Click-through rate tripled and traffic stayed flat. Why that is the right first move

On an affiliate website we work on, click-through rate from search went from 0.46% to 1.43% in three weeks. That is three times as high. Clicks did not triple. They stayed about level. This post is about why we count that as progress, and what it does not prove yet.

The rate sat at about 0.45% for three weeks before it moved, so this is a change and not noise. Read together, the three numbers say one thing: we stopped showing up where nobody clicks, and started earning the click where we do show up. About the same number of visitors now come from under a third of the impressions.

Why we chose this measure

Three findings on this site decided it.

Rank is not the prize. A broad search for one famous name sat at position 1.5 and earned a 0.86% click-through rate: almost first, almost no clicks. A specific decision search, two named options plus the word a buyer uses, earned 8.8% from position 12.5. That is ten times the clicks per impression, from page two.

Impressions can mean nothing. We studied the days when impressions spiked. The spikes were reference lookups: people checking a fact and leaving, at a 0.1% click-through rate. At one point those lookups were 65 to 75% of all the site’s impressions.

So the target is narrower than “rank first”. The aim we set with the client: page one for most of the terms that matter, and the top three only for the few where the searcher is ready to act, with those few named before starting. We wrote about the same idea earlier in 10% of searches, half the visits.

We stopped appearing for searches that never click

This is the part that surprises people. For the reference-lookup searches:

Week starting	Clicks	Impressions	Click-through rate
13 Sep	56	34,129	0.16%
20 Sep	33	7,066	0.47%
27 Sep	47	623	7.54%
4 Oct	52	492	10.57%

34,129 impressions a week became 492, and the clicks did not move: 56, then 52. Those impressions were never worth anything.

They also sat at an average position of about 8, on page one. That is why the site’s average position looked better before: it was flattered by rankings nobody used. When they went, the site’s average position went from 16.4 to 30.1 on paper, and nothing was lost.

How they went: the site had already been cut to the pages worth keeping, and in the last month about 240 more old pages were retired, each saved first and redirected to its nearest live section.

This is not a trick to improve a ratio. The pages that drew those impressions were stale, thin or duplicates, and a search engine spends its attention on a site in proportion to what it finds there.

Where we still appear, more people click

Take the searches that contain the word a ready-to-act visitor types. Call it the decision word.

Week starting	Clicks	Impressions	Click-through rate	Position
30 Aug	10	3,518	0.28%	66.9
13 Sep	25	4,303	0.58%	63.8
27 Sep	168	5,219	3.22%	54.6
4 Oct	190	5,605	3.39%	53.1

Clicks went from 10 a week to 190 a week, while impressions rose by 59% and position improved from 66.9 to 53.1. The visibility barely changed and the clicks did.

One limit on these two tables: the search engine withholds a share of queries for privacy, so figures by query never add up to the site total. They show direction, not the whole.

The counter-example

A second class of searches, the ones with the site’s main content word, shows why the rate alone is not enough. Its click-through rate rose from 2.11% to 3.28%. Its clicks fell, from 535 a week to 422, because impressions halved and position worsened.

The rate can rise while the thing you want goes down. That is why every reading in this post has clicks next to it.

What we do not know yet

• Whether the new titles caused the rise. We compared our search results with the five leaders on the same searches. Four of the five opened their description with the actual answer, where ours promised one. So we rewrote titles and descriptions to state the conclusion. Whether that is what moved the rate is a hypothesis. A test with three title shapes, assigned by a random draw, is running and has no result.
• Whether the higher rate turns into more traffic. Clicks are level.
• Whether the rate holds. Three weeks is a short series, and this market is seasonal.
• Whether more clicks become more revenue. We can see a visitor click an affiliate link. We cannot see whether they bought.

One more number belongs here. Against the same week a year earlier, clicks are still 73% down. The site lost most of its search traffic before we arrived. This is the early part of a recovery, not a recovered site.

Why this is the right first move

A site that is shown often and clicked rarely has two problems, and they have an order. If the pages do not earn the click, being shown more just multiplies a result nobody chooses. So the first job is to remove what never earns anything and make the remaining pages worth clicking. The second job is to be shown more. That is the work now.

If your own traffic line is flat, look at what is underneath it before you call it flat. Split your searches into the ones where people check a fact and the ones where people decide something, and read clicks, impressions and rate for each. It is the first thing we do in a deep audit, and the free audit is where that starts.

See how website revenue recovery works