Since the 1970s, the software industry has attempted to quantify code quality using various metrics, including Lines of Code (LOC), cyclomatic complexity, Halstead metrics, the Maintainability Index, and code coverage. The objective is to create a systematic approach to improving code, thereby reducing bugs, lowering maintenance costs, and streamlining development timelines.

However, the author argues that a single number cannot reliably determine if code is good or bad. This is not due to a lack of data, but because code quality is a multidimensional construct. While tools and CI pipelines provide constant reports, experienced developers typically identify quality by reading the code rather than relying on dashboards.

The article highlights structural failures in quantification through Goodhart’s Law and Weyuker’s theorems, suggesting that incentive systems often distort the measurements they rely on. Ultimately, the most critical aspects of quality remain unquantifiable, making the pursuit of a single quality score a futile endeavor.