Looker и LookML для аналитика
Карьерник — квиз-тренажёр в Telegram с 1500+ вопросами для собесов аналитика. SQL, Python, A/B, метрики. Бесплатно.
Зачем это знать
Looker (Google Cloud) — один из enterprise BI-инструментов вместе с Tableau и Power BI. Особенность: semantic layer через LookML, версионирование метрик. В российских компаниях используется меньше, но международных tech-company job descriptions часто упоминают.
Что такое Looker
Enterprise BI platform от Google:
- Connect к data warehouse (BigQuery, Snowflake, Redshift)
- LookML — semantic layer
- Explores (ad-hoc analyses)
- Dashboards
- Embedded analytics
LookML
Ключевая feature. Version-controlled definition метрик.
Pseudo-example
view: users {
dimension: id {
primary_key: yes
type: number
}
dimension: email {
type: string
}
measure: count {
type: count
}
}
view: orders {
measure: total_revenue {
type: sum
sql: ${TABLE}.amount ;;
}
}
explore: users {
join: orders {
relationship: one_to_many
sql_on: ${users.id} = ${orders.user_id} ;;
}
}Explores
Ad-hoc interface: select fields, Looker генерирует SQL.
- Pick measure (revenue)
- Pick dimensions (country, month)
- Add filters
- See result + SQL
Dashboards
Compose charts из Explores. Reusable, shareable.
Persistent Derived Tables (PDT)
Materialized views:
derived_table: {
sql: SELECT user_id, SUM(amount) AS total FROM orders GROUP BY 1 ;;
persist_for: "24 hours"
}Performance optimization.
Git integration
LookML всегда в Git. PR workflow для изменений.
Это big advantage: metrics reviewable, versioned, documented.
Plusses
- Single source of truth для метрик через LookML
- Version control git-based
- Powerful semantic layer — метрики defined once
- Embedded analytics для customer-facing
Minuses
- Expensive — $35k+/yr
- LookML learning curve
- Complex setup
- Acquired by Google — enterprise-focused
Looker vs Tableau
| Looker | Tableau | |
|---|---|---|
| Setup | LookML upfront | Direct connect |
| Semantic layer | Yes (LookML) | Basic |
| Visualizations | Good | Great |
| Ad-hoc flexibility | Limited by LookML | More flexible |
| Cost | $$$ | $$ |
Looker vs Metabase
| Looker | Metabase | |
|---|---|---|
| Cost | Enterprise | Free / cheap |
| LookML | Yes | No (simpler model) |
| Setup | Complex | Quick |
| Target | Enterprise | Startups, SMB |
Metabase — entry-level. Looker — enterprise.
Looker Studio (free)
Google's free data studio, рекомендация users free dashboarding. Different product. Не confuse с Looker enterprise.
Типичные задачи
Create dashboard
Define metrics в LookML → Explore → Chart → Dashboard.
Ad-hoc analysis
Explore: filter, group, pivot.
Scheduled reports
Email dashboards daily/weekly.
Embedded
Put dashboards в customer-facing products.
На собесе
«Looker experience?» LookML development, Explores, dashboards.
«LookML vs raw SQL?» LookML — version-controlled metrics. SQL каждый раз — inconsistency risk.
«Looker vs other BI?» Semantic layer + enterprise features vs simpler tools.
Частые ошибки
Overcomplicated LookML
Начать simple, добавлять complexity when needed.
Not using Git
LookML сила — в version control. Используйте properly.
Bypassing semantic layer
Raw SQL бок о бок с LookML создаёт inconsistency.
Связанные темы
FAQ
Free version?
Looker Studio (бывший Google Data Studio) — free. Looker enterprise — no.
Looker работает в России?
Предоставление через partners, but slower than international.
LookML vs dbt?
dbt — transformations в warehouse. LookML — semantic layer поверх. Compatible, но разные slots.
Тренируйте BI — откройте тренажёр с 1500+ вопросами для собесов.