Loading…

The denominator problem

Two companies report a five percent conversion rate. The first counts unique visitors to the site against completed purchases. The second counts sessions that reached a product page against form submissions of any kind, including newsletter signups.

These are not the same measurement, and neither is wrong. But averaging them, which is what every published benchmark table does implicitly, produces a figure that describes nothing. Before comparing your conversion rate to anything, you have to know what is in the top and the bottom of both fractions, and for published benchmarks you almost never can.

All four are legitimate. Quoting one against a benchmark built from another is not.

The paid search version of this, where brand traffic does most of the distorting, is in the Google Ads benchmarks piece.

Why mobile drags every published average down

Benchmark tables routinely report mobile converting at a fraction of desktop, and then attribute it to screen size and friction. Some of that is real. A large part is composition.

Mobile traffic skews earlier in the buying process, more social-sourced, more accidental. Desktop traffic skews toward people who deliberately returned, often at work, often ready to act. You are not comparing two devices, you are comparing two populations with different intent, and the device is a proxy for the intent rather than the cause of the gap.

This matters practically. A team that reads the gap as a mobile usability problem will spend a quarter on mobile checkout and see very little, because the visitors were not close to buying. A team that reads it as an intent mix problem will look at where mobile traffic is coming from instead, which is usually the actual lever.

Segment before you compare anything

The single most useful change to conversion reporting is to stop reporting one number. Segment by traffic source, by new against returning, and by whether the visitor arrived on a commercial page or an informational one.

Once segmented, most conversion problems name themselves. A blended rate that fell because the traffic mix shifted toward informational pages is not a conversion problem at all, and treating it as one wastes a quarter.

Segmenting tells you where the loss is; finding the binding funnel stage is how to decide whether that is the step worth fixing.

What to do with the number once you trust it

Conversion rate is a ratio, which means it can be improved by making the numerator bigger or the denominator smaller. The second is underrated and frequently the faster win: cutting traffic that was never going to convert raises the rate and usually raises revenue per visit at the same time.

This is uncomfortable because it reduces a number most teams are also rewarded for. It is nonetheless often the correct move, particularly for accounts running broad match paid campaigns or ranking for informational terms with no commercial relationship to the product.

For genuine numerator work, the ordering that tends to hold: fix measurement first, then the offer and its clarity, then friction in the path, then the page itself, then copy. Most teams start at copy because it is the easiest thing to change.

That ratio behaviour is general, and the metrics worth reporting covers where else it shows up.

Deciding which pages should carry commercial intent in the first place is part of content architecture work.