Anthony Watts is very fond of using Alexa statistics for showing how popular his blog is. Especially when he can use it to show that his blog is more popular than websites that spread good information on climate science.

But the problem is that Alexa uses indirect measurements to give an estimate for how much a website is visited. This makes Alexa traffic statistics basically worthless if you’re trying to do any serious analysis of visitor numbers to websites. You just don’t use it as you will almost always get something that isn’t remotely close to reality (although some businesses do use this data).

Watts should know this considering how often it was pointed out to him that Alexa isn’t reliable. Something I also wrote about in my blog post ‘Why You Shouldn’t Use Alexa Traffic Statistics‘. Yet it doesn’t stop Watts from using it to boast about his website in his blog post ‘The other divergence problem – climate communications‘ (archived here):

So, I thought I’d check and run some numbers to see if I’m shamed or not. Climateers often talk about their climate change cause being a “communications problem”. The numbers I’ve found seem to support that. Witness the new divergence problem:

These are rankings from Alexa.com Lower numbers are better, for example, Google is ranked #1.

It seems that it’s not just globally strong for WUWT, but in the USA too. WUWT is about 8 times more popular in the USA than “Skeptical Science” (SkS), and about 15 times more popular than Real Climate (RC).

Considering what I’ve already written before about Alexa traffic statistics this graph doesn’t tell you anything about the statistics for those websites. Especially when you compare the increase in ‘traffic’ that Alexa is reporting for WUWT to direct measurements:

The amount of page views on WUWT are the same as they were in April of this year. Same goes for the other statistics on Quantcast. The only exception being the number of visits, but the number of visits is not that interesting (which makes me suspect that Alexa puts a lot of weight on these statistics in their ranking algorithm). What matters is the number of unique visitors and the page views a website gets. Those are the most interesting ones if you want to figure out the reach of content on your website. Which is what Watts is doing.

This mismatch between direct measurements and the indirect measurements made by Alexa is why you shouldn’t use Alexa for comparing traffic. It’s not accurate enough for this.

And then he says this:

While I wish I could run more than 10 on the same graph,  here’s what I learned, again lower numbers are better:

Surprisingly, not only is WUWT leading the pack by a significant margin, it has now surpassed the newspaper “Grist” which has become something of a climate centric enterprise. They also have a paid staff.

No, he hasn’t surpassed the newspaper Grist:

And this is me being generous towards WUWT, the number of page views is the only statistic WUWT comes close to Grist. And here Grist still has a lead of half a million page views per month and gets more page views on most days.

This is also where it gets really interesting if you know how to read and interpret this type of data. If you look at the unique visitors per month you’ll notice that Grist has 952,484 unique visitors versus the 205,950 that WUWT gets. More people see the content on Grist compared to the content on WUWT, and thus Grist has a larger reach than WUWT.

The numbers do suggest a higher engagement per visitor for the content on WUWT, but that’s not as straightforward as you might think. If you place a comment on WUWT the page you post the comment on gets refreshed. Meaning that a new page view is logged in Quantcast. When you place a comment on Grist only a part of the page updates, which doesn’t cause Quantcast to log a page view. This lowers the number of page views for Grist.

Another detail is that a part of the visitors on WUWT are very active in the comment sections. It’s this that increases the page view count for WUWT. It’s not that WUWT has a bigger audience or more reach than Grist.

With the number of times critics have pointed this out to Watts he should be aware of these limitations the Alexa data has. The only reason I can think of at the moment why he still uses Alexa statistics is because it gives him the narrative he wants.

Watts indeed surpasses Grist in the number of page views. His website gets 2.8 million page views compared to the 1.8 million of Grist. However, there’s a reason why this number is so much bigger. This seems to be mostly caused by a part of the visitors on WUWT that are very active in the comment sections. It’s people returning to WUWT more often that increases the page view count for WUWT. It’s not that WUWT has a bigger audience or more reach than Grist.

This shows if you look at the number of uniques, people, and visits statistics that Quantcast reports:

What these statistics show is that more people see the Grist content compared to WUWT which means that it has a bigger reach compared to WUWT (uniques and people). These numbers do show a higher engagement per visitor for the content on WUWT (visits and page views vs uniques and people), but that’s also not as straightforward as you might think.

If you place a comment on WUWT the page you post the comment on gets refreshed. Meaning that a new page view is logged in Quantcast. When you place a comment on Grist only a part of the page updates, which doesn’t cause Quantcast to log a page view; this lowers the number of page views for Grist. Granted with the amount of comments Grist gets this effect isn’t big, but it’s there influencing these calculations.

The problem with this is that Watts uses these statistics to suggest that his content gets read by more people than for example Grist. But the statistics we have from direct measurements show that this isn’t the case. Watts might simply not know this, but you should be aware of this when you try to compare websites.

Same goes for his usage of Alexa statistics. But here he should know better with the number of times critics have pointed out the limitations that the Alexa data has. The only reason I can think of at the moment why he still uses Alexa statistics is because it gives him the narrative he wants.

Update 26-10-2013 @ 15:42
Today I noticed that there seems to be a horrendous bug in Quantcast when you let it compare websites. The numbers it displays during a comparison are not correct and do not match the reported statistics (they even change depending on if you compare site 1 with site 2 or vice versa). I’ve changed the above post so that it now uses the correct numbers and changed my conclusions where necessary. My apologies towards Watts for the inaccuracies this caused in my blog post.