Alexa Toolbar

[Lette]

I recently installed the Alexa toolbar and have had fun using it. BUT how accurate is it? I notice several small errors regarding our site so that makes me wonder. Are there any better tools out there to see a site's traffic?

[Red5]

Do you have access to your log files or have a stats package for your site?

[qwerty]

I believe the information Alexa gives out regarding the traffic of sites is based on data they gather from users of their toolbar. I suppose that if they get enough people to use the toolbar, their numbers will represent a decent cross-section of everyone on the web, but I tend to think the average user won't see any use for the thing, so they're really just counting the people who care enough about traffic to allow Alexa to track them.

[Advisor]

In my opinion, Alexa numbers mean next to nothing.

They have a very small sampling of people who actually use the toolbar, and that's where they get their data from. It's totally whacked, if you ask me! I wouldn't look at it for anything.

Jill

[french dread]

Alexa datas are shitty dont even bother with it. My personnal website is ranked highly cause it is a subdomain of my isp lol.
I wonder who use this toolbar anyway apart perhaps some webmasters

[Lette]

Thanks! You all said exactly what I thought! Yes, I do indeed have access to my log files and of course use them, but how can I get an idea about the traffic on other sites?

[french dread]

Ask webmasters :) Seriously there is no way to know, apart for some very big website who are studied by mediametrix etc

[Alan Perkins]

I've moved this thread to a more appropriate forum.

Lette, please only post topics uniquely related to Google in the Google forum. See this thread (http://www.ihelpyouservices.com/forums/showthread.php?s=&threadid=7575) for more details. Thanks. :)

[Bernard]

Alexa describes the inaccuracies of their own system (scroll down to Some Important Disclaimers):

http://pages.alexa.com/prod_serv/traffic_learn_more.html

[Papadoc]

Originally posted by Advisor
In my opinion, Alexa numbers mean next to nothing.

They have a very small sampling of people who actually use the toolbar, and that's where they get their data from. It's totally whacked, if you ask me! I wouldn't look at it for anything.

Jill

I disagree. They do have several million in download unless Amazon is lying. In any sampling, that is statistically a decent representation. If that is small, then the big boys are pathetic with their couple of thousand highly monitored users. Neither does Alexa pretend to quote stats unlike Comscore (formerly Media Metrix), Jupiter, and Neilsen. Frankly, I've used Jupiter info as well as Comscore when they were Media Metrix and their information is more in depth. But unless you are looking at the top 250 sites, their numbers are whacked.

You also have to know how to read Alexa. If you are on a site that is part of a subdomain, then you have no idea. But then this product is meant to help advertisers, and there aren't many sites worth advertising on that are less than 100K or that are part of a subdomain.

The only science that is out there is gotten by obtaining another site's log files (lotsa luck) and doing your own analysis. I've even see those get bogied too. Alexa is fairly dependable within the top 50K, marginally effective in the top 100K, and only serves to kick out the rest if they are giving you completely rediculous figures. Example: A site claiming 100K uniques per month whose Alexa number is 500K, is lying his butt off. However, if the site's Alexa is between 50-75K, it is within the realm of possibility.

Like any other tool, you have to use it to know it and to understand its limitations and where it fails vs. where it works well. Nothing works everywhere under every condition.

What Alexa does do is help you toss out the complete riff-raff and narrow your focus.

[mmmpph]

In any sampling, that is statistically a decent representation.

I completely agree. Alexa does not have the depth of data that a few high dollar companies have but they have more breadth. I've routinely tracked many of the larger ones and they lose their relevancy past the first few hundred, perhaps up to 500-600. As with anything, the smaller the sample, the less reliable the statistics that are based on that sample. Populations samples determine the standard deviation (SD). A high SD means low reliability.

There are two single advantages to Alexa. If their user numbers are even a fraction of what they say, it is still a decent sample. A small number of users means that any one user could throw the data drastically. They have taken this into account as well by making the number a 3-month average. Anyone can have a good or bad week, depending upon promotions, big money spent in Adwords, etc. But it is hard to keep that up for 3-months and not have it be real.

Second, this is data that is available to anyone. While it is quite limited, I suspect that Alexa will be coming up with a few more features before they go to a cost model.

The analysis that I have done with traffic and comparing them to the ranking seems reasonable under about 70,000. Beyond that, I wouldn't put any bet on the accuracy because of a reduced sample size. Theoretically though, this would be factored into the 3-month average to a degree.

I wouldn't sign a high dollar, no-opt-out contract based on Alexa, but neither would I do so with NetRankings. There's no need to do so. However, I would certainly rely on it for reducing my field and letting me know who the players were in a market that I was unfamiliar with.

As with anything, any comparisons should be done between apples and apples. Stats may not be as accurate between different types of sites. But if you are comparing two relatively similar consumer sites in the same market niche, you should be able to do a good comparison.

A critical Alexa factor that must be known is that the competition gets severe as the traffic goes up. Going from 50,000 to 45,000 isn't nearly as big a deal as getting from 5000 to 4000. It's not a straight line curve and accurate interpolation requires an extensive amount of information. The difference between number 1 and number 2 is probably enormous.

As pointed out, you also have to understand something about the site that you are looking at. Human intelligence and interpretation are required. Yahoo with its shopping, Groups, Mail, and virtually unlimited subdomains and personal pages will continue to rank #1 but that is not an indicator of the popularity of Yahoo's search function. A site that has a highly used BBS will also rank higher. Still, 99%+ of the top 50,000 are reasonably comparable if all factors are weighed. Stats are great if they are in the hands of someone that understands the method by which they were determined and what limitations they have. Stats can be dangerous if taken on pure face value with no understanding of what they mean, how to interpret them, or various weight factors that must be applied.