performance-regression-detection

Performance regression detection identifies when your application slows down beyond a 10% threshold by comparing current metrics against established baselines. It calculates a regression score from 0.00 to 1.00, tracks mean, standard deviation, min, and max statistics for each component, and requires at least 10 baseline iterations to establish reliable comparison points.

Establish baselines by running your performance benchmarks at least 10 times to capture statistical variance. The Skill records mean, standard deviation, min, and max values for each component, creating a foundation for continuous monitoring and regression detection across future test runs.

Yes, performance regression detection integrates with Serena for alerts and trend visualization, enabling continuous performance monitoring across branches and builds. This allows you to catch regressions early in your CI/CD pipeline and validate whether performance fixes actually resolve identified slowdowns.

Regression scoring generates a normalized score from 0.00 to 1.00 that quantifies how far current performance deviates from baseline. Scores above the 10% threshold flag components as regressed, helping you prioritize which slow components to optimize and track whether fixes improve performance.

The Skill records mean, standard deviation, minimum, and maximum metrics for each component across baseline and monitoring runs. These statistics enable statistical trend analysis and help distinguish genuine regressions from normal performance variation.

Use regression detection for continuous performance health monitoring, trend analysis over time, and automated alerting when performance degrades. One-off benchmarks capture a single snapshot; regression detection establishes baselines and flags anomalies systematically across development cycles.