Smells fishy: Alexa's data blunder hits Drudge, WUWT, mostly favors leftist news sites over conservative news sites

As many WUWT readers know, I have been using alexa.com for quite sometime to gauge the performance of WUWT. Reader “Pat” brought this recent strange disparity to my attention. When you see things like the Drudge report plummet and MSNBC soar, you know immediately that something isn’t right:

Those who run watchdog news websites are scratching their heads and trying to make sense of the latest data released by a California company that measures website traffic.

According to data for July through September, almost every major website – from WND to the Drudge Report and Breitbart – saw its rankings drop on Alexa.com while pro-government sites mostly went up.

Source: WND

The claim seems to be true when you look at Alexa’s publicly reported traffic graphs. Here’s the Drudge Report according to Alexa:

(Lower numbers are better, for example, Google is #1)

drudge[1]And here is MSNBC according to Alexa:

Alexa_msnbcI find that hard to believe. I find this even harder to believe, Fox News according to Alexa:

Alexa_Fox[1]And the Soros driven Center for American Progress, aka thinkprogress.org according to Alexa

Alexa_thinkprogressThe exception to that knee up for left leaning websites starting in September is CNN.com according to Alexa:

Alexa_CNN

So, this may be some sort of reporting or tracking goof. But, the oddity does not stop there.

Now, look at WUWT compared to the eco-news website “Grist”, which we long ago surpassed.

First, here is WUWT according to Alexa. Note the same “knee” in the graph as seen on the Drudgereport earlier:Alexa_WUWT

And here is grist.org according to Alexa. Same “knee”, opposite direction:

Alexa_Grist

Hmmm.

My internal wordpress.com counter for WUWT shows that WUWT had its best month ever in total views, nearly 5 million:

WUWT_monthly_trafficAnd, WUWT’s best ever day in its seven year history was in September 2014:

WUWT_bestever_trafficSomething is broken at Alexa that seems to favor one sort of website over another. What’s up with that?

It may have to do with this recent “upgrade” at Alexa that explains a new method of calculating its rankings based on what it says is a wider panel of viewers.

“We’re excited to announce that after a lot of hard work, the size of Alexa’s global traffic panel will be significantly increasing,” writes Leigh Katcher in the blog. “Over the next month we’ll be incorporating lots of new data points, which will help improve the accuracy of our metrics. As a reminder, our data panel is a sample of global Internet traffic used to calculate Alexa ranks and estimate non-Certified metrics.”

The blog posting continued, “With better data, we’ll be able to offer deeper insights into your site, your competitors’ site and overall traffic strategy. One immediate result of the additional data is that you may see your traffic rank fluctuate, especially for sites ranked greater than 100,000.”

One wonders how such an “upgrade” can affect one type of site more than others. In the WND article, they say the disparity even extends to things like Planned Parenthood versus Lifenews, two sites that offer politically opposing viewpoints.

We live in interesting times.

 

Advertisements

  Subscribe  
newest oldest most voted
Notify of

Here’s hoping its just a mistake – it would just be pathetic beyond words if a web ratings agency decided they had to cook the books to help the cause.

It’s all propaganda. Once you realize that simple fact you can see so much more that is complete BS.

M Seward

“cook the books to help the cause” — noooo! Surely not? A ‘progressive’ outfit cooking the books.? Moral authority is an unerring force for good, How can moral authority lead to cooking the books?
sarc off/

This is happening with increasing regularity. George Orwell would understand.
One of the few legitimate polling outfits left [the Gallup Poll] lists ‘climate change’ as the very least important to the public. Everything else rates higher:
http://content.gallup.com/origin/gallupinc/GallupSpaces/Production/Cms/POLL/h7fjmcrciu6d7y5yx_-zjq.png

Bill 2

So if climate change is least important to people, why is it surprising that traffic to this site is decreasing?

nielszoo

I hate to say it but you can’t trust Gallup either. They folded last election cycle to pressure from Holder’s DoJ. We are living in a post legal world. “Our” government is pressuring (and by that I mean acting as an extortionist) our banks to stop doing business with perfectly legal businesses. Privately owned businesses are being forced to take jobs that directly conflict with their religious beliefs by an out of control judiciary. Freedom of speech is being systematically turned into Progressive censorship with the full power of an out-of-control federal government with 2.8 MILLION petty bureaucrats behind it… and that’s the sad truth.

vern

Perhaps it’s time to dump Alexa. http://www.similarweb.com provides a slightly better presentation format anyhow.

Konrad

It’s not looking good for “mistake”.
Remember Alexa is run by Amazon. Something very similar happened previously with rankings for sceptic books in their climate books rankings.
This time the site bias looks too distinct to be computer error.

Brute

It is not an uncommon practice. The IMDb, for instance, fixes ratings of “ecomedia” while dissenting comments fail to make it through.
The thing to do is what’s been done now, that is, to denounce it publicly. I would bet that there very few cases where this sort of thing happens due to “orders from the top”. Most likely, it is the random eco-soldier taking matters into his/her hands.

PeterK

Eric: You mean “massage the data” don’t you?

Gary

Obviously it’s a result of station moves, equipment changes, and “adjustments”. 😉

Hector Pascal

You missed “Time of observation bias”. TOBS explains everything.

Lars P.

I think you are right TOBS adjustment seems to be the way how they did it. Nevertheless, I still wonder how the algorhytm runs…

Louis

And the adjustments are probably based on models. Why bother to measure actual results when you can just use models to get the results you want?

DD More

“Over the next month we’ll be incorporating lots of new data points, which will help improve the accuracy of our metrics. As a reminder, our data panel is a sample of global Internet traffic used to calculate Alexa ranks and estimate non-Certified metrics.”
Or, just like ‘The Hottest Year Ever’, it was hidden where no one lives.

Lester Via

Isn’t the Y axis of Alexa’s graphs upside down?

Yes… only the two of us seem to have noticed that 🙂

scot

The Y axis is the log of the rank. Lower numbers indicate higher rankings.

From the author’s OP, “(Lower numbers are better, for example, Google is #1)”

It’s not a measure of the # of views, but the popularity. Just chant to yourself “Watch out for Number 1” when you’re on their site and the graphs will make more sense.

Jim G

Would be nice if graphs had their axes labeled. I find this is semi-often a problem on this site. Not that I am complaining!! Still the best site for science, in general, as far as I am concerned.

Brian H

Deliberate, to show “better” as up (lower numbers) and worse as down. I.e. better ranking goes up on the graph. Perfectly fine as long as it’s done consistently.

DirkH

“One wonders how such an “upgrade” can affect one type of site more than others.”
Maybe they added data from Oakland?

Brock Way

I think it is a time of observation difference.

The scale at left is “upside down”, no? So the ‘leftier’ sites are losing, t’udders increasing.

simple-touriste

“You know Homer, the traditional way to cheat at golf is to lower your score.”

bonanzapilot

That’s because since lower is better, the inverted y-axis gives a more intuitive display.

Will Nelson

“V” or “straight” tail?

I wonder which collection procedure is more accurate. I’m more interested in how they changed things and what was wrong with the old way. And what is wrong with the new way.

simple-touriste

They do “homogenization”?

Mike M

You beat me to it!

Steve Keohane

What is a MSNBC?

DirkH

Microsoft National Broadcasting?

mikeishere

Microsoft National Broadcasting Comrades

DirkH

🙂

Owen in GA

The ugly offspring of the union of Microsoft and NBC (National Broadcasting Company). The child is so ugly that Microsoft abandoned it in the wilderness a few years after its birth.

Is Alexa now measuring tree ring width, and employing a new and unique multivariate analysis, to measure viewership rates.

How about an algorithm that can turn random numbers into a hockey stick/viewership curve?

Sean Peake

The Tiljander Syndrome?

Very good! [Maybe a little too much inside baseball for some…]

DirkH

Oh, I see they are owned by Amazon. Jeff Bezos having a word with his Alexa CEO I guess.

Tom in Florida

Perhaps he asked them to Mann up.

Nick Milner

A couple of those actually look like hoc… no, probably an optical illusion. (Sorry!)

Sweet Old Bob

If the data does not fit THEIR model , it MUST be adjusted…..

Robert of Ottawa

Maybe some disillusioned crimatologists have taken up new employment 🙂

TheLastDemocrat

they possibly switched to monitoring traffic indicators that represent more government activity than private individual activity.
This could be done without it being obvious, and all the while genuinely seeming as if the coverage were becoming more representative – and while actually, literally covering a greater portion or a more “geographically” representative portion of web traffic.
Overall, I have no problem believing it has been done on purpose. The Marxists have been well aware of the relevance of media consumption. This is the basis of a couple leading totalitarianism control efforts that appear well-meaning: the concept that conservatives are in an “echo chamber” and the concept that the government must ensure “diversity” in media so that the public has a sufficient knowledge base to participate in democracy.
The term is “deliberative democracy.” Sunstein The Totalitarian has a book out on this: “Democracy and the Problem of Free Speech.”
Sunstein is a leading edge for the totalitarians. He takes their wolf views and puts them in sheep’s clothing. After reviewing his views on why the govt should dictate portions of viewpoints flowing through the media, in this book, you can then review how he believes the Constitution needs to be re-written: The Partial Constitution. [You can also visit the website of this movement: find the website for “The Constitution in 2020.”]
We know Al Gore censored Google. That was well-documented: when climategate broke, many googled “climategate.” This should have prompted auto-complete to yield “climategate” if you typed “climateg ” into google search bar. But it did not. Even after this censorship was widely shared, and many people virally tried it themselves – the auto-complete failed to respond to the massive google searches for climategate.
https://www.seroundtable.com/archives/021306.html
These people are not regular U.S. Constitution-believing citizens who happen to be democrats; they are using the democrat party and label to advance their farther-left views. Enjoy.

Dire Wolf

Thanks for the information. It is always hard to keep up with our “betters” and how they want to improve on our freedoms.

Why would they release graphs with an inverted Y-axis? Misleading doesn’t even cover it. It’s just bizarre.

Huroner

It isn’t inverted: a higher score means a worse performance and lower means better. Look at it this way: if you were top your score would be 1; if you were 100th your score would be 100.

It’s their algorithms, I suppose they can cook them anyway they wish. But if they misrepresent the value of the data to advertisers they risk their own business.

The Other Phil

Huh? It makes perfect sense. Up is good, down is bad. Would you suggest the opposite?

bonanzapilot

That’s why I think they invert it.

Bloke down the pub

Did the Ministry of Truth get involved?

Kasuha

I don’t think Alexa is to blame, it is just counting page views. Something made a lot of people – or at least their computers – to view pages that are otherwise seldom visited, increasing their rank and pushing ranks of other pages down without really affecting their traffic.
It might be related to Ebola for what I know.

DirkH

Ebola makes people visit Grist?
The only thing that could cause a surge in Grist’s traffic would be if people turned en masse into middle-aged, liberal women. I’m using Alexa’s own information here.

Kasuha

I don’t see Grist getting out of its fluctuation noise level. The same for MSNBC, the same for thinkprogress. I think it’s a bit early to start spinning conspiracy theories.

Justin

A better indicator would be Google trends, which measures pretty much everything. The first link is for Watts up with that. The second link is for Thinkprogress.org. The third is for msnbc.com. The fourth for CNN.
WUWT: http://www.google.com/trends/explore#q=%2Fm%2F0bh9jm4&cmpt=q
Thinkprogess: http://www.google.com/trends/explore#q=ThinkProgress.org&cmpt=q
msnbc: http://www.google.com/trends/explore#q=%22msnbc.com%22&cmpt=q
CNN: http://www.google.com/trends/explore#q=%22cnn.com%22&cmpt=q
Also, if you enter terms for global warming and climate change, it shows a declining interest in the results graph. I think people are just not interested in the subject, which may explain why it ranks dead last in polls measuring what people are most concerned about.

Dire Wolf

Thanks. That really shows where “Think Progress” sits. What do the numbers actually count? Is it some multiplier of hits or searches?

Jeff Alberts

So WUWT is hotter in South Aftrica and India than in the US? Hmm…

People have left off believing the alarmists. That’s why it’s less of an issue with them. They want jobs for loggers more than they care about the spotted owl. Drill, baby, drill!

cwon14

Must be “models” aka “just making it up to fit a political meme”.
Sounds like consensus climate science.

Leon Brozyna

Well, I suppose Audit Bureau of Circulation they ain’t … the only metric that matters … $$$

Jimbo

Alexa does not matter. What matters is your own stats coming in – actual visitor. The only problem is that you cannot easily compare – but you can compare yourself to yourself.
PS I vaguely remember that Alexa only has data for those who bother to download their toolbar. I maybe mistaken.

Jimbo

I have my own webiste and I use Google Analytics. I compare my past performance to today. I could not give a rat’s fart what others are doing.

Google often produces curious results dependent on how political the query is. I’ve always put that down to Al Gore being a Senior Adviser to their board. Perhaps he’s now got another nice little earner as a Senior Advisor to Alexa as well?
Pointman

Jim G

Well, he invented the internet, after all!

lawrence Cornell

Yahoo search as well Pointman. Whose CEO is/was a major Obama campaign finance bundler.

Yes, I knew Google was rigging things against alternative Internet media when they announced they were going to add extra points to “expert” web sites. I don’t remember any explanation of what “expert” meant to them. Maybe Mann, probably not WUWT.
Alternative media meaning independent perspectives, libertarians, conservatives…
Seems like there may be a niche opening up. Hard to crack the Google shell, but….

Rob Potter

It sounds like Alexa is not actually counting page views, but using a sample set of “people’ and extrapolating what pages they view to the global population – a bit like Nielsen ratings or opinion polls. Such methods are prone to a major source of error – that the sample set does not represent the whole population. There are two ways to address this – try to get a “better” sample set (i.e. more representative), or weight the results based on how you think your sample is biased. Both of these allow room for positive discrimination (getting the result you want) and so there is always the need to consider the source of such sample-based polls.
If Alexa have changed their sample set or changed their weighting, then you might expect a step change in the rankings (as was mentioned above). However, since the ranking would not be a instantaneous thing, but probably based on a using a gradually reducing weight of past views together with present views, there would be a gradual change to the new levels as the older data has a gradually reducing effect on the ranking.
On the whole, I would surmise that Alexa has changed their weighting algorithm to reflect what they think is a better representation of the whole population. This could be an innocent change (you always think that more people think like you) or deliberate (as has been suggested), but regardless some explanation from Alexa is required as these ratings are used by many people to gauge public opinion and are thus of more than just passing interest.

Billy Liar

Another major source of error in polls are that they reflect the views of the sort of people who self-select themselves for doing polls.

Dire Wolf

So you think this is like overweighting Democrats in political polls (saying 35% are Dems/25% Republicans when the registration numbers are reversed)? That seems a good guess. Thanks for your insight.

Lars P.

I think Rob Potter above nailed it:
October 15, 2014 at 6:49 am
“….On the whole, I would surmise that Alexa has changed their weighting algorithm to reflect what they think is a better representation of the whole population. …”
And from the way how the graphs move it really look like they use some adjustments – as Hector Pascal said, October 15, 2014 at 7:39 am
” You missed “Time of observation bias”. TOBS explains everything.”
Wonder when they will move to models and name it “data”. That must come as next step….

Richard Dekker

That’s a lot of hockeysticks.

jayhd

It is time for Drudge, WUWT and all the other conservative sites to stop referencing the stupid things MSNBC and the other “progressive” sites publish. Those references alone probably increase traffic.

M Courtney

And to increase groupthink. You need to see what those who disagree with you are saying to hone your arguments and understand your own weaknesses.
Only God has all of the truth. And He doesn’t blog.

God has a lot of busy little bees. Besides, he turns evil against itself.
Nobody believes in governments anymore, except the ones who profit from it. Look at the reaction to Ebola. The director of the CDC oozes incompetence in his daily briefings, with stops and starts in his reading of the Teleprompter.

I would also look at an organized effort among those on the left to cook the data, sorta like an Alexa bomb (see Google bomb of George Bush for an example). As I understood it to work last I looked Alexa uses data from browser plugins to monitor traffic among others. Webmasters have long been aware that for a low ranked site downloading the plugin yourself as you visit your own site constantly could do wonders for your rank. I wonder if an organization on the left could make sure the plugin was installed on all their members/employees/users etc computers so as to skew the data with an non representative sample…

M Courtney

But why?
OK, advertisers are likely to skew their payments on hits but equally so on the demographics.
It just seems like a bit of a pointless thing to do.

ImranCan

1984 here we come

Alx

As a reminder, our data panel is a sample of global Internet traffic
Alexa has developed it’s own “data” panel. I think what this means it selects who provides the data. The term data panel is misleading, it really is a panel of people and of course changing the demographics picked will change results.
This more “accurate” panel of internet users, was made up of a much greater percentage of left leaning users. If I was an advertiser who feels their product was geared to the left, for example Climate Change Anxiety Supplements, I would use Alexa data. Otherwise, it has no value in understanding internet traffic.
BTW “Climate Change Anxiety Supplements” or CCAS probably could be a big moneymaker, I should copyright that.

As taught by the left’s godfather; “I consider it completely unimportant who in the party will vote, or how; but what is extraordinarily important is this—who will count the votes, and how. Я считаю, что совершенно неважно, кто и как будет в партии голосовать; но вот что чрезвычайно важно, это – кто и как будет считать голоса.(The Memoirs of Stalin’s Former Secretary (1992) by Boris Bazhanov [Saint Petersburg] (Борис Бажанов. Воспоминания бывшего секретаря Сталина).

D. B. Cooper

So, does this mean Alexa has hired Gavin?

Doug

But WUWT is up…..you’ve been exposed, you closet commie!

Jeremy

We all saw how google manipulated searches to make it harder to find “climategate” – those who control the story are always inclined to twist things to their agenda. Don’t believe anything you read or are told! Check the facts as best as you can and if there are no facts to check then take whatever is said with a grain of salt.

jorgekafkazar

Make that a bag of salt.

markn

Odd. Know nothing about it. Suspect there must be various degrees of corruption. Be nice to know. Guess it’s between 10 – 30% depending where you are on the planet and what your business is.

Joe Minick

If I recall… In logic class… It was called slippery slope.

Slippery slopes are covered in rhetoric or propaganda or agit-prop,

Winston

Could it be related to the level of the service plan subscribed to? Yelp apparently blackmails businesses that don’t advertise by removing good reviews. Perhaps Alexa came up with an algorithm that is “more accurate” for customers with the their most expensive plan. About the Yelp technique:
http://market-ticker.org/akcs-www?singlepost=3364742
http://www.eastbayexpress.com/oakland/yelp-and-the-business-of-extortion-20/Content?oid=1176635
http://www.datalounge.com/cgi-bin/iowa/ajax.html?t=12380900#page:showThread,12380900

Collin Maessen

Well I’ve said it before to you that Alexa cannot be trusted for accurate statistics. They use indirect measurements to give an estimate on the traffic a website is receiving, where this is coming from, and what their audience is. At best you can use indirect methods to give an indication of what is going on, but even that cannot be trusted on.
The other detail also is that it’s quite easy to manipulate/pollute the data Alexa gathers. Though I do not yet know how their recent changes will affect this.
The only thing that I can say with certainty is that you shouldn’t rely on Alexa statistics, or any other statistics gathered through indirect measures. They all have some big disadvantages that make them unreliable compared to data gathered from direct measurements.
REPLY: Colin, as usual, you miss this real issue, due to your own biases against myself and WUWT. Alexa has been doing just fine for millions of websites, it’s the change now that is the issue. It can’t simply be explained away, even if you disagree with the previous data gathering methods. Just a reminder. You’re banned from commenting here, due to your previous thread disruptions spouting your bias, and some moderator accidentally let your comment through. So there is no need to reply. – Anthony

HA !!
Nothing can save CNN from bottoming out !!
Maybe the “New” system at Alexa allows leftists to fudge rankings like they do for google searches.

jeo

Oh god, please don’t cite WND. That makes this site’s credibility go down dramatically.

John Boles

I have found over the last few years that leftist bias suffers from incredible NOBLE CAUSE CORRUPTION and knows NO limits, yes it runs very deep! Leftists will do anything for their cause, as long as they do not have to sacrifice much.

Jim G

The left lies about everything, so why not this as well? Inflation, GDP, employment, IRS, Benghazi, fast and furious, Obamacare, yata, yata, yata…. If their lips are moving they are lying.

Mark from the Midwest

It’s really pretty simple, their recruitment process leads to a biased, not representative sample. When you do heavy recruiting on Puffington Host you will get a liberally biased sample. No one in the media industry takes Alexa seriously. Nielsen and Comscore are the only generally accepted measurement platforms.

You know all those inept reporters the MSM let go because they can’t sell enough of their reports to keep them employed? Well, probably Alexa hired some of them. Ergo, Alexa’s bad reporting of internet rankings.
John

schitzree

Since the rank reflects a websites place RELATIVE to all other sites, if 5 thousand progressive sites were pushed up in rank significant then all sites they passed would see a 5k drop in their rank.
Something must be inflating a lot of progressive sites if the effect shows up even on sites in the top tier though, and it must be a very large boost. Not just +/- 500 visits.

joe2011

i was wondering this too. Is it a huge influx of fake traffic to liberal websites? they liberals were found out to be sending tons of fake email to influence conservative reps and senators in 2010 on crucilal votes iirc. Fake twitter accts retweeting bs too.

Otteryd

I think we need to wait for five years worth of data before we can say for sure. Or ten years? Or Eighteen years? Or perhaps it will descend to zero by 2035?

Ralph Kramden

If Alexa.com’s numbers can’t be trusted it would finish Alexa.

eyesonu

Serious question here. I have WUWT bookmarked on my browser and use it exclusively to access WUWT. Since I don’t use ‘google search’ or other search engines to get here what effect would/could that have on various ranking results?

Harold

Kinda looks like Mike’s trick. Mikey likes it.

ossqss

Resistance is futile! You will be extrapolated!
Sarc》
http://upload.wikimedia.org/wikipedia/en/b/bc/Borg_Queen_2372.jpg

Here’s a reality check on just how fishy those numbers are – Alexa shows MSNBC’s rankings going up, when the New York Times has just published a piece on how MSNBC’s ratings just hit a new all time low.
http://www.nytimes.com/2014/10/13/business/media/leaning-forward-msnbc-loses-ground-to-rival-cnn-.html?_r=0
“However the slide of MSNBC is especially dramatic. The Times reports that MSNBC averaged 392,000 key demo viewers per night in the first quarter of 2009. Last quarter that number was down to 125,000, representing a loss of two-thirds of their total audience.”
So, Alexa is trying to say that even though on-air viewership has just hit an All Time Low, somehow online page views are going up. I don’t think so.
It may be time for some entrepreneur to start a competing ranking service, since Alexa’s can apparently no longer be trusted.