I’ve noticed that it’s surprisingly easy to spend a lot of time optimizing CSS without seeing much real-world improvement.
Removing unused styles, reducing selectors, combining files, checking responsive rules, etc. can all make sense, but at some point I wonder whether the time would be better spent fixing larger performance issues such as images, JavaScript, fonts, or third-party scripts.
Where do you personally draw the line?
Do you optimize CSS from the beginning, or wait until Lighthouse/Core Web Vitals or real-user data shows that it is actually causing a problem?
It would be interesting to hear what optimizations have produced a noticeable improvement in an actual project, rather than just improving a Lighthouse score.