The process of selecting a subset of tests from a larger test suite, generally to run on a more frequent interval or at key points in the software development life cycle.
As test suites grow in size and test run times begin to hamper productivity, many teams choose to extract a smaller test suite or multiple smaller suites from a larger suite to run on more frequent intervals. Some teams create a suite of smoke tests - a non-exhaustive set of tests which cover the major features of the system. Other teams break the larger suite up into many smaller suites by functionality (or module).
Manual test selection is limited by human understanding of the system. It may still run more tests than needed - or worse, the wrong tests entirely - for a given change. Test impact analysis and predictive test selection offer a more modern alternative, utilizing static code analysis and/or machine learning to dynamically select the right tests for each change.