Why Global Coverage Lies

Global coverage answers: "Did tests execute enough of the project?"

Parity asks the stricter question: "Does each file's matching test cover the file it is supposed to protect?"

The Hidden Failure

Metric	Result
Project/global coverage	80%
File coverage from all tests	70%
Coverage from the matching test file	40%
Incidental covering tests	2

The 70% file coverage can look acceptable because integration, smoke, or workflow tests touched the same lines. That does not prove the matching test owns the behavior.

The Parity Signal

minimum-coverage reads all-test file coverage:

- minimum-coverage:
    min: 70

matched-coverage reads only lines covered by the expected matching test file:

- matched-coverage:
    min: 40

coverage-attribution reports how many tests covered the file and how many were non-matching.

{
  "minimum-coverage": { "value": "70%" },
  "matched-coverage": { "value": "40%" },
  "coverage-attribution": { "value": "3|2" }
}

3|2 means three tests covered at least one line in the file, and two of them were incidental non-matching tests.