Predictive Test Selection

Develocity trains a predictive model by observing code changes and test outcomes from your Build Scan data. It combines this per-project learning with training across many Develocity installations, so predictions are meaningful even before your project has accumulated extensive history.

Before each test run, the model cross-references a snapshot of the test suite and code under test with change and outcome history. From that comparison, the model selects the subset of tests most likely to provide useful feedback on the changes under test.

The model updates with each new Build Scan published to Develocity, so predictions improve as more build data accumulates. Tests that are recently new, recently changed, recently failed, or recently flaky are always selected.

A selection profile controls how the model balances prediction confidence against time savings. Develocity provides three predefined profiles:

The Predictive Test Selection Simulator in Develocity lets you assess the expected benefits and risks of enabling PTS for your builds using your historical build data. It visualizes predicted failure detection rates, savings potential, and avoidable compared to unavoidable test counts, and lets you compare outcomes across selection profiles.

Predictive Test Selection trades testing comprehensiveness for faster feedback. It suits builds where reducing feedback time is the priority, such as local and pre-merge checks.