Time to interactive

Lighthouse finds FCP, then searches forward for a quiet window of at least five seconds with no long tasks (over 50 ms) and no more than two network requests in flight. TTI is the time of the last long task before that window. It is lab-only.

Because it needs a five-second quiet window, one stray late timer can push it out by seconds, which makes it unstable. That instability is why Lighthouse replaced it with Total Blocking Time and why field tools use INP for real responsiveness.

A conference registration page paints its form at 1.8 s. A chat widget then runs a 220 ms task at 6.4 s and polls every 3 s, so lab TTI reads 9.6 s even though the form accepted input at 2.5 s in manual testing.

After delaying the chat widget until the form is touched, TTI falls to 3.1 s.

Time to interactive looks for a quiet window after load. Total blocking time adds up the excess of every long task between FCP and TTI. TTI gives one timestamp that can jump by seconds, and TBT gives a steadier sum in milliseconds, which is why TBT is the one still scored.

Treating it as a current metric. Reading it as the first moment a click works. Trusting a single lab run. Using it for field comparisons. Missing that one late tag can move it by seconds.

If an old report quotes TTI, translate it: look at TBT in the lab and INP in the field. Use the long task list in the Performance panel to find which scripts occupy the main thread between 2 and 8 s, and defer them.