What gets measured
The database first: slow queries, missing indexes, the searches that walk
the whole catalogue. Then the front end - what the browser downloads,
parses and executes before a person can click anything.
Core Web Vitals come from real visitors, not from a lab run on a fast
laptop. So do the conclusions.
What usually turns out to be true
On a large catalogue the search is the first thing to stop scaling, and no
index fixes it: at the worst point a search took eight seconds while the
warehouse waited. Moving it out to Elasticsearch took that to one.
On the public side the answer is usually weight: a page that shows text
should not ship half a megabyte of JavaScript written for a shop you do
not run.
What you end up with
Numbers before and after, on your own system and your own data - not a
promise of “up to three times faster”. Plus the list of what was changed
and why, so the next person can read it.
And an honest line about what is not worth fixing: some of it is cheaper
to live with than to rewrite.