How a comparison works
Open almost any standard GA4 report and there is an option to add a comparison. You define a group using conditions on dimensions — device category is mobile, country is Nepal, session source is google — and the report redraws with that group shown alongside the full set of users rather than replacing it.
Conditions can include or exclude, and several comparisons can run at once, each drawn as its own line or row. The unfiltered total stays on screen throughout, which is the feature that makes this useful: you are always reading a group against its own baseline rather than in isolation.
A comparison is built from dimension conditions applied to a standard report. That makes it quick and consistent, and it also makes it less flexible than a segment in Explorations, which can describe sequences of behaviour over time. If the group you want depends on what someone did in what order, you have reached the edge of what a comparison can express.
Why comparisons matter
A site-wide average is a blend of people who were never going to buy and people who nearly did. Splitting a report by device, country or channel is often the difference between a number that says nothing and a number that names the problem.
They are also the fastest honest answer to a client question. Asked whether the site works on mobile, or whether traffic from a particular country converts, you can answer inside the report you already have open, without building anything or exporting anything, and the person watching can see how the answer was produced.
Common mistakes with comparisons
The most common is comparing groups of wildly different sizes and reacting to the smaller one’s swings. A tiny group’s conversion rate moves violently on a handful of enquiries, and that movement is noise rather than news.
The second is stacking several comparisons until the chart is unreadable, at which point people start reading colours rather than numbers. The third is forgetting that comparisons vanish when you leave the report unless you save them, so an analysis you talked somebody through is gone by the next meeting.
How to act on it
Use comparisons for the two or three splits that genuinely change decisions in your business. For most of the sites I work on that is mobile against desktop, paid against organic, and one country against another — a Kathmandu business selling to the diaspora, for instance, has almost nothing to learn from a combined figure.
Keep to a small number of groups at a time and check the size of each before drawing a conclusion. Save the comparisons you return to every month so the definition stays identical, and when a split proves consistently important, promote it out of ad-hoc reporting and into a proper reporting dashboard where the whole team sees the same view.