How important is "clickability" in ranking a site?

[siorange]

For example, if Site A and Site B are ranked 1 and 2 respectively on a particular search result and MORE users click on site B rather than A, would the rankings eventually flip flop? I'm curious about this because what if site A has more relevant content but isn't as clickable as site B, would Google take that into effect in future rankings?

[DarkMatter]

I don't beleive that Google uses click frequency in it's relevancy calculations. It would be too easy for people to repeatedly click their own sites or create bots to do so.

[siorange]

Ahhh that would make sense. Are these bots common nowadays? I mean i've heard about them being used on paid listings, but is google doing anything to combat this problem?

[Pannu]

Two sides to every picture as they say; you got a point there Siorange.

[randfish]

In the recent patent, Google says they will be doing just this, but as DarkMatter noted, this could be quickly spammed if it was implemented too strongly - probably still a small ranking factor for certain search terms.

[PhilC]

Google does record clicks on the listings - they use an onmousedown event to do it. What they do with the data, if anything, is unknown as far as I know, but I wouldn't be surprise if they use it in that way at some stage.

[Marcia]

I think there is a distinct possibility that they use clickthrough data in some way, but they certainly would not want US to know about it. :D

[rustybrick]

DirectHit - hear of it? if not, there is a reason why.

[siorange]

well after doing a quick search it seems like MSN, lycos, askjeeves use it to some capacity to rank pages. But the part about it not being able to track the dial up spammers is pretty scary.

[PhilC]

I don't think you can attribute DirectHit's failure to their click tracking, RB. They came into a crowded marketplace and they just failed - they can't all succeed. Wisenut? Teoma?

[Mike Grehan]

I don't think you can attribute DirectHit's failure to their click tracking, RB. They came into a crowded marketplace and they just failed - they can't all succeed. Wisenut? Teoma?

I think you'll find that DirectHit is still folded into Teoma if you research a little bit more. Just a little hint from me that it didn't die completely.

[PhilC]

It probably is Mike. I was making the point that it didn't didn't fail because of the click-tracking, imo. It failed because not all engines make it, and it entered a well populated arena.

[Mike Grehan]

It probably is Mike. I was making the point that it didn't didn't fail because of the click-tracking, imo. It failed because not all engines make it, and it entered a well populated arena.

Here's an interesting thing. Click tracking may be a way to get demoted as opposed to promoted. I have a much longer section in my new book about this, which really is startling. But let's go back to my second edition first, and Andrei Broder:

Mike: Does it work the other way around as well Andrei. In that, if I perform a key word search and then click on the first result, if I don't like what I see and then click 'back' to the results page again, if that happens with this page on so many occasions, then in fact it could be penalised for it.

Andrei: That's correct, that can happen. You know Direct Hit was based on this idea. And they also study how to do 'inferences'. How to do data mining on the click patterns. They look at, like you say, if you click quickly onto a page and only stay on it for five seconds then it's not satisfactory, but stay on another page for 50 seconds then this seems satisfactory. At this point at least, we do click tracking in a way, to validate some of the changes we're making and for Spamming filtration, but we don't use that information for ranking. If I notice that for a particular query, like say, anaemia again, if I see that this page gets clicked a lot, I'm not going to change it's ranking just because of that. Direct Hit would do this, but we do not. So the paradigm there, I guess, is that people vote with their feet as in number of clicks. And the paradigm used at most engines, including us, that people vote with their links. So a page that has lots of links, that's a measure of popularity. The click thing may work, but very often it's a case of how good your abstract is, or how fancy the graphics are, you know, they may have a fancy site, but it may not be appropriate. So you have to be very careful how you use this kind of click thru information.

[Marcia]

I think you'll find that DirectHit is still folded into Teoma if you research a little bit more. Just a little hint from me that it didn't die completely.
http://www.directhit.com

That URL brings up Teoma.

[PhilC]

That was the idea behind DirectHit - the length of time it took a person to hit the back button. That's how pages slid down the rankings, or climbed up them.

I don't think that recording clicks on its own can provide anything realistically positive for a page. Maybe an engine could consider a click on its own as a positive vote, but that's not a realistic view, and they are unlikely to do that, imo. To my way of thinking, relating clicks to pages is a test of how long it takes the user to get back to the serp - it's always a measurement of which pages fail the least.

The big question when DirectHit first came out was, how on earth can a new page get up the rankings if the higher places are already populated by "popular" pages. DirectHit said that they gave an initial boost to new pages - they give them a chance at the top to see how little they fail. The problem with that idea is that once a new page has failed, and slipped down to where not enough people will find and click on it, it can never climb the rankings again because it will never be a new page again. It's a bad metric to base a search engine on.

Google does have the method of recording clicks in place, and they almost certainly record them. What they do with the data isn't known, as far as I know, but the sensible approach is to *make* the listings clickable by writing very clickable Titles, and even by organising the snippet to attract clicks. That way, the site gets more traffic and, if the clicks matter in Google's system, they get that benefit too. But design the landing page to keep people in the site - for both reasons.

[Mel]

It would seem more likely to me that the clicks tracked by Google would be used more profitably to refine the algo by way of measuring what was the ranking of the first page the user clicked through. Thus if they clicked on the #1 ranked link the alog might be thought to be doing a good job, but if they clicked on the number 10 result first it might be considered that the top ranked results did not appear good to the user.

[Marcia]

In the interview with MG, Andrei Broder made several references to "we." He's with Yahoo now, but I'm wondering which search engine he was with at the time of that interview.

IMHO the issue of click-tracking affecting rankings is a relevant issue for today, particularly in a negative sense, as implied by Dr. Broder.

[Mike Grehan]

In the interview with MG, Andrei Broder made several references to "we." He's with Yahoo now, but I'm wondering which search engine he was with at the time of that interview.


He was Chief Scientist with Alta Vista at that time.

[sootledir]

They've had millions of these toolbars installed for years. You think they're ignoring the data they collect?

[Iguana]

There is one place where Google definitely uses clickthrough data - that middle-of-the-page "more results" section. There is no other way I can think of that would have led to "Ashley Cole" being given a More Results suggestion relating to rumours started in a newspaper and leading to Google being asked questions by his solicitor.

Another example is searching for the little known but up and coming UK band This Et Al (one of my favourites). The is a more results section that suggests "This Et Al" (speech marks around) - very useful because the normal search picks up a lot of "et al" mentions in high pagerank sites. Checking the Overture suggestion tool I see about 80-90 searches in a month for the band. Since I have been using the band name as a penalty/sandbox test for two of my sites over the past few months then it was probably me that caused the "more results" to do that!

Seriously though, how many users will have known to use the speech marks and yet Google has picked it up for a little used search phrase. Makes you wonder about the possibility of manipulation.