Analytics: GA4, Metrica, attribution

Without analytics, performance marketing turns into a religion: “we are doing something, it seems.” works." Analytics is not Y.Metrica and GA4 “for show”, but a whole system: frontend tracking, backend events, CRM communication, attribution model, dashboards for business. This chapter is a working analytics architecture for performance in 2026.

ga4 + i.metrics: why both

Ya.Metrica is the main tool in RF-performance. Yandex Direct does not work without it (no goals - no auto strategies), audiences for retargeting are not built, are not transmitted office conversions. The metric is installed on the site on day one.

GA4 (Google Analytics 4) is formally available to Russian sites, but Google in 2025 stopped taking payments from Russian cards for GA360. The free version of GA4 continues work. Why is it needed if there is a Metrica:

• BigQuery-export of raw data for free (Metrica - for money)
• The best data-driven attribution models
• Integration with Google Tag Manager and Looker Studio
• Standard for international teams and investors

Working practice: Metrics are the main tool, the source of truth for Direct. GA4 - for deep analytics via GTM + BigQuery + Looker. Details about analytics - separate blog article.

goals and events: minimal markup

Without properly configured goals, all performance is “traffic for the sake of traffic.” The minimum set of goals that we set on the first day:

1. Micro conversions. Scroll 75%, time on site 60+ s, click by CTA. Used to optimize cold traffic.
2. Lead events. Send the form, click “Order a call”, start the chat, PDF download. These are the main performance goals.
3. Purchase. Completed payment with transfer of amount. Without this goal there is no calculate ROAS.
4. Repeat purchase. The same event, but with the “repeated” sign - important for calculating LTV.
5. Cart abandonment. Abandoned cart without payment. For retargeting.

We transmit all events to both Metrica and GA4 via GTM. The purchase amount is transferred mandatory - without it, ROAS optimization in Direct does not work.

attribution: last-click vs data-driven

The main question of analytics: to which channel should the conversion be attributed if the user came first from VK, then from Direct, then from organic? This is it attribution — a model for distributing “credit” for conversion between channels.

Main models:

• Last-click. The conversion goes to the last channel. The simplest one The most common, and most erroneous, is that it underestimates the top-of-funnel channels.
• First-click. The conversion goes to the first channel. Reverse error - overestimates the channels of acquaintance, underestimates the channels for closing the deal.
• Linear. The conversion is divided equally between all channels in the chain. Better than last-click, but ignores different weights of touches.
• Time-decay. The closer to conversion, the higher the weight of the channel. It's logical but still manual distribution.
• Data-driven. The ML model calculates the contribution of each channel based on real data. Works in GA4 and Metrica for accounts with 400+ conversions per month.

What to choose: for single-channel performance, last-click is enough. For multichannel (Direct + VK + Telegram + organic) - data-driven, or own model in BigQuery. Read more about end-to-end analytics in glossaries.

Offline Conversions: A Major Blind Spot

If your business closes deals through calls or meetings (which is the majority of B2B services, real estate, auto, medicine) - standard web analytics sees only half the picture. The lead got into the CRM, then the manager brought him to the deal, payment passed in the bank - this is all “offline” for pixel analytics.

Solutions:

• Conversion API / offline conversions. Transferring the fact of a transaction from CRM back to Ya.Metrika and Direct. The Direct algorithm learns from real payments, and not on intermediate leads.
• Callback / Call tracking. Services like CallTouch, CoMagic, Roistat assigns unique phone numbers to each channel and links calls with the traffic source.
• UTM in CRM. When creating a deal in CRM, we save UTM tags. This the basis of all end-to-end analytics.
• Yandex Audiences sync. We export phones once a week those who purchased and load them into Y.Audience - Direct begins to optimize for those who “paid” and not on “those who left a request”.

long customer journey in b2b

In B2B, a user may touch a brand 12–25 times in the 4–6 months before a transaction. Standard cookies are stored for 90 days, retargeting 180 days - this is not enough. What works:

• Cross-device tracking via logged-in users (email-id)
• CRM as a source of truth - Metrica and GA4 as “sensors” for CRM
• Separate dashboards for full-funnel: TOF (knowledge), MOF (consideration), BOF (decision)
• Attribution is not at the conversion level, but at the transaction level in CRM

dashboards and reports: what to show to the team

The performance team spends an hour a week collecting reports in Excel - this is bad. That's right: automatic dashboard in Looker Studio / Yandex DataLens / Metabase, which is updated daily. Basic dashboard structure:

1. Overview. Total expenses, leads, CR, CPA, ROAS for the period. Comparison with the previous period.
2. Through the channels. The same metrics, but broken down by Direct / VK / TG / organic.
3. By campaign. Top 10 campaigns by expenses, by conversions, by CPA.
4. Funnel. Visit → Lead → Qualified → Deal → Paid. At every stage — CR and time.
5. Cohort analysis. LTV of cohorts by month of arrival. Helps you see deterioration in traffic quality.

Dashboard for the CEO - separate, on one screen: expenses, revenue, ROAS, goals for quarter. Without charts with 30 lines, no one watches this.

analytics stack checklist

1. Ya.Metrica is installed, at least 5 goals are configured.
2. GA4 + GTM are configured, events are duplicated.
3. UTM markup is consistent across all channels.
4. The Conversion API transfers offline transactions to Metrica and Direct.
5. CRM saves UTM when creating a deal.
6. Call tracking is configured (for businesses with calls).
7. The dashboard is updated automatically, there is a separate tab for the CEO.

in the next chapter

Chapter 12 - Budgeting. How to calculate the “CAC ceiling” through LTV, how to protect budget for the CEO and how to allocate money to the stages of the funnel.