Agent skill
GA4 API Limitations, Quotas & Cardinality Boundaries
What the Google Analytics 4 Data API can and cannot do, where to find alternate data sources, and how to query within Google Cloud quota boundaries.
Capabilities & Workaround Matrix
| Analytical Requirement | Available in GA4 Data API? | Recommended Alternative |
|---|---|---|
| Multi-Step Funnel Visualizations | Partial (Aggregate e-commerce metrics only) | GA4 UI → Explore → Funnel Exploration |
| Cohort Retention Over Time | No (API does not expose runCohortReport) | GA4 UI → Explore → Cohort Exploration |
| Individual User Clickstreams / Journeys | No (API returns aggregated rows only) | BigQuery Export (Raw Event Tables) |
| Segment Overlap / Venn Diagrams | No (Aggregate dimension cuts only) | BigQuery or GA4 Explore Comparisons |
| Realtime User Event Streams | Limited (Last 30-minute summary only) | GA4 UI → Realtime Report |
| Google Search Console Clicks & Queries | Separate API (Not in GA4 Data API) | Google Search Console MCP Server |
| Historical Data Prior to GA4 Migration | No (Universal Analytics data is separate) | BigQuery UA Archive Tables |
Technical API Limits & Guardrails
When querying GA4 through an agent, keep requests within Google's structural limits:
- Maximum Dimensions per Request: 10 dimensions. Attempting to pass 11 dimensions returns an API validation error.
- Maximum Metrics per Request: 10 metrics.
- Maximum Rows Returned: 100,000 rows per query response (subject to client memory limits).
- Hourly Token Quotas: Standard properties receive 25,000 tokens per hour; GA4 360 properties receive 250,000 tokens per hour.
When to Use BigQuery vs. GA4 MCP
- Use GA4 MCP for conversational exploration, executive reporting, attribution channel audits, e-commerce summaries, and ad-hoc troubleshooting inside Claude, Cursor, or your terminal.
- Use BigQuery Export when you require row-level user journey tracking, machine learning feature engineering, joining analytics with CRM database tables, or querying data older than GA4's 14-month user data retention limit.