ReliabilityIntermediate
Compare release reliability
Which checkout release coincides with higher errors and average latency?
Compare equal 10-day release windows from the service-event table while keeping request volume, failures, rate, and latency together.
The dataset and query remain in this browser.
Published expected result
Checkout error rate by release
2026.06.4
2.44%
2.44%
2026.07.1
9.33%
9.33%
2026.07.2
2.67%
2.67%
| release | checkout_requests | errors | error_rate_pct | average_latency_ms |
|---|---|---|---|---|
| 2026.06.4 | 900 | 22 | 2.44 | 310.5 |
| 2026.07.1 | 900 | 84 | 9.33 | 453.1 |
| 2026.07.2 | 900 | 24 | 2.67 | 309.5 |
How to read the query
- Filtering to one service avoids mixing workflows with different latency and error baselines.
- Each synthetic release has equal volume and a complete 10-day window, making the comparison shape inspectable.
- The query identifies correlation with a release; it does not establish that the deployment caused the regression.
Decisions the SQL cannot make
- 1Compare equivalent rollout populations, routes, regions, and observation windows.
- 2Use a tail percentile instead of an average when the engine-specific function has been validated.
- 3Confirm a suspected regression with deploy metadata, error categories, and traces.