Negative 2000 lines of code

In the early 80's, the team behind the Lisa software decided to rate developers' productivity by the number of lines of code they would generate. It is a concrete example of what happens when measuring performance using single criteria.

Bill Atkinson quickly realized the number of lines of code was a bad metric, since it incentivizes code with poor performance and bloated architecture.[@hertzfeld1982-2000 Lines Of Code]

That's why he decided to optimize some part of the code he was working on, and submit a management form claiming he contributed $-2000$ lines of code.

That was all it took to review the performance metrics.

See: quantitative performance measurements produce undesired outputs

Found here: https://www.folklore.org/Negative_2000_Lines_Of_Code.html and referred by many different sources.