PostHog Dossier • Benchmarks • Empirical Testing
PostHog Ingestion & Query Latency Benchmarks
Controlled query benchmarks testing PostHog and ClickHouse performance across 10 million, 100 million, and 1 billion recorded user events.
| Dataset Volume | 3-Step Funnel Query | Retention Matrix | Disk Compression | Ingestion Latency |
|---|---|---|---|---|
| 10 Million Events | 85ms | 140ms | 4.8x ratio | <500ms |
| 100 Million Events | 340ms | 680ms | 5.2x ratio | <750ms |
| 1 Billion Events | 1.8s | 3.4s | 5.6x ratio | <1.2s |
Key Engineering Takeaways
1. Vectorized Aggregation Speed: Because ClickHouse uses vector SIMD CPU instructions, filtering through 100 million rows to calculate a 3-step conversion funnel completes in 340 milliseconds, compared to over 30 seconds in traditional relational databases.
2. Exceptional Disk Compression: Repeated string values (such as URLs and event names) achieve up to 5.6x compression ratios with ZSTD/LZ4 encoders, drastically reducing cloud storage costs.