See which pages and visitors convert best

Many teams build this overview by hand in an analytics tool: conversion per device, per browser, per landing page, the best and worst pages with a minimum number of visitors. Varly does it from its own data, and each row is linked to what visitors actually did on that page.

Open Traffic & pages

1. Go to Research → Traffic & pages.
2. Choose a device (All devices, Desktop, Mobile or Tablet) and a period (7 days, 30 days or 90 days).
3. Optionally click Filters to narrow down by traffic source or New or returning visitors.
4. Check Counts as a conversion. Varly uses orders when your shop is connected. Otherwise it uses the goal you chose in Click to revenue, such as a visit to /thank-you. The page then talks about leads instead of orders.

What you see

• Oran’s read: the 3 to 5 strongest patterns in plain sentences, each with a link to the rows behind it. For example: "4 of the 5 landing pages that convert worst are blog posts. None of them links to a product."
• Summary: visits, conversion rate, orders or leads, revenue (or the estimated value, if you entered what one lead is worth) and the share of engaged visits.
• Landing pages: the first page of each visit. On top, Landing pages that convert best and Landing pages that convert worst, with a small picture of the page when Varly has one. For the worst pages Varly estimates what you miss, for example "If this page converted like the rest of your site: about 14 more orders a month". Use the chips (Product, Category, Blog, …) to look at one type of page, and switch between Similar pages together and Each URL.
• Pages that help: pages that visits saw before they converted, compared with visits that didn't see them. A page that was only seen after the order, such as an order confirmation, doesn't count.
• Devices, Browsers and systems, Screen sizes, Traffic sources, New and returning visitors and Countries.

The page is long, so the row of links under the filters jumps to each part. On a computer it stays at the top while you scroll.

Patterns and Oran’s read

Varly looks for patterns in these numbers with fixed rules, so they also appear without an AI key:

• What the worst or best landing pages have in common: the same type of page, mostly visits from ads or social media, or something on the page itself, such as no link to a product, no price or no reviews. A pattern needs at least 3 pages and 60% of the group, and must be at least twice as common there as in all your landing traffic.
• Worst vs best landing pages: loading time, engaged visits, scroll depth, rage clicks or the amount of text differ by 30% or more.
• A browser that falls behind on the same device, and whether its visits hit more JavaScript errors.
• After adding to cart: a browser or device that orders much less often once something is in the cart. Varly doesn’t see your checkout, so check it there.
• Small phones, a traffic source with at least 10% of your visits that converts clearly lower, and traffic that looks like bots: at least 2% of visits, very short, hardly engaged and never converting.
• A page that helps: visits that saw it convert at least 3 times as often, while fewer than 5% of visits see it.
• A busy landing page (top 10) that converts clearly below the rest.

What Varly reads from a page (links to products, price, add-to-cart button, reviews, amount of text) comes from the latest snapshot of that page. Without a snapshot it counts as unknown, never as "no".

With an AI key, click Explain with Oran at the top. Oran reads the patterns and the numbers behind them, and tells you what may explain each one, how to check it, and what to test. Oran only uses the numbers Varly found and never invents stock levels, prices or campaigns; when an explanation depends on those, he asks you to check. Each explanation costs a little on your own AI key and shows up in the cost counter under Organization settings.

Click Turn into test idea on a pattern to add it to Ideas, with its type (A/B test, quick fix, look deeper or check), the device and the numbers as evidence. The daily research round also turns the clearest cases into ideas by itself: a browser that falls behind, blog posts without product links, and a page that helps. When a landing page falls behind on one device and another source agrees on the same page (for example rage clicks on that device), it also shows up under Connections.

How to read the tables

Each row shows its visits (the small bar is its share of all visits), the share of engaged visits and the conversion rate.

• The dot is the conversion rate. The line around it shows where the real rate most likely is. A short line means plenty of data; a long line means few visits.
• The dashed line is the average to compare with.
• Green or red means a row converts clearly or likely better or worse than the rest. The words next to it always say the same thing, so you don't have to rely on color.

Browsers, screen sizes and traffic sources are compared within each device by default. Phones usually convert lower than computers, and some browsers are mostly used on phones. Comparing Safari with Chrome on mobile tells you more than comparing them across all devices. Switch to All devices together to see the overall picture.

How Varly decides

• Visit: one session, ending after 30 minutes without activity. Only visitors who gave consent are measured.
• Conversion rate: visits with at least one conversion, divided by all visits. It's per visit, not per visitor, so it can be lower than in other analytics tools.
• Engaged visit: lasted 10 seconds or more, saw 2 or more pages, or converted.
• Enough visits: a row needs enough visits to expect about 10 conversions at your site average, with a minimum of 100 and a maximum of 2,000. At a 1.5% conversion rate that's about 670 visits. Rows with fewer are grey and say Not enough visits yet. They never appear in the best or worst lists.
• Clear difference and Likely difference: each row is compared with all other visits, with the same Bayesian method as Varly's A/B tests. Because this page compares many rows at once, some will look different just by chance. So the bar is stricter than in Click to revenue: clear from 97.5% sure, likely from 90%, and only when the gap is at least 15%.
• Best and worst order: the best list is sorted with a small correction toward your site average, so a page with 120 visits and 3 lucky orders doesn't top the list. The table still shows the real numbers.

Common problems

• "Only 41% of orders could be linked to a visit." Conversion rates are then lower than they really are. This happens when many visitors don't give consent, or when orders reach Varly without the visit they came from. See Why are my orders not linked to visits?.
• Everything is grey. There aren't enough visits per row yet. Choose 90 days, or All devices.
• No conversions shown. Varly hasn't seen orders or a goal yet. Click Choose what counts as a conversion and pick your thank-you page, or connect your orders.
• Loading speed and errors aren't on this page. You find them per page on Analyses.