Amazon’s Alexa traffic reporting service has little credibility left among people who follow traffic trends. Most analytics services, like Comscore, don’t measure small sites well, but they tend to get it right for the larger sites. Alexa seems to get everything wrong, no matter how large or small the site.
Example: In August Alexa said that YouTube passed Google itself in total page views. They were wrong, but their data continues to perpetuate this alternate reality.
Now, another embarrassing error. Alexa says that Facebook, on a steady growth curve for the last two years, now has a larger audience than MySpace. This isn’t as ridiculous as the YouTube/Google error, but it’s still way off. Comscore says that worldwide MySpace uniques are 109 million/month, whereas Facebook is at 86 million. Compete.com, which measures traffic using similar techniques as Alexa, stills says that MySpace is larger than Facebook.
Thanks for the tip Mark.
See all
Post
They don’t even track all websites. Quantcast is much better.
http://www.quantcast.com ENJOY!
Interesting, but not too surprising. Facebook users are more likely to be interested in things like web traffic, so they’re more likely to have have Alexa’s toolbar installed. Of the thousands of people building/marketing facebook apps, a good % probably have it installed. 1000s of users frequenting your site with the TB installed will spike the heck out of your Alexa score, even at this level.
That’s the problem with Alexa data, it’s not anywhere near a random sample. The more internet-marketing & tech-saavy the audience, the more inflated a site’s traffic will appear. Some engineeer from Google did a study, good read:
http://norvig.com/logs-alexa.html
One problem with Alexa (internally) is that while we all know their data depends on people with the toolbar installed, they don’t seem to be doing anything to get more people to install it. As such, their stats become more and more irrelevant overtime…
http://www.compete.com, and quantcast from above, http://www.ranking.com are other services.
Alexa also has a redirect problem, where people put redirect to alexa then on to the website to boast scores. Gaming the system
yeah, i think the problems with Alexa are well know. What I don’t get is why they don’t make any effort to fix those problems.
Anyone in the SEO business has long known that a diaper wearing monkey throwing darts at random numbers is more accurate than Alexa. If I hear someone bragging about their Alexa rating, to me and many others, their credibility goes right through the floor.
Alexa may be far from perfect, but consider that comScore gets its figures through downloads of an “internet accelerator” that some places have referred to as spyware. Before the investors jump in defending comScore (Fred Wilson is on the board) they naturally deny this, instead referring to their program as “researchware” but at least consider the distribution model: a program that is marketed as a download accelerator to PC users. Do you really think that PC users who download an internet accelerator make for a sample audience that is that much better? If anything comScore stats weigh against heavy tech and sites that may find favor amongst first adopters…and most certainly against sites favored by Mac users. comScore gets it right only as far as PC users who download internet accelerators get it right, and nothing more. The overall usage figures are then applied on estimates of sample size applied to overall market…which makes it, like any stats program, an educated guess at best. If comScore says a site has 16m users it may only have 100,000 comScore users in the particular market then applied statistically to how many they think it would have based on this…but again, the sample is narrow and not necessarily representative of the broader market.
Alexa has been an issue since way before Web 2.0 explosion. I’ve been dealing with the crap for years - probably since 2000ish. I always enjoyed when someone would try to get me to sign millions of dollars to their service based on Alexa charts. Those days I enjoyed
Duncan, very informative. Thanks for sharing…
btw- isn’t Fred Wilson a VC; not a technologist, CTO or an authority in the space?
Sincerely,
NY Giants Stink
Thanks for the credit and link Michael!
Compete aggregates data from 10+ sources, including ISP logs, ASP’s, toolbars, Panels, etc to form its 2 million person sample.
Check out this interview with Guy Kawasaki about Compete for a whole lot more - http://blog.guykawasaki.com/20.....ons-2.html
I can confirm in more than 10 cases for my stats that the stats of Alexa is wrong…
NY Giants Stink
thx.
Fred Wilson (VC) is an investor in comScore and also sits on the board, least he did about 12 months ago unless he has stepped down in the mean time
It’s ok for most people I guess. I used Compete also. How accurate is Alexa? No idea, but when you dealing with millions of sites who knows.
Of course it’s wrong. Like a broken clock, it’s correct twice a day, but that’s about it.
My site is clearly bigger than myspace.
Comparing Alexa stats (international) with Compete (US-only) without even mentioning that discrepancy seems almost as silly as using Alexa for important decisions.
*walks away from the dead horse*
The main problem with Alexa is that they deal with irrelevant stats, like overall rank and reach, which say very little.
And of course the Alexa system is too easy to game to get into the top sites list with very little real traffic.
Compete and others are more useful due to showing the approximated no of uniques, even if they way underestimate them, but you learn quickly how much they underestimate.
Not to mention that it shows MySpace traffic as unrealistically constant since May
Did anyone hear of Syntryx? Alexa’s method is simply not capable of dealing with the large mass of websites out there. Compete is definitely better, especially now with their new Search Analytics. Quantcast is a different type of a bird. I don’t think trending is their true focus nor vision. They got something else in mind. There’s are few more alternatives out there, however I think that Syntryx unique technological approach to competitive analysis enables comparison of both Long, Mid, and Front tail websites is more complete.
I don’t blame Alexa for not being able to properly assess trends for the mid and long tail domains; however it’s disappointing that they are incapable of providing proper trends for the front tail domains where their methodology should have been sufficient.
Compete isn’t any more accurate than Alexa, it all really depends on how many people you have using a compete toolbar.
Quantcast is about as accurate as most javascript tracking scripts, but there is a major problem, most websites don’t include their code.
If you went by Compete data, I should be charging the same for advertising as Techcrunch were in June this year….
http://siteanalytics.compete.c.....?metric=uv
I don’t know why Google hasn’t came up with a way to measure Internet traffic, I think they would have a better idea. If only they could partner up with Yahoo, and MS Live Search.
I think all of these guys were in make believe statistics class. Compete, alexa, hell even nielsen and arbitron, every ratings company under the sun use such horrible methods that I’m surprised any ad campaign has ever succeeded.
The counting wars were over a long time ago. All counting has a frame of reference. If you work with the frame then insight may follow.
Look for consistency in data collection methodology whatever it is. Strongest comparisons will be trends for like with like by collection sample, site audience, site structure, site use and so on for any metric.
Ranked position is always whatever for any counting methodology - very subject to the what, how, why of the methodology. The tighter the online methodology to get ranked position right - the less useful the data seems to become.
Alexa is very useful for it’s global coverage and length of service. Be nice to Alexa.
Alexa is right on the facebook thing, more people are using it since they found out that I invested in it!
http://fakesteveballmer.blogspot.com
http://www.alexa.com/data/deta.....ftware.com
Alexa has our 1wk at 195,917 and our US rank at 60,260.
That’s pretty good for a website that only does outsourcing to a handful of clients.
BTW, if you haven’t heard, BeerCo is coming to town next year.
I figure a search released anywhere else will never get used or seen. Getting beat down by 75% tax didn’t help either. See you all next year.
My website went from 298,987 to 1034 in one day and then back to 490,871 the next.
It is worthless. The only people who look at it are out-of-touch CEO’s.
i DO NOT tamper with my alexas ranking for one reason only.
I use alexa for myself, i want to know how good my site is doing based on their users. i could care less what people think of my alexas ranking and this is why i dont try to bump it up (and i can easily)
Alexa although not accurate it is the most accurate of what is available today.
A lot of very good points here - Alexa is easy to game, and biased towards one set of data. As is Comscore, Compete, Hitwise, etc. But this kind of data is absolutely necessary, and nobody seems to have any idea for making it more accurate.
To make it about 98% accurate, the browsers would have to build in the code across all operating systems - with screams and howls from the data protection lobby.
Hence, for the foreseeable future, you have to look at all of these sites, and use a bit of common sense mixed with a fistful of salt.
Alexa is 100% right - every time. Spot on. I set my watch to it every day - twice each day.
I have worked in the research group of a large publisher for over 8 years. I have been tasked with trying to find out what tracking service is the best and have been working on this for almost 3 months now, without help from any of the services. We want a true independent analysis.
I have found that you get what you pay for. Compete is terrible, Alexa is worse, Quantcast is bad (though people seem to like it), and the services you pay for are the best. Of course the smaller the site the harder it is to measure and I understand this being a researcher. I just think people forget that free data is usually crap.
Alexa, Compete, Comscore, they are all way off. The problem with Alexa is that it is so easy to game, so that just one person can really affect a site. As to why they don’t do anything about it- I’m guessing they just can’t come up with anything.
Alexa is for trend watching, not statistics. And the best tool for trend watching there is.
I develop a site that has grown steadily since I came on board. This growth has been reflected in Alexa at an uncanny accuracy.
It is not fair that Techcruch continues to monitor Alexa accuracy by focusing on the behemoths of the web. I don’t think we need to use Alexa to gauge these trends. There is a need for something that does it well.
@21 (Andy)
Don’t forget us little people who knew you when…
You know how the old saying goes?
Don’t trust a statistic that you haven’t faked yourself.
Alexa, Compete, Comscore, it’s all the same garbage.
They all seem to be using the same flawed tools, relying heavily
on spyware toolbars and “magic” extrapolation. Keep in mind
that only the most clueless (”bottom of the barrel” users) install
those toolbars. So if your site doesn’t attract a reasonable
portion of mouthbreathers - too bad, you’re off their radar.
Complaining about alexa is like complaining that there is only
crap on tv.
None of them are accurate. if you look at stats from Hitwise and comScore for the same website, they have very different numbers.
Also, many of the services measure things in different ways (daily unique, monthly unique, visits, visitors, pageviews, US-only etc)
For a comparison of different competitive analystics tools, try this post: http://lifeisaventure.wordpres.....he-biases/ and http://lifeisaventure.wordpres.....fic-stats/
Its important to be in business than getting business affectively right.
http://tekno-world.blogspot.com
I don’t think alexa data is accurate, if Comscore is using similar techniques obviously its stats is going to be inaccurate too.
Very clever mike:) y dont u and duncan write a book..”Hundred ways and more to highlight Facebook”. No publicity is a bad publicity.Especially for a web 2.0 company.
Yeah, Alexa is mixed up. I dunno why so many people rely on it. And TextAdsLinks are using it to calculate the value of a link — they should consider using something else
Alexa is the biggest load of crap since since George bush said Iraq had wmd’s
I have to agree with RAJ Michael and Duncan have a little soft spot for Facebook.
But then again so do I
Alexa is like dessert to lunch. Nice, but not necessary.
I’m sorry, compete is better than alexa? Whaaa?
I just tried to compare 2 sites of mine that get exactly the same traffic (about a million uniques a month each). Compete told me one site gets 5 times more traffic than the other.
It is absolutely no better than alexa, not even close.
Personally I like Hitwise’s model - they have proxies which sit at ISPs and watch traffic flows. It doesn’t get much better than that unless you’re Narus or the NSA
(I don’t work for any of them).
Comscore is not the holy grail of user statistics.. I myself prefer google trends which does indeed show facebook is very close to catching myspace..
http://www.google.com/trends?q.....amp;sort=0
wow… i can’t believe i agree with Chris Sack!!!
j/k chris. it’s not the first time.
sorry for the double post:
although…. http://www.google.com/trends?q.....amp;sort=0 hehehe.
OK, so the general consensus here is that *none* of the top measurement firms (Comscore, Nielsen/NetRatings, Compete.com, Quantcast) produce data that can be considered consistently accurate. Do you mean to tell me that Wall St has been making huge investment decisions on companies like Google, Yahoo, EBay and others based on completely useless data?
Comscore, Compete.com and Nielsen regularly put out search engine marketshare data, and those announcements are picked up by Wall St and send search engine stocks moving. How is it possible that they could be wrong when looking at such huge sites?
How can the myspace usage be flat for so long? To me it implies no data for that period