[neilk]

Your own perception of why you switched may be partially in error.

Google has done a lot of research to show that speed matters quite a lot to users. When results appear in less than 250ms the user is much more satisfied than even 500ms. Anything less than a quarter of a second is thought of as "instantaneous" and the user starts thinking of it very differently.

Here, they are taking it down to less than 100ms, which is now operating faster than the user can even perceive (for that matter, it's faster than many users can type). So the real effect is that they have now eliminated the need to click on "Search". So now Google is even more interwoven into your thought processes, and other search engines are going to seem unbearably slow pretty soon.

[nanairo]

True. But you'll have to agree that such argument have a diminishing marginal return pattern. 500ms vs. 250ms may be great, but what about 100ms vs. 50ms? Once you reach the point my brain is the bottleneck, what's the point?

On the other hand it could be distracting to get something happening in the screen while you are thinking the best way to form your query. Usually when I am about to ask someone a question and that person tries to give me the answer much before I have finished, it can get a bit annoying.

So I am not convinced this change in speed will bring much benefit, and it risk (from the sound of it) to be distracting. However this is just my gut feeling. I'll still give it a try, and I am sure other people will have different opinions.

It's good to see Google trying to improve the search interface in any case.

[kwamenum86]

Experiments have shown that 100 Ms is the threshold.

[jyothi]

I think speed beyond a point doesn't add much to the experience. One would spend at least 2-5 seconds to check out the results and minutes to find the right article, few milliseconds would not matter.

Instead if they return 'good results' in 1000 milliseconds also I would continue to use them.

[kwamenum86]

If you examine the empirical evidence this is patently false. Milliseconds, on average, on the whole, in the long run, at scale, however you want to phrase it, do matter.

[jyothi]

in mission critical applications. I really hope there is some good competition for google.

I have to stress again on what the first parent commenter. I am a more genuine SEO person myself and sometimes it is just so annoying that google doesn't even invest in cutting down all the spam on first page result.

Almost seems like they are focusing just on the technical scale & adwords impressions. Given the talent at Google it shouldn't be hard to put in checks for gamers without anything useful on the site.

PS: disagree but down vote somehow doesn't make sense. There is nothing offensive or irrelevant in my comment.

[Eight]

For now

[daychilde]

The nice thing, though, is that there's an option to turn it off for those that don't like it.

I'm not sure if it'll prove to be too distracting vs. mildly helpful for me, yet.

[peterwwillis]

it's funny you bring this up because we were just discussing site performance at work today. what i took away from it is this: speed is mostly relevant when there's a competing product providing similar value and you risk losing business due to inferior user experience.

if google were far better than their competitors at providing what users want (relevant search results they don't have to dig through) they could take 5 seconds and it wouldn't matter. the added value of getting what you want on the first try at the top of the page would keep people using that product.

yes, 500ms can seem like an awful lot when you come from a world of 250ms. but if the results returned from the 500ms are significantly better, you bet your ass users will sit through double the time to get better, more reliable results.

[andreyf]

yes, 500ms can seem like an awful lot when you come from a world of 250ms. but if the results returned from the 500ms are significantly better, you bet your ass users will sit through double the time to get better, more reliable results.

Are you sure?

[lhnz]

Are we talking about opinions here? If so, I have mine, and: I agree with him. I might spend 500ms waiting for my search results but my greatest annoyance is still having to spend 10 minutes trawling through SEO 'content' to find a legitimate website.

[andreyf]

Are we talking about opinions here?

Sorry, that was the point. We could talk about opinions, sure. But all major search contenders do significant user research into how people will react to much less significant changes than Instant Search. This change had its but user-tested and dogfooded for quite some time.

[peterwwillis]

IANASEOE, YMMV, HBD, ETC. but for us, yes. the site of my company was balls slow yet managed amazing feats of capitalism, mostly due to a faithful userbase in the face of faster and cheaper alternative sites. value won out.

an example: Steam, the game delivery system, takes longer to deliver and properly execute games than it takes to buy one from the store, install it and run it. but you get more value from steam so people sit through hours or nights/days worth of downloading to run its games.

[andreyf]

Good point about Steam. It really does suck balls compared even to the vast majority of web apps. Ditto for iTunes: slow as balls sometimes, but I still use it to buy music sometimes.

Still, a competing product will win both over if it can get sub-100ms response time for all of their UI interactions. Buy and listen to a song? Imagine hearing it instantly after you click "buy & listen".

[dedward]

The results would likely have to be significantly different.

Google and others have studied the response time -vs- user stickiness issue extensively, and it's good stuff to read - users don't even know what they want.

Example (paraphrased) - the majority of users will say "I would rather see 20 results on the page than 10" - but actual A/B testing shows that the users shown only the 10 results (faster, less information to process) are overwhelmingly more likely to continue using the service.

What people think they want when asked is very often not what really happens in practice.

So if a search engine is giving you results you hate - response time is obviously meaningless - but otherwise, it's everything.

[bad_user]

> what i took away from it is this: speed is mostly relevant when there's a competing product providing similar value and you risk losing business due to inferior user experience

Even if you don't have competition, you leave the door wide open if requests take 5 seconds.

There's also indirect competition ... like a lot of people type the name of the service they want directly into Google's search box ... a practice which would stop if it took 5 seconds, because accessing your local bookmarks would be faster.

[_urga]

Less than 100ms is widely considered to be instantaneous. And it is instantaneous if you're coming from a world of 250ms. But people can in fact distinguish between 20ms and 100ms. A better definition of instantaneous, for the purpose of UI design, would be that 100ms is the threshold at which people start to take notice that something is NOT instantaneous. It's not enough to target sub-100ms. One has to target sub-50ms. Anything more than 50ms ruins the magic.