Conversion rate

When results decline, there are only two places the cause can be: before the click, in the auction and the targeting, or after it, on the page. Conversion rate separates them in a single comparison.

If clicks and spend held steady while conversion rate fell, nothing about the auction changed. The same number of people arrived and fewer of them did anything, so the cause is on the site: a page change, a broken form, a slower load, a price rise, an out-of-stock product. If conversion rate held and clicks fell, the page is fine and the problem is in delivery.

This matters commercially as much as technically, because the two sit on different sides of the agency boundary. One is yours to fix and one belongs to the client, and arriving at that conversation with a number is a completely different experience from arriving with a hypothesis.

It is also the fastest available answer when a client says the campaign stopped working. Most of the time it did not.

Google Ads divides conversions by clicks. GA4 divides by sessions or users depending on which report you are looking at. Those are three different denominators, so three different percentages come out of the same underlying reality and none of them is wrong.

Clicks and sessions are not the same event either. A click that never finished loading is a click and not a session. One visitor clicking twice is two clicks and possibly one session. Consent rejection removes the session and leaves the click, which is why the gap between the two is widest in Europe.

The numerators differ as well. Each tool applies its own attribution and its own window, and an ad platform reports a conversion on the date of the click while analytics reports it on the date it happened.

The practical resolution is not reconciliation, which is unachievable, but consistency. Pick one source for reporting, say which it is on the report, and use the others as context.

Conversion rate is more sensitive to who is arriving than to anything on the page. Branded search converts several times better than generic search, remarketing converts better than prospecting, and returning visitors convert better than new ones. None of that is about the page.

So a campaign deliberately expanding into a broader audience will show conversion rate falling, by design, while total conversions rise. Reported without that context it looks like a decline, and the client asks what went wrong with a strategy that is working.

The same arithmetic runs in reverse and is more dangerous. Cutting back to the efficient core raises conversion rate and reduces volume, which reports beautifully. An agency judged on conversion rate has an incentive to shrink the account.

The defence is segmentation. Report conversion rate by campaign type rather than as one account figure, so an expansion shows up as a new segment with its own expectations rather than as a fall in the headline.

A page with 40 visitors and 2 conversions reports 5%. The honest confidence interval around that figure is wide enough to include 1% and 15%, which means the number is compatible with the page being excellent or terrible.

This is the most common analytical error in agency reporting, and it is invisible because the percentage looks equally precise either way. A figure of 5.0% carries no indication that it came from two events.

The practical rule is to look at the conversion count before the conversion rate. Under roughly thirty conversions, treat movement as noise unless it is very large. Under ten, do not report a rate at all.

The same applies to testing. Deciding a variant won after fifteen conversions each is not a result, and acting on it produces a confident wrong answer, which is worse than no answer.

Not a site-wide average. Site-wide figures include direct traffic, returning customers and people arriving from an email who already intended to buy, all of which convert far better than a cold click from an ad. Holding a paid landing page to that standard guarantees it looks bad.

Not an industry benchmark either, for the same reason as every other benchmark in this glossary: it is measuring a different traffic mix on a different offer at a different price point.

The comparisons that carry information are the page against its own history, two variants against each other in a properly powered test, and the same campaign before and after a specific change. All three are available without buying a report.