Meryl.net Home About Blog Clients Services Contact

Meryl's Notes Blog

Things wordy, geeky, and webby

Alexa Web Search – Beta

Friday, May 10th, 2002 at 7:53 AM Category: Meryl's Notes Blog 4 comments

Entry originally posted on May 5. Reposted today because of the lively debate and Webreference adds to the fray.

Another ego-surfing tool: Alexa Web Search has a cool feature where you can find out the ranking for a site in all the sites that exist on the net. Just enter the site name into the searchbox and it’ll give you the results.

<edit>Nick took a closer look at Alexa and provided honest feedback:

“It ranks site based on traffic. What they fail to mention is that their databases are terribly outdated (by my test it appears to be 1 year and 1 month out of date). In addition, the ranked numbers are based on a site-wide rank and not per page rank, we are not sure if they use the highest-ranking page on the site or the lowest for the base. Let’s hope these bugs get fixed before it goes beyond beta. As always, we prefer to measure a site’s creditability by the number of readers as opposed to the number of hits, after all, any banner view in a pop-up can count as a hit.”

He’s absolutely right about the hits. I could reload a page 10 times and give it 10 hits. Now, don’t go running off and trying that to your own sites. Do you really want to up the hit numbers like that? OK, it’s on your head.</edit>

<edit on 5/10/02>Amazon’s new service combines the power of Google, the rankings of Alexa, the Wayback Machine, and Amazon’s back-end. Will Amazon
rank the Web like they do for books? See what Andy of Webreference has to say.</edit>

Tags: Leftovers

Subscribe to this here blog: RSS or E-mail

4 comments

Posted by Geoff on May 8th, 2002, 12:22 PM
Let’s be careful not to spread false unsubstantiated rumors. If you read their help, you’ll notice a few things… 1) they update their rankings daily, and 2) their ranking information rules out multiple views of the same page by the same user, and 3) their rankings are based not only on hits, but also on number of users.
Posted by Nick on May 8th, 2002, 2:27 PM
If the content were updated on a daily basis they would have information about the current feature. Instead they are reporting information about a feature that was published in 1999. Furthermore the information in the database itself for 1999 is also completely incorrect at best. No research was conducted to evaluate the information to ensure it was accurate. The user sampling is also very small and limited given the methodology for obtaining the information. At best, the figures on Alexa can serve as a ballpark generalization of traffic, but it can by no means function as a metric in which to determine popularity among “users.” It would be better advised to use data from resources such as Fosters and Jupiter for more accurate metrics.
Posted by Geoff on May 8th, 2002, 4:30 PM
I can’t speak about all of their data, but at the risk of repeating myself, the traffic rankings are updated daily. Fosters and Jupiter might have more accurate metrics but their methods also rely on sampling… and a small sample at that. I checked with Alexa and they have had over 10 million toolbar downloads… certainly not a small or limited sample size, and well over 100 times larger than the alternative ratings companies.

Anyway, skepticism is good. Second opinions are great. Keep your eyes open.
Posted by Nick on May 9th, 2002, 10:47 AM
Here is my point. I don’t have Alexa (but I used to have it installed in the past on a different computer)… no one in my group here at work (20+ developers and managers) have it… in fact I would go so far as to say no one in this company (20,000+ employees) have it installed. What kind of sampling is that? Instead of posting nothing but skepticism I would like to propose a solution: Alexa should post the facts… and exact count of total systems with the application installed in addition to an estimated ratio of number of users using it vs. those who are not based on previous surveys that have estimated the usage of the Web. If it’s a 1 to 100,000 ratio I’d be impressed… but not satisfied until I saw something like a 1 to 500 ratio… anything short of that is as valuable as me wagering stock on the outcome from an 8-ball.