A site-wide figure is an average across visitors who behave nothing alike. Splitting it four ways, by device, by source, by landing page, and by new against returning, routinely turns an unremarkable overall number into an obvious and specific problem.

The problem with a single number

A site converting at three percent sounds like a site converting at three percent everywhere. It almost never is.

It is far more likely to be six percent on desktop and under one on mobile, or eight percent for people arriving on a service page and near zero for people arriving on a blog post. The overall figure is the weighted average of those, and it describes no actual group of visitors.

Averages are useful for tracking a direction over time. They are close to useless for finding out what is wrong, because the problem is always concentrated somewhere and the average is designed to smooth exactly that out.

The four splits worth running

By device

Start here, because it finds the most problems. A large gap between desktop and mobile conversion is common and usually mechanical: a form that is awkward to complete on a phone, a number that is not tappable, or a page that is slow on a mobile connection. If mobile is most of your traffic and half the conversion rate, that single gap is your biggest available improvement.

By landing page

People arriving on different pages have different intent, so comparing them is comparing readiness rather than page quality. What you are looking for is a page receiving substantial traffic and producing nothing, which is either an intent mismatch or a page with no way to act on it.

By source

Search traffic and social traffic behave completely differently, and blending them hides both. A channel sending volume with no enquiries is either the wrong audience or the wrong landing page, and separating it is how you find out which.

By new against returning

Returning visitors convert substantially better in most service businesses, because they have already decided you are a candidate. A high overall rate carried entirely by returning visitors means new visitors are not being persuaded, which is a different problem from the one the average suggests.

What the combinations reveal

PatternUsual cause
Mobile far below desktopForm usability, an untappable number, or mobile speed
One landing page with traffic and no enquiriesIntent mismatch, or no way to act on the page
One source with volume and nothing elseWrong audience, or arriving on the wrong page
New visitors far below returningThe site persuades people who already know you and not strangers
Everything uniformly lowA site-wide problem such as a broken form, worth checking first

The trap in segmenting

Splitting far enough always produces an interesting-looking number, and at small volumes most of those numbers are noise.

A segment with eleven sessions and one enquiry shows a nine percent conversion rate, which means nothing. Before drawing a conclusion, check the size of the group. As a working rule, ignore any segment with fewer than a hundred sessions or fewer than about thirty enquiries, and treat anything below that as a hypothesis rather than a finding.

This matters more for small business sites than for large ones, and it is the specific version of the small-numbers problem described in what your analytics cannot tell you.

A ten-minute routine

Four splits, three months of data, and it will normally identify one specific thing worth fixing. That is a considerably better use of an analytics session than watching a site-wide figure move by fractions of a percent and wondering what it means.


Frequently asked questions

Why is my overall conversion rate misleading?

It is a weighted average across visitors who behave nothing alike. A site at three percent is commonly six percent on desktop and under one on mobile, and the average describes no actual group.

What should I segment my analytics by?

Device, landing page, source, and new against returning. Those four splits find the majority of problems on a small business site.

Why is my mobile conversion rate so much lower?

Usually mechanical rather than mysterious: a form that is awkward on a phone, a number that is not tappable, or a page that is slow on a mobile connection.

How small is too small to draw conclusions from?

Treat anything under about a hundred sessions or thirty enquiries as a hypothesis rather than a finding. Splitting far enough always produces an interesting number, and most of them are noise.

Should returning visitors convert better?

In most service businesses, yes, since they have already decided you are a candidate. A good overall rate carried entirely by returning visitors means new visitors are not being persuaded.

What does it mean if every segment is low?

A site-wide problem rather than a segment-specific one, and the first thing to rule out is a broken form or an enquiry route that is not delivering.

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.

Overall numbers looking fine and enquiries not coming?

We split your data four ways to find where the loss is actually concentrated, rather than watching an average that hides it.

Start a Conversation