For me, it’s always look at the real user data first before trying to lab test a page. The real user data often points to different bottlenecks than the ones we assume. Or just test in general - never assume something is faster, just test before and after.
Ship less, do less, compress, priority, split, chunk, CDN, etc, etc. All these things, although all are exactly what you should do; in real life is not that easy.
The nomenclature online is full of abbreviations and tips and tricks and ways of improving every byte and every aspect of your performance online.
But most websites and most teams don’t build products, have fast turnarounds, low budgets, 0 know how about web tech, and most CMS solutions out there are not built to help with any of this unless you spend a lot of time and energy.
So, for me, the first rule is to make sure everyone understands that web performance is not about speed. It is about access. It’s about your users / customers and everyone else that comes to your website. Everything else is a by product of your happy users.