email: ebobnar at gmail.com

Affordable SEO Consulting for the Results-Oriented Online Business.

Welcome to my blog. Here's a list of my best posts, as well as a complete archive of everything I've ever blogged about.

Main | The Pros and Cons of Autolinks »

February 18, 2005

The Search Engine's Point of View (SEPOV)

Randfish over at SEOmoz Blog has a great post on how to approach the site optimization process from the perspective of a search engineer.

This is something I've been thinking about a lot lately. No matter how many Information Retrieval papers we read, or how much observation and testing we do, or how many sites we optimize, we never truly know what factors a search engine is taking into account when they rank pages. And even when we figure out what works, it still could (and often does) all change in the blink of an eye when a search engine decides to update its algorithm. Then it's back to the drawing board, testing and crunching numbers until you think you got that perfect formula worked out again.

It seems like an easier tack would be to imagine yourself as an engineer working at Google, and reflecting on which decisions you would make in designing a search engine that produced the most relevant results while minimizing the impact of spam.

Taking this perspective, it would make sense to discount reciprocal links entirely, along with "run-of-the-site" links, links from off-topic pages, and links from directories. These types of links are simply too easy to manipulate. This leaves you with two other types of links:

1) Links from authoritative, on-topic pages, i.e. "quality" links.

2) Links from non-reciprocal, non-authoritative, on-topic pages, i.e. your "quantity" links.

To rank well, you either need a few big, authoritative links, or lots of little links. In both cases, all the links would need to be on-topic, and none of them should be reciprocal.

It seems that a search engine could base their entire ranking algorithm around these two types of links and be almost immune to spam. I suppose there would still be the problem of people paying for links, but this could be minimized somewhat by only counting those links that were strictly on-topic. Finding links to buy which match up with the content on your pages can be challenging, especially if you're targeting a niche market.

Most on page factors could be discounted almost entirely. As long as a keyword is mentioned somewhere on your page or in the anchor text pointing to you, you should be ranked for that page. Keyword density should be almost meaningless, as it tends to be nowadays, anyway.

Randfish lays out some other, more interesting ideas, such as tagging known SEOs by IP address, tracking the sites they visit frequently, and discounting links coming from those sites.

More on this later. In the meantime, engineers at Google and Yahoo are thinking about this kind of stuff everyday. We should be, too.

Like this article? Get lots more by subscribing to my feed.