A strong answer. “First I'd separate which kind of slow, because they have different causes. Slow to first content is a network and bundle problem. Slow to interactive is usually JavaScript execution blocking the main thread. Slow after loading is rendering or a data problem. A good Lighthouse score with unhappy users usually means the third, because Lighthouse measures load rather than interaction.

So I'd profile the actual interaction rather than the page load. In practice the causes I find most often are: a request waterfall, where three sequential fetches each wait for the previous, and the fix is parallelising or moving the join server-side; over-fetching, where the endpoint returns far more than the view needs; and a long task blocking the main thread — a large synchronous parse or an expensive render on a list that should be virtualised.

Re-renders and memoisation I'd check last, because they are the most commonly assumed cause and rarely the actual one, and memo everywhere adds its own cost.

And I'd measure with real user monitoring rather than a lab tool, because the users complaining are probably on worse devices and networks than my laptop — which is usually the whole story.”