Forum Moderators: open
If you are referring to indexing how many people visit certain sites compared to others, how can this be acheived. Direct Hit failed - it never got beyond doing this reliably for a small number of one-term queries. Alexa's attempt uses a highly skewed sample, and even Alexa's supporters acknowledge that below say the top 1,000 or 10,000 sites things get very smudgy.
USe the Google toolbar? - another highly skewed sample.
Use "karma" type ratings like happy and sad faces and all the rest? - a recipe for spam city.
And even then you are restricted mainly to a site basis rather than a page basis.
and with millions of sites, how can google check the popularity of each one? - especially as the ranking of niche, specialist sites is Google's speciality? First Google would have to theme much better than it can now, so you are not comparing
apples to oranges.
Finally, it would involve the "...tyranny of the popular..", a bugbear of modern life where the most useful information in the world is hidden from public view because savvy marketers can make a transient lowest-commom-denominator talent like Brittany Spheres more popular than enduring talent.
It would be a step back in my opinion, as before Google has managed close to the impossible - to cut through a lot of the marketing hype to real useful information that may not indeed, be popular.
..some kind of popularity based algo should be included
I think for more frequent spidering it makes sense.
A site getting a disproportionally high visitors amount may be worth while respidering and reindexing more often.
In a way Pagerank has a bit of popularity in it:
Indeed, one way of viewing Pagrank is that it puts a number on how easy (or difficult) it is to find particular pages by a browsing-like activity.
from: www.almaden.ibm.com/cs/k53/www9.final/
What about those like myself that wish to deliver content with javascript rather than PHP or other server sided program?
The web has become fast enough to deal with other mediums than just text, yet Google relies 100% on text for its indexing.
Lots of people search on widgets, and they go to the widgetcity site. It is hugely popular. It happens to have a page on widget history, but it's pretty weak. None of those people searching on "widget" go to the Widget Museum site, because they just want to buy them, they don't care that Ole Olson created the original widget out of straw and lutefisk in his barn in 1863.
Now if you search on widget history you will still get that lame page on the widgetcity site instead of the widget museum.
With the current system, even if thw widget city page has higher PR, the widget museum site will probably rank higher for "widget history" because it will have many more external links with that link text.
I don't think there'd be anything wrong with doing it, they could do it without using toolbar info. For example, if the destination of a result is www.example.com, they send you to google.com/redirect?url=www.example.com, and they have all the stats they'd ever need on which listings get the best response.
The things that get clicked will be the things that appear most relevant to the searcher. If you were searching for SARS research, you're not going to click on Britney Spears results just because she's popular in general. You'll click where you think you're going to find what you're looking for.
Also, PR definitely has a lot more than 10 values, Google just chooses to define toolbar values that way.
In a way Google's very foundation is popularity measurement, only they base it on webmasters votes, not on users votes.
Including popularity among users in the algo in a meaningful manner would require a lot more than just a simple CTR style measurement.
Popularity in itself says nothing about the usefulness or relevancy of a result to a query.
Essentially Google would have to get a grip on how satisfied users are with a result the user clicks. Time spent on the page would be a bare beginning to measure that.
I don't even start thinking about the privacy issues involved.
I cant see yet any solutions that will be "better" than the present systems. Or ones that are scalable. For Adwords, there are a far less datapoints and people pay for the service, which means more budget can be spent on it. To clicktrack all clicks on google, analyse them and incoporate in the database would be a massive undertaking. And as i argued before, still open to abuse.
And how do you operationlize "popularity"? Number of unqiue hits? Amount of time spent on site? Amount of pages veiwed? Number of returning users? One of the major problems with systems like this is that popularity can be more due to marketing than the site itself. Savvy marketers can very well draw viewers to a site, but if its not very good they leave quick! If you just counted visitors, that site would be seen as "popular", but in the event, its not - it was just well marketed.
Remember with Google at least, an interactive site like yours can be optimized by almost all known techniques that are very influential. Title tag, incoming hypertext links, link popularity from other sites. That's even before you start adding text on page to augment your other material.
We have to optimise for all promotional options. Google is just one, and it specialises in plain text sites, because generally people want to find info quick and less often want to go to a slower loading multi-media presentation.
I know Google accounts for a lot of SE traffic now, but that will not always be the case.
One thing you have to do, if you want google traffic is make compromises.
One thing we did was to change our home page to almost all text, small, and very simple. Then the more complex stuff goes on pages with links from the home page.
On sites that we dont want to make any compromises at all, we dont worry about Google and use Adwords or paid advertising.
The "predated" backlinks shown in www2, www3, and sj appear to imply that G are VERY seriously comtemplating the new popularity algo as the strange behaviour seems to run only on secondary datacenters. Have any one seen the dance in www? For me, I don't see any dance since the update began around 48 hours ago, which is not the same as before. Very curious to see what would be the new popularity algo which might mean another round of sweating again
"Methods and apparatus consistent with the invention provide improved organization of documents responsive to a search query. In one embodiment, a search query is received and a list of responsive documents is identified. The responsive documents are organized based in whole or in part on usage statistics."
from one of the Google patents as Rubble88 posted here: [webmasterworld.com...]
