GA4 or Yandex Metrica - what to choose for Russian business?

Both. GA4 - for international traffic, Google Search Console and export to BigQuery. Metrica - for analyzing the Russian audience, Webvisor and integration with Yandex Direct. On projects with a budget of 300K ₽/month, I install both counters and check the data once a week. A discrepancy of more than 15% is a signal of a problem with tags.

How to set up GTM yourself in 1 day?

Create an account on tagmanager.google.com, copy the container code to the site (in and ). Add a GA4 Configuration tag with Measurement ID. Add a Metrics tag with the counter number. Set up the All Pages trigger. Publish. All. Basic setup takes 2-3 hours, if the site is on Tilda or WordPress - 40 minutes.

What is Last Click attribution and why does it lie?

Last Click gives 100% of the conversion value to the last channel before the purchase. If a person saw a VKontakte banner, then clicked Direct retargeting, then went directly - Last Click will record the entire conversion for the direct visit. Direct and VK will seem “non-working”. In practice, this leads to budget cuts for channels that actually work at the top level of the funnel.

What is server-side tracking and when should it be implemented?

Server-side tracking is when events are sent not from the user’s browser directly to GA4/Metrica, but through the server. Browser blockers and iOS 14+ cut up to 30-40% of client events. Server-side gets around this. It’s worth implementing with a budget of 500K ₽/month or when you see that the data in the accounts is 30%+ less than in the CRM.

What metrics should be displayed on the main dashboard screen?

Of the 8 projects that I am currently running, they all have the same thing on the first screen: sessions, conversions, CPA by channel, ROAS (if e-com), and % change from last week. No bounce rate or viewing depth - these are metrics for a sitemaster, not for a marketer.

How to set up goals in Yandex Metrica?

Go to Metrica > Settings > Goals > Add a goal. For e-com - a JavaScript event for the purchase event from the dataLayer. For lead forms - a JavaScript event on the submit button. To view pages, use the 'page visiting' condition with the URL /thank-you. After setting, check in debug mode: Metrica > gear icon > Tag Debugger.

Hotjar or Microsoft Clarity - which is better for heat maps?

Clarity - free, no limits on session recordings, direct integration with GA4. Hotjar is paid from $32/month, but the UX of the interface is better and there are surveys. For most projects, Clarity covers 90% of the needs for free. I use Hotjar only if I need in-app surveys or NPS.

How to check that analytics is working correctly?

Three checks: 1) DebugView in GA4 - go to the site in incognito mode and watch events in real time. 2) Metrics Debugger - Chrome extension shows all counter events. 3) Compare GA4 and CRM data for the last month - a discrepancy of more than 20% in leads indicates a problem with tags or attribution.

All articles

PerformanceOctober 7, 202514 min

Marketing Analytics 2026: GA4, Metrics, Attribution and Dashboards

How to build marketing analytics from scratch: GA4 vs Yandex Metrica, GTM setup, channel attribution, heat maps and web viewer. Real stack for a team of 1 person.

Without analytics, performance is a budget in the dark. Below I discuss not “how to install GA4”, but specifically: the minimum stack that is actually used, why Last Click lies in almost half of the cases, and when server-side tracking pays off in a month.

I have been working with marketing analytics since 2018. On the current 8 projects - GA4, Metrica, GTM are installed everywhere without exception. The numbers in the article are from real offices and correspondence with about 15 other fellow marketers whom I helped set up a stack over the past year.

Nice numbers in your advertising account and real sales are different reports. Without end-to-end analytics, you optimize the first, thinking that you are optimizing the second.

1. Minimum analytics stack 2026

At the start of a project with a budget of up to 500K ₽/month, four tools are enough. Exactly four - no more, otherwise you are drowning in settings instead of looking at the data.

GA4 — the main tool for event analytics, funnels and export to BigQuery. Yandex Metrica — cross-validation of data and Webvisor. GTM — a tag manager through which you manage both counters without a programmer. Heatmap — Hotjar or free Microsoft Clarity for analyzing boarding behavior.

This stack covers 80% of the analytical tasks of an average project. Looker Studio for a dashboard - I add it only when a report is needed for a client or team; it does not produce data by itself.

2. GA4 vs Yandex Metrica: when to use what

The question “what to choose” is wrong. Correct: why each of them is needed separately.

Function	GA4	Yandex Metrica
Event model	Yes, flexibly via dataLayer	Yes, but weaker in terms of customization
Recording sessions	No (needs Hotjar/Clarity)	Webvisor - built-in, free
Heat maps	No	Yes, built in
Funnels	Yes, powerful exploration funnels	Yes, easier but faster to set up
Integration with Yandex Direct	Limited (via conversion import)	Native, direct
Google Ads Integration	Native	No
Export to BigQuery	Yes (free up to 1M events/day)	No
Audiences for remarketing	Google Ads / DV360	Yandex Direct / Ya.Audience
Real time	Delay 24-48 hours for some reports	Almost real time
Cost	Free (360 from $50K/year)	Free

My practice: I read Metrica when I need to quickly see “what’s happening right now” - Webvisor, heat maps, funnels via Direct. I open GA4 for in-depth analysis - cohorts, cross-device paths, attribution, export to BigQuery.

On one of the e-com projects (electronics, turnover ~12M ₽/month), the discrepancy between GA4 and Metrica in the number of transactions was 18% - with an identical code. Reason: GA4 sampled part of the sessions, Metrica did not. That's why I install both and check them once a week. A discrepancy of more than 20% is a signal that there is a problem with the tags somewhere.

3. GTM: tags, triggers, variables - in 30 minutes

GTM scares marketers with its interface - but for basic setup there are three concepts, no more.

Tag — what we send (GA4 Configuration, GA4 Event, Metrica). Trigger — when we submit (pageview, button click, form submission). Variable — what we substitute in the tag (page URL, button text, value from dataLayer).

Basic setup in 30 minutes: create a container, copy the snippet to the site, add a GA4 Configuration tag with Measurement ID, a Metrics tag with a counter number, assign an All Pages trigger to both - publish. This will be enough for the first two weeks.

The next step is setting up the dataLayer. The developer adds calls like this to the site dataLayer.push({'event': 'purchase', 'value': 4900}}),GTM intercepts and sends to GA4. Without dataLayer, you are blind to events - you only see pageview and indirect indicators.

On an EdTech project, implementing a dataLayer for 7 key events (registration, course start, lesson completion, purchase) took the developer 4 hours. After that, we saw a real funnel for the first time - and discovered that 40% of users quit the course in the 3rd lesson. Without this data, advertising would be optimized instead of the product.

4. Heatmap and Webvisor: finding holes in conversion

Heat maps and Webvisor - the only way to understand what the user is doing on the page without inventing it from the bounce rate numbers.

Three case studies where a heat map explained what numbers could not:

On the landing page of a SaaS product, users actively clicked on a decorative block with icons, mistaking it for a button. The CR of the form was 0.8% - after adding a CTA to this block it increased to 2.1%.
On an e-com product card, 70% of clicks on a photo went to the lower third of the image - a description hidden from mobile users accidentally appeared there. Redesigned - conversion on mobile +34%.
Webvisor showed that YAN users leave the landing page after 8 seconds, without even scrolling to the offer. The lead block was moved higher - CPL decreased from 1,800 rubles to 1,200 rubles.

Microsoft Clarity - free, no session limits, directly integrates with GA4. It's enough for most projects. I use Hotjar only when I need built-in surveys on the page.

5. Attribution: why Last Click lies

Attribution is the most misunderstood topic in analytics. And the most expensive in case of error.

Last Click says “the last channel gets all the value of the conversion.” The user saw an ad on VKontakte → then clicked on Direct retargeting → a week later he typed the brand into the search directly → bought it. Last Click will record 100% of sales as direct entry. VK and Direct will seem “non-working”.

Attribution model	Logic	When to use	Main disadvantage
Last Click	100% → last channel	Never as main	Kills the top level of the funnel
First Click	100% → first channel	Entry point analysis	Kills the bottom of the funnel
Linear	Equal shares to all channels	Start, little data yet	Does not distinguish between channel contributions
Time Decay	More close to conversion	Short transaction cycle (1-3 days)	Underestimates reach channels
Data-driven (GA4)	ML based on real paths	With 300+ conversions/month	Need volume of data
Position-based	40% for the first, 40% for the last, 20% for the rest	Long transaction cycle (B2B)	Arbitrary weights

According to my observations on 6 projects with deployed multi-touch attribution: the transition from Last Click to Data-driven opened up 25-40% of “invisible” conversions - those that Last Click attributed to branded search or direct visits, although in reality the role was played by media or VKontakte at the entrance. This is not an increase in conversions - this is a correct accounting of those that already existed.

6. Server-side tracking: when and why to switch

Server-side tracking — when events are sent not from the user’s browser, but from your server to GA4/Metrica. The browser transmits the event to your endpoint, the server forwards it further.

Why is this needed: ad blockers (uBlock, AdBlock, built into Safari ITP) cut up to 30-40% of client tags. iOS 14+ has limited data transfer from the browser. On one fintech project, we saw: in CRM 890 leads per month, in GA4 - 560. The 37% gap is a blind spot when making budget decisions.

After switching to server-side, GA4 began to record 820-840 leads - the increase in visible conversions was 25-40%, depending on the month and type of traffic. The budget didn’t change, sales didn’t change—it’s just that the analytics finally showed the real picture.

When to switch: with a budget of 500K ₽/month, or when the gap between CRM and analytics is more than 20%, or with a high share of iOS traffic (mobile applications, lifestyle niches).

Technically, this is setting up GA4 Server-Side via Google Cloud Run or sGTM (server-side GTM). Cost: $50-150/month for hosting, 1-2 developer days for integration.

7. Dashboard: which metrics to display on the first screen

On all 8 projects I have the same first screen of the dashboard. Not because “it’s customary,” but because I tried different options - only this one stuck.

Sessions/unique users - % change from last week
Conversions (goal) - absolute number and CR%
CPA by channel - breakdown into at least 3 sources
ROAS or Revenue - only if e-com with transaction tracking
Funnel by key stages - from the first visit to conversion

Bounce rate, viewing depth, time on the site - I remove it from the first screen. These are metrics for a UX audit once a quarter, not for weekly monitoring by a marketer. When “everything is on the first screen,” attention is scattered and important things are drowned.

Dashboard in Looker Studio assembled in 3-4 hours using ready-made sources (GA4 + Metrica). I set up automatic PDF distribution on Monday at 9:00 - the team sees the week even before the planning meeting.

8. What to do right now

If there are no analytics at all, install GTM today, it’s 30 minutes. Through GTM you can add GA4 and Metrica in an hour. Another two hours - goals for the main conversion events. This is enough for the first month.

If you have analytics, but the data does not match the CRM, first check the UTM markup on all paid channels (missed tags = lost attribution), then compare attribution models, then look at the server-side.

Don’t waste time on a “beautiful dashboard” before you understand the quality of the data. A beautiful dashboard with dirty data is a costly illusion of control.

If you want to analyze your specific situation, write to Telegram or through form. Starting consultation - 0 ₽.