Half a second delay caused a 20% drop in traffic

I heard Google also found that increasing speed increased users' perception of the quality of the search results. The reason was that when the site was faster, users did more searches. Naturally they found more stuff. But they attributed the gain to Google's cleverness rather than the extra work they'd done themselves.

Couldn't one come to another conclusion here? In fact several other conclusions. For one, since users were getting more results on the page they didn't need to go to a second results page, thus causing the drop-off in traffic.<p>If you're going to come to some grand "Aha!" moment, you should define your experiment. Were these the same users? on the same day? searching for the same stuff? What browsers were they using? Did they have their windows maximized? What was their screen size? ... and so on.

Another well-illustrated example of the fact that users don't know what they want, you have to A/B Test it out of them.

I think the magnitude of impact might be due to the search engine market. Results of different search engines are usually indistinguishable (to most web users), so the most important factors become design and speed. In other markets the balance may be different.<p>I'm not saying that speed isn't important for everyone, but having that killer feature, even if it slows down your site, might be what sets you apart from your competitors.

Maybe PG should try this in kicking his HN habit, introduce throttling features that will increase the time taken to render HN rather than focusing on optimizations, I know it would help a lot of other folks as well ;-)

I can understand this. Walmart is the most painful website to browse on the internet, because you go on usually looking for something or the price of something and it takes what feels like an eternity to get <i>anywhere</i>.

So, where does this fit into net neutrality? ISPs can certainly sell QoS services to customers.<p>For some applications, this is important (e.g. gaming). For others, it could be a nice competitive advantage. But then with more &#38; more of the ATM frames dedicated to QoS, that'll leave more contention (thusly, higher latency &#38; drop) for the rest.<p>I'm just worried about who's going to use these facts for what ends.

It's pretty much a non-news -- I think everyone is aware of this -- but backing it up with actual tests and data is <i>great</i>. Thanks for sharing the post

so, why not load the first few results quickly, then fill in the rest later through ajax? It'll take more than .5 seconds to eyeball the first ten results.

Sorry to bring back the table vs. css layout debate, but since Google use tables for some parts of their layout, can we assume that this is because it renders faster (and therefore makes more money for them)?