• Register
• Login

SEO Theory - SEO Theory and Analysis Blog

Algorithm analysis, Web community relationship analysis, SEO practices and techniques, industry news, etc.

• Home
• About SEO Theory
• SEO Glossary

AlexaRank, Compete, Google Toolbar PR, and other SEO quackery

Posted by Michael Martinez on July 31, 2007 in Competitive Analysis

First impressions are hard to recover from. They are also not very trustworthy.

I usually form an opinion of someone’s SEO capability within 30 seconds of reading the subject line of their forum posts. If their opening sentence or paragraph mentions Alexa, Google Toolbar PR, or some similar popular “ranking” value, I usually write them off as having no substantial knowledge fo search engine optimizaton.

Now, there are quite a few “A list” and “B list” SEO bloggers and forum operators out there who still bring up Alexa rankings, Google Toolbar PR, and a few other SEO buzz expressions. I give none of them any credibility for understanding what to do with numbers. I put no stock in their analyses.

I have the impression most of the well-known people in search engine optimization avoid saying anything favorable about Alexa and the Google Toolbar PR. They have learned to distrust these numbers.

Even as recently as a few months ago I have seen discussons where well-known people have talked about their rankings on Alexa. High Alexa rankings have been shoved into people’s faces as some sort of vindication of competence and skill for years. Those numbers really mean nothing. Alexa’s data is a fringe snapshot of Web traffic and it will be more accurate for some sites and less accurate for others.

Unless Alexa can offer a consistency of accuracy, however, its rankings are meaningless babble with no relevance whatsoever to any useful comparison of site visits and visitors. If you consider (for the sake of discussion) that an Alexa ranking of 50,000 is at best 50% accurate and an Alexa ranking of 256,000 is at best 10% accurate, what sort conclusion can you draw about the quality of either Web site? None, really.

The problem with Alexa rankings is that we have no way of knowing how far off their estimates are from reality. Every Web site that Alexa reports on potentially receives more or less traffic than Alexa suggests it may be receiving. There no signals that tell you whether the Alexa numbers are close, too high, or too low.

And the same is true for Compete’s monthly traffic estimates. Both services probably have a better statistical sampling for large, well-known domains like CNN, Time, Whitehouse.gov, Yale.Edu, and Match.com. But when you get down into niche fields, less generally useful or appealing verticals, Compete’s data is no guesswork is no longer trustworthy.

Which is not to say that neither Compete nor Alexa can be used for any sort of analysis. You can certainly look at the trends in growth and decline of estimated site visits. Do those trends match other data you can track? If so, then you may be able to gain some insight into who visits your domain. For example, if you get 30,000 visitors a month but Alexa estimates only 1600 visitors and Compete estimates only 3,000 visitors doesn’t that suggest to you there may be different populations that have greater impacts on either Alexa or Compete?

Analyzing other people’s guesswork means your own analysis can never be more informed than guesswork, and it’s not fair to say that you’re making educated guesses. You’re making guesses based on guesses and ignorance, just as the guesses made by Alexa and Compete are based on ignorance.

If you can isolate trends for small groups of Web sites I believe you can get some useful guesswork out of Alexa and Compete. You can also check your visitor counts against their estimates to see if you’re improving your footprint in their user populations (which are drawn from two primary sources: users of their tool bars and people who subscribe to the ISPs from whom they purchase click data).

But as far as comparing Alexa ranks and Compete traffic with other sites goes, you might as well hang a shingle on your door that says you’re a qualified dinosaur doctor. You have as much credibility that way.

The Google Toolbar PR gets ripped on at least a weekly basis in the SEO blogosphere, but people continue to talk about their sites’ PR when they participate in forum and blog discussions. The very fact that someone says their site has a PR value of anything tells you immediately they don’t know what they are talking about. And if you point out the obvious to them, they may say, “Well, yeah, I know it’s about pages but I mean my home page has a PR of X”.

Toolbar PR is not based solely on links, as many SEOs claim. Hence, it’s not a “logarithmic derivation of internal PR” as many SEOs also claim. And even if it were a logarithmic derivation, so what? When you classify 25-30 billion pages on an 11 point scale, what does it matter who gets a PR of 8 or a PR of 9? Those values don’t affect your search results.

Nonetheless, every “cool SEO tool” that comes out — even after years of leading SEOs saying the Google Toolbar PR values don’t mean anything — reports Toolbar PR. Why do they do that? There is no statistically meaningful way of integrating Toolbar PR values into an assessment of optimization, visibility, or link strength. I roll my eyes every time I see a Web browser reporting Toolbar PR.

I think the reason so many people fall in love with these measurements is that people are more comfortable working with numbers than with vague concepts. Abstract ideas are often forced to conform to concrete limits. I remember many of my mathematics professors rolling their eyes at students because entire classes wanted to know which numbers they could plug into variables to prove equations were correct or incorrect.

The need for plugging in numbers holds you back when you’re analyzing a complex system. If we assume for the sake of discussion that Alexa, Compete, and Google agree that your site probably gets a lot of links and a lot of visitors, what does that tell you about how many queries for which you’ll rank first?

Absolutely nothing.

What does it tell you about how well your site will perform against someone else’s site?

Absolutely nothing.

What does it tell you about what value your own links will pass to other sites?

Absolutely nothing.

When you know nothing about two Web sites, looking at guesswork from other perspectives that work with the same amount of ignorance doesn’t provide you with any insight. Now, some people would be quick to point out that Alexa and Compete buy ISP click data, which is true. But how many clicks does site A receive versus site B? They don’t provide you with that data. Their numbers don’t imply what that data is or may be.

Just as you cannot determine what your internal Google PageRank is from the Google Toolbar PR value, you cannot determine how much click-through data your ISP (or anyone’s ISP) sold to either Alexa or Compete.

We can fairly reliably discover two things about Web sites: Where they rank for specific keywords and whether most people should be satisfied with what they find on those sites. We can also make some working guesses about how much traffic keywords get and about how that traffic breaks down.

We won’t be 100% correct but we can build reasonably functional data models. And those data models can tell us more than looking at arbitrary rankings in tool bars and aggregator sites.