This is totally fair and I agree with that. What I was trying to say and didn't express particularly eloquently is that you need to consider both "abstract" measures of code quality (performance, test coverage, complexity, rate of regressions/defects, etc.) and specific product metrics. You can deliver "high quality code" that checks off all the abstract metrics, but if it doesn't actually solve your business problem then you've basically just succumbed to Goodhart's Law.