INP is not FID with a new name
FID measured the first interaction. INP measures the worst one, and that is a completely different bar to clear.
FID was generous by design: it measured input delay on the first interaction only, and only the delay, not the work. Most sites passed it without trying.
INP measures the full latency of the worst interaction across the visit. A page that is fine on load and terrible after the third filter click now fails, correctly.
Look at your worst interaction, not your average
The metric is a high percentile of the worst event. Averaging your interactions hides exactly the one being measured.
Comments
3 Newest firstWorth adding that the worst interaction is often a filter or a sort on a long list, which nobody profiles because it is not on the load path.
We passed FID for three years and failed INP on day one. Same site, same code. That was a humbling morning.
That is the common story. FID was measuring the easiest possible moment, and almost everyone cleared it by accident.