A VC« comScore Responds To The IAB | Main | TargetSpot »
What's Wrong With Alexa?
Why have many "web 2.0" service seen a significant dropoff in traffic in alexa in the past six months?
I used to find Alexa valuable for at least relative numbers. But given what's been going on recently, I am not sure it can be trusted for anything.
Check out this chart for delicious and digg:
These two charts are almost identical in their shape. How can that be?
Here's comScore's data on digg and delicious:
April 24, 2007 Venture Capital and Technology | Comments (23)
Comments
It seems like Alexa has been making changes in the way it measures site traffic, but even that doesn't explain the parallel. (Your Alexa image didn't load for me, so I went to the source - http://tinyurl.com/2cc6ln .) It's interesting to see that Alexa regards del.icio.us and delicious.com as different sites.
Posted by: Aaron | Apr 24, 2007 9:08:32 AM
first, i dont see your alexa chart. second, i went to alexa and did it myself and dont see the close correlation/similarity that you claim (it shows delicious growing and digg flat to down). maybe i'm missing your pt or miscalculating. pls clarify.
Posted by: mmay | Apr 24, 2007 9:36:02 AM
pls ignore, i used the same URL that you had used in the comscore screen above ("delicious.com"). i assume comscore does not consider the two URL to be different like alexa does??
Posted by: mmay | Apr 24, 2007 9:38:40 AM
Fred,
I think it has something to do with the alexa toolbar and usage overseas. Everything is measured relative to the total Alexa population. However, I think the toolbar population is rapidly skewing towards Asia. You can see it it many of the popular sites in the US.
Posted by: Brad | Apr 24, 2007 9:42:18 AM
The same pattern appears when looking at a lot of European sites, since some time in the second half of 2006. Strange that Alexa isn't talking about this, as it's been thoroughly trashing the value of their traffic data and landing some people in hot water.
Posted by: NF | Apr 24, 2007 10:07:54 AM
I can't stand Alexa! It's too confusing
Posted by: Jason P | Apr 24, 2007 10:36:18 AM
Those two charts actually could both be accurate if more people are visiting those sites, but coming back less often (since Alexa tracks daily uniques and Comscore tracks monthly). That would make sense if, for example, more of their traffic was starting to come from search engines (and less from "active users").
Posted by: Timothy Bryce | Apr 24, 2007 11:07:41 AM
I saw the same thing for blog patterns as far back as 6 months ago. I attributed this to the cyclical patterns caused by days of the week and holidays, but this is clearly strange.
It has to be that sample population is just way too different for these two services. Could it be that most of Digg crowd is Firefox? I'd expect del.icio.us to be same way, but who knows...
Alex
Posted by: Alex Iskold | Apr 24, 2007 11:10:04 AM
Three factors:
1) Expension of Alexa toolbar outside US. US share of Alexa base dropped from 37% in 2004 to only 14% in 2007. Alexa Reach is a relative number (share of the global pool)
2) Alexa reach represents daily uniques whereas Comscore etc tracks monthly uniques. As others have pointed out, Alexa number depends how often a user visits a website.
I wrote some analysis of Alexa a while ago ...
http://lifeisaventure.wordpress.com/2007/03/27/how-to-use-alexas-traffic-stats/
Posted by: frank | Apr 24, 2007 11:36:30 AM
Bottom line, DO NOT trust Alexa data.
There are many people who believe this, here's my rant: Bad Business Decisions with Alexa
Posted by: You Mon Tsang | Apr 24, 2007 1:04:31 PM
Fred:
Here is Yeeyan's Chinese translation of this post:
http://en.yeeyan.com/articles/view/thunder/806
I am a co-founder and translator of Yeeyan.
I've appended a chart at the end of the translation, comparing three top Chinese portal sites: QQ, Sina and Sohu. And the chart echoes your observation.
We will further translate comments to this original post and comments to the translation, into their opposite language.
I am curious to see by facilitating this kind of cross-language discussion, whether we can solve this puzzle in an interesting way and possibly quicker, especially when Asia/China usage of Alexa toolbar is believed to be a major factor in the equation.
BTW, we now have a dedicated AVC group at Yeeyan, where readers/translators can focus their interest on AVC, and they can subscribe and have discussion.
http://en.yeeyan.com/groups/show/avc
Posted by: Thunder | Apr 24, 2007 2:02:46 PM
Brad, NF, Timothy Bryce's comments have been translated into Chinese. Links posted by frank and You Mon Tsang have been conveyed to Chinese readers as well...
Posted by: Thunder | Apr 24, 2007 4:46:45 PM
This is very interesting!
Posted by: Doug Dosberg | Apr 24, 2007 7:39:43 PM
I agree that Alexa is good for little more than trends - and your posts reveals that even that can be flawed.
For your own data, your logs remain the best measure of traffic and data. Google Analytics is great for high-level information.
For other sites - I've found that Quantcast.com is the closest of the available, public third-parties. Compete.com is next, but my main issue with Compete is that it is monthly rather than daily or weekly. This is particularly problematic when, like today, its the end of a month and the most recent data is nearly 30 days old.
Alexa is based on toolbar adoption and user-behavior which, by definition, has selection bias.
That said, as long as people continue to measure traffic and brand awareness via Alexa, it will be powerful and it's up to the view / audience to decipher the real meaning.
Ryan
beRecruited.com
Posted by: Ryan | Apr 24, 2007 11:07:59 PM
Alexa toolbar was blocked as adware by many anti virus software in China.
Alexa toolbar lost about 1/3 of installation since last year.
Posted by: Che Dong | Apr 25, 2007 1:25:07 AM
This comment was translated from Chinese, orignally made By Si Yu(思域):
I compared top 4 Chinese portals: QQ, Sina, Sohu and 163, the shapes of their reach and the values of their p/v are very similar. Traffic ranks are quite different because of the rank is proportional to Reach^2 * P/V.
Similarity of the chart might be because of the following:
1. All top 4 portals have similar user base, therefore there are roughly same level of Alexa toolbar users visiting all top 4 portals.
2. All top 4 portals have similar content, which is news focused. Therefore they have similar P/V.
3. All top 4 portals have positioned themselves similarly in terms of targeted audience, marketing and contents.
4. Alexa is screwed.
Posted by: Thunder | Apr 25, 2007 2:45:11 AM
There are other interesting facts about Chines e Internet usage from Si Yu's another comment to the Chinese translation:
- Most Internet companies have some semi-legal contents, such as gambling and porn
- These contents and therefore the usage were greatly impacted by government regulation
- Dip in Januray and March on Alexa were identified as the Department of Police's tighening control on Internet porn. But once loosened, it sprung back quickly.
- Impact of news events and cycles like weekend usage is very obvious.
Posted by: Thunder | Apr 25, 2007 3:15:34 AM
This comment was translated from Chinese, orignally made By Thuner:
@Si Yu: The reasoning was great. However it did not explain the similarity between Digg and delicious. What's even worse is that Flickr's pattern is identical too! Fortunately, facebook sees a totally different chart. What is common among Digg, delicious, and Flickr, but different in facebook?
Posted by: Thunder | Apr 25, 2007 3:20:27 AM
Since Alexa is IE only and Firefox market share is continuously growing, being up to 30-40% in Europe in many countries, Alexa is losing users fast. That also shows on the graphs. Alexa is very unreliable for anything outside the US.
Posted by: Jüri Kaljundi | Apr 25, 2007 3:54:33 AM
Jüri, there's a Firefox extension that sends data to Alexa, just like the IE toolbar does. Of course the FF extension does more than that, but even still probably over 99% of the Alexa data comes from IE users, it is no longer a 100% IE thing. Here's the extension:
http://www.quirk.biz/searchstatus/
PS: I have nothing to do with that extension - I mention it just FYI.
Posted by: RBA | Apr 26, 2007 12:42:56 AM
hey... where are the "two" charts; only one is showing up.
Posted by: Richard Muscat | Apr 26, 2007 11:50:54 AM
I have compared stats from alexa, trafficranking.com and comscore and they all vary widely. I see that certain sites that have alexa ranks of 900 have less traffic then sites ranked 38,000 on alexa.
Even the trends dont coorespond to reality. I have figured out one thing. If the site has a high alexa rank, it has a whole lot of webmasters that are using it.
Posted by: Rafael Cosentino | Apr 26, 2007 12:11:52 PM
i have been using alexa from last 1 year and i have found it is very statistical in terms of traffic.
dipankar
Posted by: Dipankar Chakraborty | Jun 18, 2007 1:53:14 PM
Main RSS Feed of All Posts
VC and Technology Posts Only
