Google's has a new metric: Speed Index. What is it and how is it different from Page Speed?
1) Speed Index vs Page Speed
Take a look at Google's documentation on their Speed Index. What it's measuring is how quickly the page contents are visually populated. The main difference between Speed Index and the regular old Page Speed is this:
Speed Index measures how quickly stuff appears up on the page. Page Speed measures how long everything takes to 100% load.
In other words, the Speed Index is how long it takes for you to start seeing stuff, Page Speed is (basically) the time it takes for the browser's little spinning wheel to stop spinning. If you want to get technical, Page Speed is when the onLoad() event fires.
Speed Index is a better indicator of user experience because if the content I want to see is on the page, as a user, I'm good to go. A lower Speed Index number is probably a good thing.
I don't really mind that a Flash ad is taking a few extra seconds to load. So this might be a small ranking factor. But please don't rush off and tell client's developers to optimize their Speed Index. It's new. Let's allow the dust settle a little. You'll get better results if you focus your efforts on the site's overall performance anyway.
[Added Note: Read Patrick Meenan's comments below. He is a Google Engineer. Patrick, thank you for leaving a comment!]
2) Google spells out Speed Index's algorithm explicitly
Did you notice they actually spelled out in plain sight the algorithm they use for this signal? (Hint: it's the last paragraph).
The Speed Index algorithm uses a histogram from video frames taken as the site loads.
That's a really simple idea. But if you're not a programmer you might not know this is really hard to execute in real life. Actually, let me rephrase that: It's not easy to execute at Google's scale.
Google specifically states that there are flaws in the algorithm:
There are some cases where it will not really be accurate but the tradeoff has proved to be very worthwhile for the vast majority of pages we have tested it against.
No algorithm with a dataset that constantly changes, like (ahem) the Internet, will ever be perfect. However, I'm sure they will continue to tweak it.
Personally, I'm thrilled that they published the technique itself. The black box isn't so black anymore.