Is the Emporer Naked?

Are we all too accepting of Google's algos?

[ChrisXenon]

I know we have no practical alternative right now, except to play the game according to Google's rules, but I wonder if there is scope for feeding back to Google - somewhat forcefully - on the way things are.

In essense, the problem is that computers are very bad at understanding website content. They use a set of very simplistic metrics which are an extremely poor substitute for comprehension.

It is upon this heaving, lurching foundation of those very simplistic metrics that all of SEO, and the bulk of real-world web-use, not to mention virtually all web-based businesses are built.

Link popularity simply STINKS MIGHTILY as a way to determine if a website is a really great online shop - or a really awful one. It's a reasoning error to assume that a good shop will acquire many more links than a bad one.
And investing years to accomodate this fundamental stupidity - by building link popularity, is a truly sad and tedious way to pass time and spend energy.

HTML tweaks to move keywords up the page whilst preserving a table layout; padding text to make the body text as big as the leading website, re-naming pages to contain keywords - all of these are idiotic consequences of the fact that the search engines are so dumb.

Keyword Prominance, keywords link text, and on and on - are simply VERY BAD measures of anything significant - and of course, they are easily manipulated so that very bad sites, tweaked by very smart or very energetic people - get to the top.

And yet this is the state of the art on our planet today, and Google is the best of the lot. There is no real practical alternative here, nor any in site, so Long Live SEO.

But if someone finds a better way - the world is going to love it, they're going to become very rich, and Google is going to have to think again.

Chris

[Monkscuba]

The laser semiconductor physics site is an excellent one to look at when discussing Googles algo. It has a very good glossary of terms with photos of B.S. Interestingly if you search for the singers name, the laser site still ranks pretty well (No. 12 for me). The site above it in the SERP has NOTHING to do with Ms B.S., so I guess has a pile of incoming links with B.S. as link text.

Sometimes the SERPS are full of B.S. ;)

[Powdork]

Hello Monkscuba,
I knew to search for laser semiconductor physics because that was the title presented when I did the Britney Spears search. I'm not looking at the serps now but I presume your talking about that which rhymes with thavejik.mom. More proof the dissatisfied with results form is naked too, has been for several months. ;) I'm not saying SPAM here either, Googlebombing, perhaps.

<added>Yes, allinanchor is practically a Google serp mirror</added>

[Monkscuba]

Yep, googlebombing. But I guess a site appearing in an unrelated search only because of link text must be considered as canned pork luncheon meat?

Indeed if I search for "my name" (with quotes) my site comes up top. The cache says :

These terms only appear in links pointing to this page

Oops! I did it again.

We should have a little WW competition to see who can get highest in a B.S search without having any relevent content. Winner gets a free B.S.C.D.

[killroy]

One thing people tend to forget is that for every commercial subject (say "buying shoes") there is 100s or 1000s of commercial sites ALL OVER THE WORLD coverign it any 10s and 100s of 1000s of informational sites on the subject, and only 10 first page rankings. So no algo can treat all of them "fairly" at the same time, and often relevancywill be very close accross the topresults. Neither paid for listings nor different algorithms are going to change that.

The only real improvement you can have besides of course user interaction (i.e. more accurate keywords such as a locality) is to include more meta information, such as prices in thiscase, or shoe model names or brands and so on. But this is all information that needs more infrastructure in place (such as semantic web) and still doesn't really address the query in any more detail.

The big problem really is that short queries that aren't on a particular person, object or part of speech will NEVER capture truly unique results, by design.

The only true advance I see in this area is to replace the meta information with a semantic analysis that "infers" that meta information. But to be honest many pages do not truly include enough material to provide the context.

One major flaw in Google's view of the web is still that it looks at it as a series of documents, which isn't true. Many sites are applications and do not fit into keyword analysis and themeing systems.

For example I run a local directory. Because of its structure it ranks high on very competitivy keyhrases, but is useless because its very local. Furthermore it doesn't fit into dmoz style catigorisation since it covers 1000s of topics equally. It's just not a document, but an application. And without explicit meta information there is simply not enough context to algorythmically extract all the relevant features for a good result match.

SN