Break it down by what you can change: visitor type, device, source, landing page, country. A blended rate points to no one in particular and triggers no action. It's the class average: it reassures the principal and helps no student.
The benchmark
The most revealing breakdown is often the first one: new vs. returning visitors.
In the 2026 Contentsquare benchmark, returning visitors convert at 2.9% and new visitors at 1.7%. On the stores the authors audit, the gap is often wider: at one brand whose sales peak in winter, returning shoppers converted four to five times better than first-time visitors, and 85% of those first-time visitors came in on mobile.
The only segment that really mattered was therefore "new visit, on a phone," and that's exactly the one the ads were bringing in.
The rule
A segment earns a line in your report if someone at your company can act on it. Otherwise it's decoration.
The steps
- New vs. returning visitors. Shows whether the problem is convincing people or bringing them back.
- Mobile vs. desktop. Shows where to look at usability.
- Source. Shows whether paid traffic finds what it was promised.
- Landing page type. Product page, collection, home page, campaign page.
- Country, if you sell abroad.
- The volume of each segment. Before you read it, count its orders, and write the result as one sentence for leadership:
"This segment gets 80 orders a month, so we'll review it quarterly."
The pitfall
Slicing too thin.
A segment with 100 orders over the period has a rate that swings by about 20% in either direction by pure chance, because the noise in a count grows with the square root of the count: ten orders of difference out of a hundred is well within normal variation. You'll end up commenting on noise. Below a few hundred orders, extend the period or group segments together.
Second pitfall, less obvious: your segments don't cover all your orders.
At one fashion brand, Shopify Analytics tied a device to only one order in two, because the rest were placed without analytics consent.
Before comparing mobile and desktop, look at the sum of the two: if it's far from the total, you're slicing only the half of your customers who consented.
