Google is built on old assumptions (or How to beat Google at search)

Perhaps its because I've been awake since 3 am wrapping presents for the kids, or perhaps its because I need more (or something in my) coffee before facing a house full of in-laws on 4 hours of sleep, but it occured to me this morning that Google is bass-ackwards with their ranking algorithm.

Let's look at the recent trend: paid links. Why are paid links bad? Because they skew Google's results. Google likes a nice democratic approach to web popularity, and paid links are simply purchased votes. If the same thing happened in a presidential campaign we'd be howling, but that's not my point. Remember, Google was built back in the day when search wasn't so great and tools like link directories were actually helpful. In those days, paid links were pretty rare, and a link from site A to site B really was a pretty honest vote, so Google was essentially built on the backs of link directories.

But what's really important in gauging a site's popularity and usefulness? The same thing that all the Anti-PR preachers have been preaching about: visitors! How honestly useful is a site that no one visits? How good can a site be when 100% of its visitors stay on the page for less than 3 seconds? Conversely, how popular is a forum when it has 10,000+ members with over 300 online and posting at any given time? Can't we easily say that one site is better than another based on its number of visitors and the duration of thier visit? Isn't it a simple matter to rank a site's popularity based on a simple computation of (# of visitors) * (length of visit)?

How do we implement this? With an easy bit of javascript (probably quite similar to urchin). Yes that will mean that some people will slip through the cracks, and that's fine. It is their perogative to not be tracked. Yes it means that it won't be 100% accurate, but how accurate is counting back-links or computing a page's keywords? With one small javascript module and a simple formula, you can quickly find and rank the sites that people not only visit but also spend time at. It is then a simple step of generating keywords for that page and attaching this popularity rating to it. To make things even simpler for webmasters, if you want a page indexed, you simply add the javascript to your page. Pages without the required code simply wouldn't be tracked.

There would be no need for spiders because a page either has the code and is indexed or it isn't. There would be no need for quality guidelines because if the page is a poor quality resource visitors won't stay. There would be no concerns over paid links because they have absolutely zero weight in the ranking formula and are reduced to their intended traffic generating purpose.

Its time for search to be simple *and* return results that are really useful, so someone please either point out the obvious loophole in this idea or offer me 10 shares when your new search engine has its IPO.

 

Googlexa?

I think the fatal flaw with your plan is it kills new startups. A lot of people rely on organic search engine traffic to build up their site's popularity. Just because Site-A doesn't have as many visitors as Site-B doesn't mean Site-A isn't a whole lot better than them. Example: MySpace is the most popular social network, but Facebook is at least 100x better.

The second flaw I can see is that it will invite a whole lot more black hat approaches to artificially influencing search results. How easy would it be to just buy one of those 10,000 visitors in 1 day auto-surf packages that guarantees visitors will stay on the site for xx amount of time.

I think that site traffic could potentially work as a factor in their algorithm, but definitely not as the sole-factor. Google just needs to put less weight on links and more on the quality of the content.

P.S. I'm sure there are enough people using the Google toolbar for them to get a decent idea of web traffic.

 

Ummm Google Toolbar?

 

What if they find exactly what they were looking for(like an answer) on that page in less than 3 seconds and then just closed it?

Or what if competition starts some bots to visit your site and close it in 0.5 seconds?

Or what if competition starts some bots to visit their own site and navigate on it for hours?

 

ermm that interesting...

 

My guess would be: Randomize!
One day you are first, the day after you are in the middle of nowhere.
More democratic than this........
Sorry, I am under effects of Prosecco.........

 

Googlexa, lol! I like that. I'm sure its available and I'd register it if it wouldn't get me sued!

I don't see this killing startups at all. A new brick and mortar establishment with lots of hype and no customers isn't a success, so why should a site with a huge PR and no visitor be a "success"? Think of it like bar-hopping: You want (are searching for) a bar (website) with hot+chicks and cheap+beer (keywords). If a tour guide (search engines) monitor the behavior of other customers (other visitors) who wanted the same thing and know that they spend 2 hours at Joe's Bar and Grill while only spending 5 minutes at Duffy's Tavern, I could fairly safely assume that Joe's has more hot+chicks and cheep+beer than Duffy's, so I would direct you to Joe's over Duffy's.

Quality is subjective. What if I think MySpace is 100x's better than FaceBook? What if I think they both suck? No, an algorithm needs to be as objective as possible.

I didn't think about the auto-surf things though. That would artificially inflate our stats, even if they or a bot only stayed .5 seconds. Perhaps the formula would have to be a little more than a simple number times duration.

Finding what you're looking for in 3 seconds and leaving does sort of pose a problem, but probably not in the way you're thinking if I understand your first comment about bots visiting and closing a site in .5 seconds. First off, if the formula is a simple # of visitors times the length of the stay, then a bot hitting a page 1000's of times would only inflate the popularity of that page, so doing it to a competitor would be a bad idea regardless of the length of its stay. Doing it to your own site would be a good idea though as you point out later in your post. It would also lead to people obfuscating their pages and hiding the crucial information in an effort to get people to stay longer. That's akin to hiding the bar so I can't order a drink soon enough.

Because of the correlation with real world customer behaviors, I think this idea definately has merit, but it obviously has flaws too. The Google toolbar is a great start, and Google obviously has the resources, code, and clout to implement this kind of change. I just hope they also have the brain power to work out the kinks and implement something like this. If they don't maybe someone with more free time than me can figure out how to work up this idea and make a hamn-dandy search engine. Google's great, but it certainly has room for improvement.

 

Good post.

Looking at all the responses so far, we should realise that there is no simple solution. I am no fan of google, but have to admit that judging an importance of a site based on its incoming links, was a good idea at first. It worked well until SEOs', Blackhats and the rest of the rats started buying and selling links to such an extent, that they own! the search results.

To clean up this mess, google played around with its PR, which in my opinion, was a bad move which could bite back because it has cause many to lose faith in the green bar.

 

The green bar has been lying for a long time. Even when it worked properly, it wasn't a measure of a websites popularity.

 

i agree... "green gold" was a lie, all the time.

three or four month ago i started thinking that google came up with its google's analytcs with the intention of a social "this site works for me" measuring... just like you posted above. Im pretty sure they are doing it... and you dont need javascript... just google analytics measuring your visits...