StatCounter and Net Applications still disagree on IE vs Chrome

We have reported in the past on the differences in methodology between StatCounter and Net Applications when it comes to finding out which web browser has the most market share among PC users. Indeed, StatCounter has made no bones about how it feels that its method of monitoring page views on websites is more accurate than Net Applications' method of using unique visitors. StatCounter also feels that its use of tracking three million websites is also better than Net Applications, which only uses 40,000 sites.

This week, both StatCounter and Net Applications showed their data for the web browser market share in September 2012. Not surprisingly, both still seem to differ on how much Google's Chrome browser is used compared to Microsoft's Internet Explorer.

Net Applications' data shows that IE 8 is the top dog on the PC web browser circuit, with 24.09 percent, followed by IE 9 with 19.45 percent. Chrome 21, which got updated to Chrome 22 last week, has a 14.32 percent of the web browser market share.

StatCounter's numbers are very different. They show Chrome 21 with a commanding lead at 27.34 percent. IE9 is second on this list with 17.93, followed by IE8 at 13.06 percent.

If there is one thing that both Stat Counter and Net Applications are close to an agreement of, its Mozilla's Firefox and its place in the web browser standings. Firefox 15 is fourth on both lists, with Net Applications giving it a market share of 10.91 percent and StatCounter giving it a little bit more at 13.01 percent.

Sources: StatCounter and Net Applications | Images via StatCounter and Net Applications

Report a problem with article
Previous Story

Windows 8 has five times less pre-release users than Windows 7

Next Story

Office Web Apps touch-based viewing features explained

29 Comments

Commenting is disabled on this article.

@ still1

"i am totally with global stat on this. google did confirm it last month that according to their stat chrome is the most used browser and is backed up by global stat."

Well - if Google confirmed Chrome is number one browser in the world, then it must be true! LOL

Now, once MS says it has number one browser - then you'll be really confused...

there is something wrong with both of them. how all chrome users are using chrome 21 and nobody on chrome 20 or 19 whereas on IE is a different story. first of all as for me I am using all browser because I am a web developer. I know many corporation must use IE because is mandatory by their infrastructure systems probably because its easier to maintain. only home users (who are free to choose browser) should be counted. I believe most of them at least power-users do not use IE and perhaps use either firefox or chrome.

S3P€hR said,
there is something wrong with both of them. how all chrome users are using chrome 21 and nobody on chrome 20 or 19 whereas on IE is a different story. first of all as for me I am using all browser because I am a web developer. I know many corporation must use IE because is mandatory by their infrastructure systems probably because its easier to maintain. only home users (who are free to choose browser) should be counted. I believe most of them at least power-users do not use IE and perhaps use either firefox or chrome.

Probably because Chrome updates in the background and IE doesn't

Few months ago I had some e-mail exchange with StatCounter in which they recognized that they were not capable to differentiate the chromium based browsers. They are reporting, e.g. Chromium and Comodo Dragon browsers both to be Google Chrome.
Also, they make no difference between Mozilla browsers like PaleMoon and Firefox, for example.

With all do respect, please find below the last of my e-mails (copy/paste), in which I deleted the name of your employee. I am sure you can find all the other e-mails.
---------------------------------------
Hi,

I'm afraid this user agent string does not contain any information which confirms that it is in fact Chromium instead of Chrome, and so we cannot apply this to our system.

- Signature deleted -
---
StatCounter Global Stats

On Wed, 23 May 2012 06:25:57 +0100, ctinro wrote:
> Useragent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.1 (KHTML, like Gecko) Chrome/21.0.1136.0 Safari/537.1
>
> Mobile Device: No
> Tablet: No
> RSS Reader/Feed Reader: No
> Crawler/Robot: No
> Browser Name CHANGE: Chrome to Chromium
> Browser Version: 21.0
> OS: Win7
> Hardware Vendor:
> Hardware Model:
> Screen Width: Unknown
> Screen Height: Unknown
>
> Comments:
> as discussed in (previous) e-mails. As you can notice, THE original Chromium is also reported as Chrome which, most of people will associate it with Google. I am sure that is the same situation with Waterfox for Firefox (I shall not test it), as happened with ComodoDragon and PaleMoon. And I assure that these are strong communities of users. Regards.

The useragent you submitted

Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.1 (KHTML, like Gecko) Chrome/21.0.1136.0 Safari/537.1

is identical to a Chrome useragent. It has nothing to identify it as Chromium. So we must detect it as Chrome.

However, if you submit a Chromium useragent such as

Mozilla/5.0 (X11; U; Linux i686; en-US) AppleWebKit/534.16 (KHTML, like Gecko) Ubuntu/10.10 Chromium/10.0.648.133 Chrome/10.0.648.133 Safari/534.16

it will be detected as Chromium.

statcounter received some google money and NA had some of MS's?
well, i like MS better so NA is right - so much for statistics

stats counter is pretty crap at all there stuff and probably net applications is not better, these stats sites should just go away they are useless at what they do and each have there bias.

ACTIONpack said,
I will be happy when IE7 and 8 will be under 5%. IE9 was the first good browser from Microsoft since IE5.5

IE 6 was not really that bad at the beginning. The problem is it took 5 years for MS to update it.

ACTIONpack said,
I will be happy when IE7 and 8 will be under 5%. IE9 was the first good browser from Microsoft since IE5.5
The problem with IE 9 is that it's slow even on hi-spec computers, crashes more frequent and has a bare-minimum UI that isn't really appealing to users. Chrome on the other hand has proven to be faster, more reliable when it crashes, have a better UI and more options to personalize it than IE.

Pubudu Kodikara said,
This is driving us crazy... who tells the truth?

Well, considering one is based on pageviews and one is based on unique visitors. I'll go with the unique visitors, especially when knowing Chrome preloads many websites the user will never visit. creating a vast amount of pageviews that arent real.
Firefox also does this but ALLOT less.

Also the first time Chrome came even close to IE's values, was right after the release of their preloading bullsh*t. Coincidence much?

StatCounter Global Stats said,
Prerendered pages (which are not actually viewed) are not included in our stats.

http://gs.statcounter.com/faq#prerendering

Our stats are based on internet usage trends. To accurately measure usage, we have to base our stats on page views (and not unique visitors). Please see our video explanation for this below:

http://youtu.be/J_OO2rcZ7bU


And you see at the point where it was launched there is some increase in Chrome.
And you strictly refer to Prerender, asin http://www.chromium.org/develo.../design-documents/prerender
Which is not the same as preloading pages when you type them, preloading URLS you might click on websites (FF also does/did this) etc.
Which is what I was refering too, as these tactics create ALLOT of fake hits.

FF is not in the pre-loading list. (they might've dropped it in the meantime and I'm currently not in the mood to go figure this out).
But worst IMHO is that you guys clearly state the prerender "link rel=prerender" tag. And do not say a word about Chrome's pageloading when typing in bar (i.e. loading several websites when you're only accessing 1).
Or the pre-loading Chrome and FF do (not sure about opera/safari) when it sees links on the site and thinks 'hey the user might wanna go here, lets start loading'.
These tactics create a vast amount of fake pageview traffic. Reasons like this is why I rather stick with the statistics that base it on unique visitors (which is also distorted a bit thanks to this, but allot less then pageviews)

And just wondering, as I could not find it on the FAQ.
Do you guys actively track IE9 Compatibility mode as IE7 or IE9? Considering it sends out an IE7 user-agent (albeit with Trident5.0 and NT 6.*)

IE9 with Compatibility mode User Agent:

Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 6.0; Trident/5.0)

EDIT: Playing around with dev tools in IE10, it did show IE10 on http://gs.statcounter.com/detect in Compatibility Mode (I have no IE9 to test on currently).
However it does fool the detector when using dev tools and manually selecting IE9, IE8 or IE7

Edited by ShadowMajestic, Oct 2 2012, 12:01pm :

Shadowzz said,

Well, considering one is based on pageviews and one is based on unique visitors. I'll go with the unique visitors

NetApplications doesn't just go with unique visitors for only 40k sites, it also manipulates the numbers based on government data and so forth, which basically means they're fudged.

StatCounter provides the raw unfiltered (mostly) stats from millions of web sites.

I know which I prefer to believe.

Shadowzz said,

especially when knowing Chrome preloads many websites the user will never visit. creating a vast amount of pageviews that arent real.
Firefox also does this but ALLOT less.

StatCounter has said that they don't count preloads. So you're just spreading FUD here.

Some points of interest:


Prerendering was announced by Chrome in June 2011. This change did not have any significant impact on our stats.
Chrome is currently allowing the detection of prerendering behavior via its Page Visibility API.

Shadowzz said,

Also the first time Chrome came even close to IE's values, was right after the release of their preloading bullsh*t. Coincidence much?

All statistic counters agree that IE usage is on the decline. Eventually NetApplications, even with its skewed data, will succumb to IE losing its lead. It's only a matter of time.

Shadowzz said,

And you see at the point where it was launched there is some increase in Chrome.

As stated in our FAQ, this is not the case. The introduction of prerendering did not have any significant impact on our stats.

http://gs.statcounter.com/faq#prerendering

Shadowzz said,
But worst IMHO is that you guys clearly state the prerender "link rel=prerender" tag. And do not say a word about Chrome's pageloading when typing in bar (i.e. loading several websites when you're only accessing 1).

Prerendered pages which are not viewed are excluded from our stats. It does not matter if they are triggered by a link or typing in the omnibox.

If you have any further queries please send them to us at http://gs.statcounter.com/feedback and we'd be happy to answer them.

StatCounter Global Stats said,

As stated in our FAQ, this is not the case. The introduction of prerendering did not have any significant impact on our stats.

http://gs.statcounter.com/faq#prerendering

Prerendered pages which are not viewed are excluded from our stats. It does not matter if they are triggered by a link or typing in the omnibox.

If you have any further queries please send them to us at http://gs.statcounter.com/feedback and we'd be happy to answer them.

i am totally with global stat on this. google did confirm it last month that according to their stat chrome is the most used browser and is backed up by global stat.

simplezz said,
NetApplications doesn't just go with unique visitors for only 40k sites

You realize you can sample 90% of the entire internet population by just monitoring one site (for ex. Google), and have only 1% worth of sample rate by monitoring millions of sites (each of which have only 1 or 2 visitors a day), don't you? So the number of sites monitored alone says NOTHING about how good even the raw sample set of a monitoring service is. And since SC avoids at all cost to differentiate their users, you can't even tell how much of the internet population they're monitoring. Even with millions of sites monitored, for what we know that could be very well below <1%.

simplezz said,
it also manipulates the numbers based on government data and so forth, which basically means they're fudged.

Wrong again. What you call "manipulation" is called "normalization", and contrary to what you state it does not "fudge", but rather "corrects" the the raw samples, which are inherently prone to sample bias. Actually that's how all real statistics works: in order to gain results which are representative for the entire population you have to weight your raw samples, because only in the rarest cases will you be able to gain raw data from a group which has the same composition as the one you want to project your result numbers onto.

So by weighing their numbers Net Applications doe exactly that what anybody who has any clue about how statistics work would do to get a meaningful results; and because SC doesn't nothing alike, their numbers are totally useless, even if you don't take into consideration that they don't differentiate between users, and can't even properly filter out number-skewing hits like those originating from pre-loading of pages.

Now I understand you hate IE, but also hate the thought of belonging to a minority - so you prefer and rather want to have those statistics true that show other browsers, not IE, as the most or more popular ones. Still, that changes nothing on the fact, that Net Applications does good statistics (both because they differentiate between visitors and also because they weigh numbers), and that StatCounter's numbers are totally useless and meaningless (because they count pageviews and because they don't even weigh and correct their raw samples).

osm0sis said,
Looks like Net Apps actually shows IE8 as top according to your pics..

Also according to the words if you read them...