Forum Moderators: Robert Charlton & goodroi
In my eyes it makes sense, as a couple of people have mentioned, no matter how many links you have, if Google ranks you well and then people come to the site only to leave 2 seconds later then there is obviously a problem with it. So - Google thinks, ok, if a user is not staying on the site for any length of time at all and then perhaps clicking the "back" button to go to the search results, this is something we need to take into account when ranking a site.
A lot of people haven't installed Google analytics as it's just giving more and more data to Google about users and what they actually do in your site. The way I used to see it, is that if I don't have analytics, and if the user is not going to click back to google, then google effectively will have no idea what people are doing on my site (unless they use cookies etc). It can't track user paths through my site, because without my permission that would be a bit underhand.
So what behaviour does Google use in the algo to rank sites? If their data is very limited what can they actually do?
But... and here is the big caveat.. companies like Hitswise buy ISP data on a huge scale and it gives them complete history / user paths, user behaviour etc etc. The ISP data has information on which keywords people type in, what sites they use and click on, how long they spend on those sites etc etc. (Please correct me if I'm wrong here!). Even if a user is not going through a SE the ISP data can completely profile the internet behaviour of users.
So, wouldn't it be possible for Google to just buy this data, just as Hitswise does? In which case they may have access to complete information about what everyone does on the internet. How good would this be for their algo.
I'm not quite sure what to make of this possibility. In one way it doesn't surprise me, in other way it scares me slightly. It makes me glad I'm white hat, but it makes me slightly nervous that I'm not doing everything I can in the right places, perhaps not working quite hard enough at visitor retention etc etc.
Who knows - maybe google knows absolutely everything, in which case the only way forward in my opinion is to go back to basics - get the best product you can ever get, blow your competeition out of the water with it, and then customers will come, they'll like it and spend time on your site (depending on what you do and sell of course!) and links will come naturally. You may have to remember SEO when considering architecture aspects and some of the content but it won't be anywhere near as difficult as a lot of people are finding it now.
Pipe dreams eh... who wouldn't love the perfect product / service / good! (actually isn't this how Google got to be the number one SE?! I'm not saying they are perfect by any means, but they give better results than other SE's.. and everyone seems to use them)
As you mentioned, if Google's going to integrate a use behavior component, they certainly need a significant source of data and not just a thin selective slice. Also, as Matt Cutts has mentioned, it's quite problematic to use "time on the page" - sometimes a user can see exactly what they need in a quick glance - an acronym definition, for instance.
While discussing how to integrate manual ratings into the SERP calculations, the 2006 Human Editorial Input patent [webmasterworld.com] makes mention of three main scoring components used in the Google algorithm:
...the applicable editorial opinion parameter of a web site may selectively affect one of the scores used in determining the final ranking (e.g., the text match score, the connectivity-based score, or the popular opinion score) [emphasis added]
I translate that as:
1. Query dependent or relevance factors - on-page, anchor text etc
2. Query independent factors - PageRank, quality/trust/burstiness of backlinks
3. User data
I've got a site which ranks very well, and some affiliate pages on this site, with some slightly related content (not 100% the same as the main focus of the site). Some of the affiliate pages yo yo something chronic, even though they provide a really good resource, and has constantly updated unique content and some deep links to them.
For the main keywords for the site we're rock steady though.. it makes me think that there is some user behaviour (i.e. people clicking away from the site) which could be causing this..
Would that be possible - a page that has the majoirity of traffic leading away being seen differently.. mm.. but no, it still provides a really good resource, even if people leave
As you mentioned, if Google's going to integrate a use behavior component, they certainly need a significant source of data and not just a thin selective slice.
In my country, at elections times, some polling agencies interview a few thousand people (say 2,000 or 3,000) and then can predit the results of the election (over 100,000,000 voters) within a margin error of 3%.
This is made with statistics, and Google certainly has some good statisticians in staff.
Highly unreliable...
I'm going to give an extremely common search that I do at least 10 times a week.
Searching for the Lyrics of a song
- Situation 1. Find the song, see the couple words I wasn't understanding and click back - 3 seconds on site
- Situation 2. Isn't the song I'm looking for. Spend average of 1-3 minutes on site searching for lyrics. Find them
- Situation 3. Isn't the song I'm looking for and instantly know site doesn't have the song. - 3 seconds on site.
A garbage site is a garbage site. They pay people $10 an hour (or nothing if they are Goog-webmasters slaves) to locate these.
But somehow they are going to develop AI to determine what the user's intent is for searching?!
lol, this reminds me of Man's eternal search for the Fountain of Youth.
The day Google masters this technology is the day of the apocalypse.
Even the greatest marketers of all time are able to "read their customers minds" at no greater than 20% accuracy
Another possible source of user data - Google now owns DoubleClick.
Think you all have this reversed.
Google is more interested in selling user data than incorporating it (to any serious degree) into their algo.
It's much more profitable and what they've been logically leading to over the past 3 years.
IF they were still a edu project, I could see them being naive/ambitious enough to attempt a hardcore implementation of user analytics into their algo, but those days are gone and past and the truly brilliant minds have retired in style.
