Alexa, Amazon's Most Flawed Idea

Rub3X writes "The Alexa ranking system is naturally flawed. The data should never be treated as accurate, as it's easily manipulated, and not supported for most browsers in the world. It's an estimate, and nothing more. " I've been saying that forever, but unfortunately for me, since it's a number on a website that is considered "Real" to some, I'm supposed to take it seriously. I imagine this is a problem for many webmasters out there.

Comments

Error in article

[tont0r]

According to the article:
"Alexa has no support for FireFox, Opera or Safari at all. "

According to Alexa's Wiki:
"Users running any browser except Internet Explorer and Mozilla Firefox are not represented. Thus users of Opera, Safari, mobile phone (WAP) browsers are all ignored. Nevertheless, this is still the vast majority of the browser market."

So its half right :P

But is supported for the #1 browser

[Anonymous Coward]

Alexa may not be supported by most browsers, but it is supported by The One Browser that most people use and almost everyone has.

Re:But is supported for the #1 browser

[daeg]

It doesn't matter, though, since the distribution of toolbars is not uniform across all Internet users. A good example is the website I work on. We know our traffic, yet Alexa under-reports us. We also know a local competitor's traffic -- both sets of numbers are generally public information that advertisers use. They have a nice site but get about 1/2 of our traffic, yet Alexa over-reports them over us by a factor of 3-4.

You can pull accurate statistics if and only if your data points are distributed correctly. Because Alexa has no way to randomly and accurately assign toolbars to users, their data is not reliable in any form.

A similar example is how political polls are taken. You can get accurate numbers with 1,000 adults if, and only if, those 1,000 are random throughout the entire population. You can skew the poll numbers by polling 1,000 Democrats or Republicans only instead of 1,000 random. Your results are only accurate to your surveyed population -- in Alexa's case, their numbers are only accurate so far as "Rank ### amongst Internet Explorer 6.0 users who speak a limited number of languages who have voluntarily installed our toolbar to submit their surfing habits to us for analysis and are subjected to trade secret methods of ranking".

The only way that you could pull accurate numbers would be through all ISPs selecting random data points to find what hostnames people were using. It would have to be filtered, though, to produce accurate numbers in terms of actual "website hits" instead of just "website requests". Keep-alive would further impede accurate results. As would proxies, DNS caches, and HOSTS files.

The people that matter

[truthsearch]

Everyone who owns or develops web sites knows this. Anyone who hints in a forum the numbers may be accurate immediately gets slapped down. It's the non-technical advertisers who don't know this. And they're the only ones who care about this ranking in order to gauge how much to spend on purchasing web site advertising. Since almost no web sites publicly display traffic info advertisers find Alexa rankings very convenient and probably just don't understand why they'd be useless.

Until advertisers "get it" or a much more accurate public metric is made available, Alexa rankings will unfortunately matter to web sites that are supported by advertising.

Yeah and MOST for slashdot is not IE

[technoextreme]

The problem is that statisically it's nice to say that 30% does not make a majority but Im sure that spreads changes from website to website. Imagine looking at the statistics for a Linux website. The majority there better not be IE.

Another reason to dislike Alexa

[jandrese]

Alexa also rewards webmasters who write badly broken IE only webpages, forcing people who normally use Firefox to switch over to IE for that webpage.

Re:The data shows there are problems

[Anonymous Coward]

I wonder would the obvious spike in users at digg & /. be due to the introuction of an alexa plugin for mozilla firefox at that time (May 2006)?

www.stevecastle.org

Just askin'...

Useless or Used Wrong?

[logicnazi]

Now it's clear that the rankings from this system are heavily skewed and misses a substantial portion of the user base.

This suggests it is useless as a way to estimate how much to pay for advertising on a web site (though since this is usually per click/per display I don't see why ranking matters here). However, it doesn't show that this data can't be usefull for other things. For instance it could be quite usefull to know what other sites the users (or IE users) of a site visit.

In other words the data seems useless for any statistical analysis but it could be quite helpful to know what sorts of users visit a site. Sure slashdot's traffic might be underrepresented but I bet you the data still show that slashdot users are quite likely to go browse gadget purchase sites or programming related sites. If you want to know where to advertise your new fancy gadget or a fancy new programming enviornment that would be very usefull information even if it wouldn't support a rigorous statistical analysis.

BZZT.

[Rob T Firefly]

One fact TFA and the Slashdot title both got wrong, is Alexa wasn't Amazon's idea. Until Amazon bought it in 1999, Alexa was the commercial offshoot of archive.org for three years. Alexa is still what gives the Wayback Machine its web crawls.

don't see the point

[jasen666]

Alexa is flawed from the start.
What impetus or benefit would a user have to install a toolbar that tracks them? Other than out of charity to help out this company? I don't get it. Nor do I particularly trust them. Just one more thing to help crash IE.

WTF is Alexa?

[Yvan256]

Seriously, why bother writing two or three sentences anymore? Just put a single link on a single word, that's even less helpful and even less work for the editors.

WTF IS ALEXA?

Another case of "I don't want to waste 30 seconds to explain WTF the news is about, let 50K users waste a few minutes and slashdot a website trying to figure out what it is".

Re:The data shows there are problems

[stevesliva]

All of the variance was Alexa anomalies.

In the past three years, the one Slashdot article with Alexa in the title was in December 2005. No doubt a few slashdotters took a quick look at the toolbar, and the quickly decided it was worthless.

Re:The data shows there are problems

[badasscat]

f so, it kind of makes the case that Alexa data is less than useful.

It's not "less than useful".

In fact, this is both a completely obvious and a completely stupid article submission. The "duh" tag is appropriate, both because none of the current ranking/statistics systems are accurate, and because despite that, they are still useful.

When you're looking at numbers like total reach, or you're comparing one web site with another, nobody needs statistics that are 100% accurate. I don't need to know if CNN has 4 million unique visitors per day or 4,409,765 unique visitors per day. You're using these services to get a general idea. If I'm running a web site, for example, I know what my own stats are - I don't need Alexa to tell me. But I can still use Alexa to tell me the basic gist of a competitor, and if they're not as accurate as internal stats would be, what does that matter?

Moreover, Alexa's stats are no more or less accurate (or easy to manipulate) than those of major organizations like Nielsen. The fact of the matter is any system that's not using actual server logs is going to have some inaccuracies (and if you think otherwise, then you've just bought into marketing spin). You live with it and accept it. The main difference is that Alexa is free, whereas other stat compilers charge thousands of dollars per year.

Re:masked domains

[Rob T Firefly]

Unless the toolbar goes off of what you type into the address bar rater than the urls that are actually loaded, would I would consider more likely. Do they say how they do it?

I don't know, but off the top of my head I doubt page-ranking services would count other sites loaded in an IFrame. Otherwise I could create one of those useless domain-squatting pages that just exist to throw ads at people who click or type wrongly, load a bunch of actually useful, respected sites in IFrames, and use all that content to boost my Alexa rank/Google rank whatever else.

Re:The data shows there are problems

[dumbfounder]

I think they changed to a different statistical model at that point, there are a ton of sites that make a jump on that same date. It is a good thing that they continually refine what they have (because it is FAR FAR FAR FAR from perfect) but they should have a little asterisk there letting people know what happened.