Performance
Speed Index
Also called SI.
Speed Index is a Lighthouse lab metric, in milliseconds, for how quickly the visible part of a page fills in. Lower is better, and it summarizes the whole load as one number.
How it is measured
Lighthouse records a video of the load, compares each frame with the final frame, and computes visual completeness over time. Speed Index is the area above that completeness curve. On mobile, under 3,400 ms is good and over 5,800 ms is poor.
It is lab-only, with no field API. WebPageTest also reports it from filmstrip frames, so numbers differ between tools. It carries 10% of the Lighthouse performance score.
Worked example
Two versions of a pricing page both have an LCP of 2.8 s. Version A paints a header at 0.9 s, the nav and copy at 1.4 s, and the table at 2.8 s, for a Speed Index of 1,900 ms. Version B shows a blank white screen until 2.7 s and then everything at once, and scores 2,900 ms.
The LCP is identical but the feel is not. The team adopts A's approach: render the plan table on the server and stream in the comparison chart later.
How it differs
Speed Index measures the whole progression of the visible page. First contentful paint is one instant, the first text or image. A page can have a quick FCP and a slow Speed Index if it paints a header and then waits for the rest. FCP ignores what happens after it, while Speed Index averages across the load.
Common errors
Expecting to find it in field data. Comparing Speed Index across tools. Forgetting that carousels, animations, and cookie banners change the frames. Optimising it with a fake splash screen. Reading it from a desktop run for a mobile audience.
In practice
Use it as a tie-breaker when LCP and FCP look fine but the page feels slow in the filmstrip. Look at frames at 0.5 s intervals and find the blank gap. Do not make it a primary target; LCP and INP are what your field data can actually track.