Event tracking для аналитика

Зачем это знать

Без tracking — нет аналитики продукта. «Мы не знаем, нажимают ли на кнопку». Event tracking — foundation product analytics.

Analyst — часто designs, owns tracking plan. На собесах могут спросить.

Что такое event

Action пользователя или system:

• User click button
• Page view
• Purchase
• Error occurred
• Video watched 10 seconds

Captured с properties (context).

Event structure

Standard

{
  "event_name": "product_viewed",
  "user_id": "u_123",
  "timestamp": "2026-04-23T10:30:00Z",
  "properties": {
    "product_id": "p_456",
    "category": "electronics",
    "price": 1000
  },
  "context": {
    "device": "iPhone",
    "os_version": "17.0",
    "app_version": "2.3.1",
    "country": "RU"
  }
}

Типы событий

Identify

Link user к session:

identify(user_id, {email, plan, ...})

Track

Action performed:

track('product_viewed', {product_id, ...})

Page / screen

page('/products/iphone')
screen('Product Detail')

Group (B2B)

User part of organization:

group(org_id, {org_name, plan})

Tracking plan

Document of what trackable.

Columns

• Event name
• Description
• Required / optional properties
• Owner
• Use case

Example

Event	Description	Properties
product_viewed	User views product	product_id, category, price
added_to_cart	Item added	product_id, quantity
checkout_started	Checkout form opened	cart_total
purchase_completed	Order placed	order_id, amount, items

Naming conventions

Consistent

Pick format:

• snake_case (most common)
• camelCase
• Consistent verb tense

Verb + noun

• product_viewed (past)
• order_placed

Not click_product_button — too implementation-specific.

Event vs property

Event = action. Property = detail about action.

• Event: video_played
• Property: video_duration, video_category

Properties

User properties

Attached к user profile:

• signup_date
• plan
• email
• lifetime_value

Updated периодически.

Event properties

Specific к event:

• product_id
• amount
• session_duration

Context

Device, platform, app version. Usually auto-captured.

Tools

SDK

• Amplitude SDK
• Segment
• Mixpanel SDK
• PostHog
• Google Analytics 4

Server-side

Для reliability (adblock не affect):

analytics.track('purchase_completed', properties={...})

Hybrid

Client для UX events, server для critical (payments).

Implementation

Client-side

// React example
import { track } from './analytics';

function onAddToCart(product) {
    track('added_to_cart', {
        product_id: product.id,
        price: product.price
    });
}

Server-side

# After DB transaction
analytics.track(
    user_id=user.id,
    event='purchase_completed',
    properties={
        'order_id': order.id,
        'amount': order.total
    }
)

Pitfalls

• Missing events (bug, adblock)
• Duplicate events (retry logic)
• Wrong user_id (not logged in)
• Privacy (PII в events)

Data quality

Testing

• QA each new event
• Verify в test environment
• Check production regularly

Monitoring

• Event volume alerts (drop)
• Property completeness
• Schema compliance

Validation

Tools: Segment Protocols, Iteratively.

User identification

Anonymous

Pre-signup users. Device / session ID.

Merge

When signs up → link anonymous events к user_id.

«Merge» operation critical. Lose pre-signup funnel без.

Privacy

PII

Don't put names, emails, phones в event properties plain text.

Hash or ID reference.

Consent

GDPR, Russian ФЗ-152 — user must consent tracking.

Cookie banners, consent UI.

Anonymization

Older data — anonymize.

Real-time vs batch

Real-time

Events fire → immediately в analytics.

Good для monitoring, alerts.

Batch

Collected → sent периодически.

Efficient bandwidth.

Streaming platforms

Kafka, Kinesis — event stream. Analytics consumes.

Использование

Product analytics

Amplitude / Mixpanel — funnels, retention.

Data warehouse

Events + enrichment → DWH. SQL queries.

Marketing

Attribution, retargeting.

Engineering

Debug issues, error tracking.

Common mistakes

Tracking all, analyzing nothing

Over-track → overwhelming.

Focus: what questions want answered?

Bad names

button_clicked_on_screen_4_v2_new — unreadable.

Clear, semantic names.

No documentation

Team adds events ad-hoc. Confusion.

Maintain tracking plan.

Missing properties

«I want break down by X, но X not в properties». Missing opportunity.

Include key properties upfront.

Inconsistent

Two events doing same thing. Merge.

В companies

Russian

• Яндекс Метрика (free), AppMetrica
• MyTracker (Mail.ru для games)
• Alchemy / internal tools

International

• Amplitude
• Mixpanel
• PostHog
• Segment (aggregator)

Data model

Warehouse tables:

-- Events raw
(event_id, user_id, event_name, TIMESTAMP, properties_json, context_json)

-- User properties (current)
(user_id, plan, signup_date, ...)

-- Event schema registry
(event_name, property_name, type, required)

SQL query examples

Count events per type

SELECT event_name, COUNT(*) FROM events
WHERE TIMESTAMP >= CURRENT_DATE - 1
GROUP BY 1 ORDER BY 2 DESC;

Funnel

See SQL для воронок.

Drilldown properties

SELECT
    properties_json->>'category' AS category,
    COUNT(*) AS events
FROM events
WHERE event_name = 'product_viewed'
GROUP BY 1;

На собесе

«How instrument product tracking?»

Process:

1. Identify key user journeys
2. Define events per journey
3. Document tracking plan
4. Implement SDK / server-side
5. QA
6. Release
7. Monitor

«Pitfalls?»

Missing events, bad names, PII leakage, identification gaps.

Связанные темы

• Amplitude для аналитика
• Mixpanel для аналитика
• PostHog для аналитика
• Product analytics глубоко
• Data governance

FAQ

Сколько events?

Quality > quantity. 20-50 typical core events.

Real-time нужен?

Обычно batch OK. Real-time — specific use cases.

Tools ok без analytics platform?

SQL + warehouse. Works. Requires more engineering.

---

Тренируйте — откройте тренажёр с 1500+ вопросами для собесов.