User behavior analysis begins where traffic stats stop: a page can pull in steady search traffic and still perform poorly — visitors arrive and leave almost immediately without reading on or clicking further. Site-wide averages often hide this because one or two problem pages get smoothed over.
This skill compares bounce rate, time on page and browsing depth across pages to find exactly those cases.
What user behavior analysis does
For a site connected to Yandex Metrika (optionally a specific set of pages), it reads bounce rate, time on page and browsing depth for pages with meaningful traffic, then flags anomalies — bounce rates well above the site average, or time on page implausibly short for the content: a ten-minute guide, say, where the average visit barely outlasts the page load.
What you get
A two-part chat report: "losing people" — a table of page, traffic, the problem metric versus average, a hypothesis (snippet mismatch, slow load, unstructured text, no next step), and what to check or fix; and "benchmarks" — two or three best-performing pages worth copying from. It closes with a recommendation on which page to start with.