I launched a new project and, for its first week, ran Google Analytics and Yandex Metrica side by side. I wanted to compare them on the same real traffic instead of choosing one from screenshots, feature lists, or habit.
After that week, I found myself opening Yandex Metrica much more often. The reason was not a dashboard metric. It was Webvisor, Yandex Metrica's session-replay feature. At that early stage of the project, I cared less about another aggregate chart and more about a simple question: what are people actually doing on the site?
My first reaction was that Yandex Metrica was “winning by a mile.” I still prefer it for this kind of work, but that sentence was more emotional than precise. One project and one week are enough to describe my experience. They are not enough to prove that Yandex Metrica is objectively better than Google Analytics for everyone.
The part of Google Analytics that felt abstract to me
In the original version of this article, I described Google Analytics as abstract. That needs a qualification. GA4 is not shallow: it has reports, Explorations, segmentation, and User Explorer for drilling into individual user activity. The issue was the workflow I needed at the time.
With event and aggregate analytics, I can learn that users reached a page, triggered an event, dropped out of a path, or behaved differently as a segment. That is useful. But when I am trying to understand an unfamiliar UX problem, there is still a mental step between the metric and the interface the person actually used.
Session replay shortened that distance for me. Instead of only looking at counts, I could inspect the sequence of interaction in an individual visit: navigation, scrolling, clicks, pauses, and changes of direction. That did not automatically explain why the user behaved that way, but it gave me much better material for forming a hypothesis.
Webvisor did not eliminate guesswork
This is the most important correction to my original wording. A replay does not tell me that a user is frustrated. It shows behavior from which I may infer friction. Repeated clicks, backtracking, a long pause, or abandoning a flow can all look suspicious, but none of them is a mind reader.
That distinction matters because session replay is easy to over-trust. Watching one strange session can create a very convincing story in your head. The correct next step is not “I found the cause.” It is “I found a hypothesis worth checking.”
The workflow I now prefer is simple:
- Use aggregate analytics to locate an interesting page, flow, device class, or segment.
- Watch a small set of relevant replays instead of random sessions.
- Write down what is directly observable and separate it from my interpretation.
- Make a change only when the evidence is strong enough to justify it.
- Use analytics again after the change to see whether the measured behavior actually improved.
Session replay makes the middle of that loop much more concrete. It does not replace the rest of the loop.
Why it was especially useful on a new project
For a newly launched site, many UX questions are still unknown. I do not yet have months of stable baselines, mature funnels, or a long list of known failure modes. In that situation, qualitative observation can be unusually valuable because it helps reveal what questions I should even be asking.
This does not mean replay is more important than aggregate analytics. They answer different questions. Aggregate data is much better at telling me how common something is. A replay is much better at showing me what one concrete interaction looked like.
That combination is stronger than either view alone: first find the pattern, then inspect examples; or first notice a suspicious interaction, then check whether it appears often enough to matter.
Session replay has limits
There is another reason not to treat Webvisor as ground truth: the recording itself has technical limits. Yandex documents that Session Replay does not necessarily contain every visit, recordings are retained for a limited period, and some browser or page technologies can affect what is captured or how it is replayed. Dynamic styling, cross-domain iframes, private browsing, canvas content, and Shadow DOM are examples of areas where limitations can matter.
So if a replay looks broken, that does not automatically mean the user saw a broken page. And if a session is missing, that does not automatically mean nothing happened.
Privacy also deserves deliberate attention. Session-replay tools can observe much more than a normal pageview counter. Passwords and sensitive fields have protections, but the site owner still has to configure collection responsibly, avoid sending identifying data that should not be sent, and respect the consent and privacy requirements that apply to the site.
So which one would I choose?
For early UX debugging on this project, I preferred Yandex Metrica. Webvisor made it faster for me to move from “something looks odd in the numbers” to “I can inspect the interaction that may be behind it.” That was the decisive advantage in my first week.
I would not turn that into “Google Analytics is bad” or “Yandex Metrica is the better analytics platform.” GA4 has a different strength profile and a much deeper event-analysis workflow than my original short post gave it credit for. If acquisition reporting, custom event analysis, integrations, or a broader measurement stack are the main requirement, the comparison can easily look different.
I also do not think running both is pointless. During a comparison period, two tools can reveal differences in how they model sessions, users, and events. The numbers should not be expected to match perfectly, because analytics platforms do not necessarily measure and process everything in exactly the same way.
What I actually learned after one week
The useful conclusion is narrower than my original “Yandex wins by a mile,” but I think it is stronger.
Yandex Metrica won my attention because it made individual behavior easier for me to inspect. That changed how quickly I could move from a metric to a concrete UX hypothesis. Google Analytics remained useful, but for the problem I was focused on during launch, I reached for Metrica first.
Webvisor did not turn guesswork into certainty. It turned vague guesses into better-informed questions. For UX work, that is already a substantial improvement.