02.06.09
Gemini version available ♊︎Summary: Lies, Damn Lies and Net Applications (Fake ‘Statistics’)
DEBUNKING MISCONCEPTIONS and erroneous consensuses is one of our key goals here, which is why we wrote the following posts earlier this week: 1) Microsoft and Apple Pay Net Applications (Hitslink); 2) Net Applications: the Big Lie, Boosted by IDC|IDG et al
Since these posts were initially written we’ve learned two more things; the first is that Net Applications changes its own statistics after publication, with evidence here.
Why is Net Applications (Hitslink) changing its browser stats after publishing them?
[...]
When I got back later and reloaded the page, I noticed that it had gone down to 0.5% or so the last few hours. I still left the page open, and returned a little later. To my surprise, the page was no longer showing the same numbers for the same time. It’s as if it had never shown 0.5%.
The second thing we have learned is that one of the executives at Net Applications used to work for Microsoft.
And to summarise the points that were raised before:
- Microsoft and Apple put money on Net Applications’ table, so rather unsurprisingly, the results satisfy both companies. GNU/Linux, on the other hand, is not able to pay Net Applications for favourable bias. Might the inclusion of logs from Apple sites (maybe even Apple.com) explain the high Hackintosh share?
- Net Applications uses a Microsoft stack for hosting, which is not so common
- Net Applications does not count actual traffic (e.g. page requests), based on its very own explanation
- Net Applications admits its statistics are flawed (skewed)
- Net Applications keeps its methods secret and the dataset likewise
When scientists who are merely “hired guns” get exposed, the validity of their output is no more. Net Applications fits this description and it absolutely must not be seen as ‘gold standard’. The output may be designed to discourage adoption of products from their clients’ competitors. IDC is doing the same thing with Microsoft (for servers) and those spreading the message even turned from journalists into Microsoft employees. █
“Mind Control: To control mental output you have to control mental input. Take control of the channels by which developers receive information, then they can only think about the things you tell them. Thus, you control mindshare!”
–Microsoft, internal document [PDF]
ᶃ Gemini Space
This post is also available in 
Content is available under CC-BY-SA
Roy Schestowitz said,
February 7, 2009 at 9:19 am
Addendum:
One source made a claim that we could not confirm. It said:
“When net applications was soliciting sites to be part of their counter, it gave a pretty good clue as to why the counts are a bit “off”. The wanted sites to serve up this ActiveX control that would got to user’s browsers and count the individual users, sending the stats to Net Applications’ counter via http post requests.
“This of course is a problem because most people who use other browsers use them because they don’t want Active X invading their PCs all the time.”
Jose_X said,
February 7, 2009 at 12:13 pm
I won’t accuse them of doing this without knowing, but I have almost zero doubt in my mind that active x or similar enough tricks are something of a pre-req.
In short, I expect the 0.8% (Linux) to come from visitors to ms-friendly sites.. and not be a representative figure of the general browsing on the Internet.
When groups hide their methodology, it’s difficult to believe their “trust me”.
For every such “fact” put out my Microsoft or by journalists repeating those numbers we need to put out the alternative numbers.
Roy, maybe you want to keep here a list of alternative samplings (whether scientific (preferably) or not). This way we can all point to this page each and every single time anyone mentions the 0.8% number. We can be just as honest yet unscientific about this. We can write to such authors quoting the 0.8% figure to state there is doubt about that and point to our page.
Preferably, each entry will include the details of how the stats were accumulated. And Linux numbers don’t need to be big. People don’t need to be ashamed if some months the numbers are low. We know it’s a minority that use Linux daily.
We may even rank the entries based on who appears to have put the most effort to be fair.
Also, preferably the website would be of topics not related to OS or FOSS, but, given the potential to cheat with things like active x (clearly an OS requirement), I wouldn’t reject those entries (ie, I’d accept bias entries to help prove a point.. without methodology the deck might be very or even very very stacked against Linux).
Carla (LT) and many others might be willing to do monthly tallies or something. I’ll try to bring it up at some point when I’m over there.
Hey, I’ll bring this up on thetuxproject.com since this is marketing. We can try to get volunteers to always complain to authors posting the other number and to comment pointing to this alt page.
We can also publish our own methodologies (unscientific most likely since we lack $$) which others may want to use. A weekend breakdown is useful since many businesses require IE use.
You do very decent stats Roy (don’t have handy the link to your last breakdown), so perhaps BN will want to kick things off (and host it).
You may want to also tag each entry that requires platform specific things.
Roy Schestowitz said,
February 7, 2009 at 12:42 pm
This sounds like something that would require a lot of work. Can’t writers just be told off for referencing figures that are essentially sponsored by Microsoft and are inaccurate (by admission from the ‘researchers’)?