list next quarter. An inventory without a calendar stays a one-off curiosity; bound to a routine, it becomes the shopfront’s guard.

Frequently Asked Questions