Sphinn - Think Like a Search Engineer

43
Sphinn It!

Think Like a Search Engineer

Went Hot: January 4, 2008 - 2:34 am
Posted By: DoshDosh 549 days ago
Topic Type: News Story (Jump to http://www.seomoz.org)
Category: SEO

What do Dan Thies, Aaron Wall, Dave Naylor, Dave Davis, and select others like Hamlet Batista have in common? They think like search engineers. Here are some ideas to get you started in the same mindframe.

16 Comments

Who Desphunn This (0)

Comments

from Gab 549 days ago #

Votes: 0 | Vote:

+ -

I had a feeling this would wind up as one of your submissions Maki :).

from emanuelh 548 days ago #

Votes: 1 | Vote:

+ -

You present us with a classic decision-making issue and reach the wrong conclusions. The mindframe you propose is to think like a search engineer (class of 1998?) but the real challenge is to think like Google's team of engineers (class of 2008). And this is impossible

I may think of a an issue demanding my attention for ten minutes and make what it seems to be a rational decision. Given an hour I may consider additional details of the pros and the cons and make the opposite decision. Given a month I may find out that the first decision was much better after all.

Google can afford a practically infinite length of time and a very large team of engineers and other specialists. Their task is not really to design a new, ideal, search engine but to perfect the existing one. They have to compromise between the conflicting pressures of improving relevance and resisting manipulation by SEO workers. Which is actually done patch upon patch, tuned a little bit to the left and then just a tiny turn to the right.

And what is relevance after all? Perhaps like IQ it is no measure of the amorphic term "intelligence" but merely the ability to pass IQ tests successfully. So "relevance", for Google's team, is even more circular because it is measured as the ability to show up highest in Google's own search results. Moreover, they know everything possible about their ranking algorithm (and I'm sure they're sometimes surprised too), while I know at most a few percent of what I would like to know.

It is only when you think as an engineer that you find out that SEO cannot be done as an engineer.

from jennosborne 548 days ago #

Votes: 1 | Vote:

+ -

"Moreover, they know everything possible about their ranking algorithm (and I'm sure they're sometimes surprised too), while I know at most a few percent of what I would like to know."

I disagree with you on this point. I doubt very much that any one Google Engineer has that much of the secret sauce. It would be foolish for Google not to spread the knowledge out into areas of expertise. Surely there have been engineers who have left before? If any one Google engineer knew that much about the algorthm then there would be far less debate about the specifics.

I think that the reason the playing field isn't level is because they hire the smartest people in the world. You may be in their league but I certainly am not :)

from emanuelh 548 days ago #

Votes: 1 | Vote:

+ -

We don't disagree. I wrote they know everything possible about their ranking algorithm referring to the team as a whole.

And even if they do'nt really hire the smartest people in the world - it's so much easier to hide than to spy.

from Gab 548 days ago #

Votes: 0 | Vote:

+ -

I'm not sure I understand what your criticism as to being outdated refers to. Care to clarify?

from emanuelh 548 days ago #

Votes: 1 | Vote:

+ -

Sure. What is out-of-date is the SEO worker's former ability to cope with the complexity of Google'a algorithm while performing a task such as estimating reliably the amount of workhours required to lift a web page from # 11 in a search engine's SERP to # 10, through thinking as a search engineer . And I don't blame anyone - it is the algorithm's complexity (a term that refers equally to the amount of information we don't have) that has grown exponentially since 1998.

There was actually one search engine at that time (Excite?) that used to show next to each listing its relevance score (in terms of up to 100% - another outdated absurdity).

from emanuelh 548 days ago #

Votes: 0 | Vote:

+ -

- continued -

When I say that back in 1998 one could think as a search engineer I mean that a well-trained and experienced SEO worker could create a simple and adequate enough model of the search engine's algorithm, and even deduce individual relevance score differences. It was possible because the search engines were so simple. Nowadays, thinking as a search engineer, one must realize that his best model of a search engine is not adequate enough to perform even the task of estimating the relevance score gap between two adjacent web pages listed in a SERP.

I have written on this issue in many comments here in Sphinn.

from Halfdeck 548 days ago #

Votes: 1 | Vote:

+ -

I'd have to agree with emanuelh. The gap between what SEOs know and reality will continually widen. Some SEOs will insist that Google's algorithms can be reduced to things like anchor text, linkage, authority, or keywords in the TITLE. Some will say you don't need to know a thing about Google's secret sauce; SEO is dead - its now all about traditional marketing and having the right alliances to spread your word. But as there's a huge chasm between what we hear about Paris Hilton and reality, there will always be a big gap between what SEO "experts" think they know and what is actually going on in Google's datacenters every day.

More importantly, at the end of the day, its a search strategist's ability to triple a company's quarterly profits that matter. You can make that happen without knowing a thing about search. Many website owners that pull +100k uniques/day from Google know nothing about search engine optimization. Some of them even believe META keywords still matter. So what? They are raking in the traffic and they know how to convert traffic into sales. That trumps SEO knowledge any day of the week.

from Gab 547 days ago #

Votes: 0 | Vote:

+ -

Well said Half. You need to be able to make money most of all. And you mentioned Paris Hilton - bonus points right there!

Emmanuel, thanks for clarifying. I see your point, but I do think you can do a lot of experimenting to learn that stuff.

Cheers
Gab

from emanuelh 547 days ago #

Votes: 0 | Vote:

+ -

Gab, can you describe a real or hypothetical experiment followed by reliable conclusions?

from Gab 546 days ago #

Votes: 1 | Vote:

+ -

I personally am too lazy, but I can point you to some who aren't:
http://www.slightlyshadyseo.com/?p=131
http://www.allthingssem.com/seo-services-company/
http://www.allthingssem.com/category/seo-experiments/

from sza 546 days ago #

Votes: 0 | Vote:

+ -

emanuelh, in order to have reliable conclusions, one's experiments would need to advance in very small steps. (Like what SlighlyShady described in one of his articles.)

Such experimentation is very time consuming, and in the end may not amount to anything useful.

Most probably, you'll never see any reliable conclusions that are worth knowing, because the guy who invested time and effort in reaching them will keep them to himself.

All we're going to be left with is the results of non-conclusive experiments, which still make relatively good linkbait and fodder for discussion.

from emanuelh 546 days ago #

Votes: 1 | Vote:

+ -

Well Gab, an experiment modifies the relevance score of the page undergoing the experiment. But you cannot observe the relevance score - all you can see is change in ranking. And you can have no idea whether the change in the ranking was caused by your experiment or by the fact that a web page, ranked above the page undergoing the experiment, lost some relevance score.

These three posts do not even refer to relevance scores, only to rankings. I humbly advise you not to be so credulous or not to be so lazy.

from DanThies 543 days ago #

Votes: 0 | Vote:

+ -

Right - you can attempt to isolate an experiment as much as possible, run multiple experiments in parallel, etc. but I've seen many "results" reported that completely fail to account for even obvious uncontrolled variables.

from Gab 537 days ago #

Votes: 0 | Vote:

+ -

Emanuel, you're right that rankings aren't the be-all end-all. You can control for outside influence too, however, if you use made up words that are highly unlikely to be used outside of your test. aogtouabeorudspptk for example. optimize for aogtouabeorudspptk and aogtouabeorudspptking (the conjugated version), and chances are other sites' dropping relevance won't be an issue as there won't be any others..

from emanuelh 536 days ago #

Votes: 0 | Vote:

+ -

By experimenting within an arena (SERP) that contains only two pages I get no quantitative results either. And if the relevance score gap between them is zero I guess that the search engine determines randomly who is first.