Linux’s Market Share: Is There Any Way To Know?
One thing many of us would like to know, I am sure, is how many people out there use Linux? The usual numbers, those from Net Applications, would indicate that less than 1% of people use Linux. More recently, though, there have been some more optimistic numbers. Canonical is claiming that 11% of businesses use Ubuntu. (If that is the case, just imagine how many use Red Hat or Suse.) Gartner says that Linux’s market share is 4%, putting it about even with the Mac. And Context says that almost 3% of PCs sold in the UK have Linux pre-installed.
While none of those numbers are huge, they mean the difference between Linux and Mac OS X being about equal and Linux being a speck of dust in the sea. The fundamental problem is that it is really, really hard to know how what the marketshare of Linux, or any open-source software, is. After all, one download might never be used, or only be used for a short time, and another might be used to install Linux onto 100 computers.
In addition to the standard market share numbers I mentioned above, there are at least two others projects trying to determine how many Linux users there are.
The first is the Linux counter. Their idea is to get people to visit the website and register as a Linux user. Obviously, this has some flaws in that anyone can register many times and most people will not bother to register.
The second project is the Open Source Census. The idea is similar, except that you download an application. This presumably gives them better control over double submissions, but it cannot be perfect and it is still voluntary.
Short of every Linux distro calling home, which I hope will never happen, there may never be a way to know how many Linux users there are. Perhaps the downloads of some cross-platform application would give a rough idea, but that application would have to be something that a Linux user, a Mac user, and a Windows user would want equally. If anyone has a better idea, I would love to hear it.
August 14th, 2008 at 10:58 pm
All that matters to me if that app A runs on Linux.
From what I hear, the more linux users, the higher are the chances!
August 14th, 2008 at 11:15 pm
Although it’s difficult to tell how many people are using Linux, Microsoft must have a very good idea of how many people are not using Vista.
August 15th, 2008 at 1:02 am
Ask google, everyone uses google, and browsers leave their version and running os in the logs…
August 15th, 2008 at 2:44 am
the distros could count the number of computers connecting to their update system.
they could make a unique hash for each computer that connects used only for the purpose of counting the users.
August 15th, 2008 at 2:48 am
http://fedoraproject.org/wiki/Statistics
August 15th, 2008 at 4:26 am
There is such application, it’s called Pixel image editor, but as far I know there are no official stats for these 3 operating systems… http://www.kanzelsberger.com
August 15th, 2008 at 4:42 am
Well actually there is such an application I think. When I tried the “Songbird” application for the first time, it seemed like it had something hidden inside that called home and in the main interface you would see in a webpage a pie-chart of how many Songbird users are on Mac, Windows, or Linux. They seemed to have removed that, perhaps people didnt like that kind of spying, but anyway I thought it was interesting to see the numbers, as this application was cross-platform from its beginning and seemed to be desired by all parties. It used to show about 11% using Linux, 9% on Mac, and 80% on Windows.
Sounds quite logic and right to me… But surprising is the fact that Linux outnumbered Macs!
August 15th, 2008 at 5:29 am
Even your last idea wouldn’t work because most linux users get their software from their distribution’s repositories, not from some website where you can count the number of downloads.
August 15th, 2008 at 6:36 am
Thank you for the article. In a short search a while ago, I tried to find out the market share of Linux as well. The only data, I could find are: 2%, 0.8%, 1.91%, 3.9%, 14.1%, and even 46%. Please see details below:
2%: http://www.webmasterpro.de/portal/webanalyse-systeme.html
0.8%: http://www.xitimonitor.com/en-us/internet-users-equipment/operating-systems-august-2007/index-1-2-7-107.html
1.91%: http://www.w3counter.com/globalstats.php
0.82%: Netapplications
3.9%: http://www.w3schools.com/browsers/browsers_os.asp
14.1%: http://www.heise.de/newsticker/Marktanteil-von-Firefox-kratzt-an-der-20-Prozent-Marke–/meldung/109877 (website of one of the largest and best German computer magazines)
46% http://www.gulp.de/kb/tools/trend.htm Project offers for IT freelancers with the knowledge Windows respective Linux. Gulp is the largest database where software projects look for computer freelancers in Germany.
August 15th, 2008 at 7:05 am
Case and point:
Well I must admit, I have installed the GNU/Linux operating system on at least 6 different computers using the one download. Then, my dad passes it on to his friends who in turn copy that disc and pass it to their friends, and so on.
Some of them are, and possible will never be, connected to the internet since gNewSense (Ubuntu) works out of the box without the need to download additional drivers or even software–it’s all included.
So it looks like that the internet can’t be used to count them either.
August 15th, 2008 at 7:38 am
How about adding up the numbers of users of the software repositories from the major distroes?
August 15th, 2008 at 8:57 am
Another route to calculate the actual installed base of Linux users comes from Canonical’s statement that 12M unique IP addresses hit their update server last year. Assuming that the average IP address hosts one computer (conservative), Ubuntu has 1/3 of the desktop market (from DistroWatch), and Gartner’s estimate of a billion computers in use worldwide, the overall Linux installed base comes out to about 3.6% - reasonably close to Gartner’s report of 4%.
August 15th, 2008 at 9:18 am
How about skype downloads? Firefox is not good, because it is preinstaled on most Linux distros.
August 15th, 2008 at 9:37 am
Hard to say overall, but the best bet for measuring *desktop* usage would seem to be hits against Google. Or perhaps hits against the top site in each major world region. Yahoo in parts of East Asia for instance (iirc).
However getting that data probably isn’t going to happen.
August 15th, 2008 at 9:44 am
The only way to try to quantify linux use is by statistical methods (random sampling, etc)
August 15th, 2008 at 10:16 am
Just about everyone uses Google on a regular basis. Google could record each unique IP address that visits it’s home page, and what operating system was used. It is not necessary for an end user to have any special software installed on their PC for this. For example if you visit the download page for Adobe Flash the site automatically recognizes what operation system you have. Google could do the same thing and record the results. The same PC would never be recorded twice, because Google would remember the IP address, and if you change operating systems then that could also be reflected in the data.
August 15th, 2008 at 11:09 am
+1 for counting via the distie updates - in turn the disties should be collaborating together to get the message out in unison i.e. each distie posts a figure to a central counting site.
The other place to look is here:
http://www.w3schools.com/browsers/browsers_os.asp
August 15th, 2008 at 11:35 am
Using Google’s unique IP address log isn’t going to give an exact picture because: a) gateways and routers can be used by multiple PCs; b) some users have a dual boot PC; c) most ISP clients use a dynamic IP address that could change several times over given period of time.
Other interesting points are: as stated before, not all PCs are connected to the Internet; you can have your browser report anything (OS) by changing it in its preferences.
August 15th, 2008 at 11:36 am
Valve Steam gaming on Wine users is well known 1% of all Steam users so reports saying that that at all Linux is only 1% PC is unreal. Also saying that more than 1 from 10 PC’s have Linux is quite unreal too….I think real number of Linux users is somewhere between 4% and 10%.
August 15th, 2008 at 12:28 pm
Regarding Songbird, I would tend to suspect the number of Mac users would be higher, except that I get the impression that Mac users tend to stick to Apple-branded software - so a smaller proportion of users will bother to examine a “third-party” media player.
As for measuring Linux use - I think the best you can do is to pick SOME way of measuring it (OS reported by browsers, inferred from sales or downloads, voluntary reporting, whatever), get a figure, and then prefix that figure with “more than” because there are always some people who spoof their user-agent string, or legally re-use downloaded distributions, and so forth.
August 15th, 2008 at 12:46 pm
Hi all,
The google thing is nearly perfect, but I see some flaws here
1: My IP changes everytime when I reconnect to the internet (or automatically after 24 hours)
2: If one is not connected to the internet, but still using his computer
3: If linux user are more afraid about their freedom and so not using any google sites
4: A shared internet connection (in houses, companies) where some computers have windows, others have linux, …
5: On one computer multiple users are working (so effectively n users are using that OS, but only one is counted)
Conclusion:
We only can count the computers that have pre-installed an OS.
August 15th, 2008 at 8:46 pm
I have loaded ubuntu on maybe 10 or 12 pc that i scrapped together. since feb 07 that i get from friends,i load and give away, but i setup everyone with my internet address throuh a linksys router or straight throuh my dsl it is not many but i try. tom
August 15th, 2008 at 8:50 pm
I only been with linux since 2/07 and just then started to learn thanks tom randolph
August 15th, 2008 at 9:00 pm
I love *nix, but honestly, who cares how many people use Linux? Enough people use it for it to be considered as a release target for an application/game. That’s all that matters.
August 15th, 2008 at 10:50 pm
“Conclusion:
We only can count the computers that have pre-installed an OS.”
Not true. I bought an XP computer, then put Linux on it (which, unfortunately, still gave money to Microsoft). Others have done the same.
You’ll find a very, very insignificant amount of computers sold with Linux pre-installed, more often than not.
August 16th, 2008 at 3:10 pm
actually, the percentage of linux users is 100%.
not on the desktop, granted, but a lot of the processing is moving from the desktop to the web. “everybody” uses google, so “everybody” uses linux.
August 16th, 2008 at 9:22 pm
@ Nathan Samson
Google keeps cookies in your browser. It’s their business to know who’s who & what they like. Most seriously, that is the core of their business. I’ve been astounded by the precision of some of their targeted advertising. I’ve searched at home for a new embedded box to run Linux, come in to work & had a box with the exact specs I was looking for pop up in the google ad words (I was on a windows box). Google know I run Linux at home.
I find it interesting that they’re backing an ‘open’ (-ish?) Linux smart phone platform. I’ve been _searching_ for one for years.
August 17th, 2008 at 10:05 am
*nix #’s would be incorrect if u rely on google. Many users run proxies, hide-ip, tor, etc.
August 17th, 2008 at 10:37 am
I read through quite a few of these reviews and I saw some flaws in them - 1. Ask Google - comparing those who use google to those who dont - I think the former will still turn out higher. My mom uses computer, but they hard go to google. They ask me which site is good for what - and they visit just a few of those. There are, I am sure, too many people like them. 2. Counting number of PCs connecting to update servers - there are too many countries in the world where net speed is not good and many people find it expensive to connect using dial ups. They use the computer in whatever condition they receive it. So a large fraction will go unnoticed there as well. 3. Counting the pre-installed boxes - Not possible - too many people wipe off their hard drives (especially the vista ones) and put linux - so did I. 4 Counting the downloads - I download tons of distros and dont use them for more than 1 day sometimes. Morever I download the same ones more than ones sometimes. Sometimes I lend my CD to whole group of people. However, I can still bet on this method - I just get the feeling that the number of copies not used can even out the number of copies used multiple times. I am not sure though. Some statistics Ph. D. can answer this better I guess.
In short it may not be possible ever to determine how many people use Linux. Hey does anyone know how many people use Pirated Windows XP???
August 17th, 2008 at 4:58 pm
Well, Net Applications in no way claims to represent reality. They only claim to show “trends”. I blogged about it, and for marketshare, you can basically discount Net Applications as invalid. Nobody has addressed my questions, Net Applications is silent, and more and more evidence leads me to believe they are an op in market of primarily Windows shops and sites measuring Windows apps thereof.
I wrote this blog to instantly counter those who always say “Linux is 0.8%”. It never fails to quiet them down.
http://mostly-linux.blogspot.com/2008/07/why-is-net-applications-gold-standard.html
My feeling is, with China and Red Flag Linux going leaps and bounds (but not measurable for obvious reasons), I would probably put Linux at 3% min, 8% max.
August 17th, 2008 at 11:54 pm
Linux Counter has an application to download too. It doesn’t prevent falsification (not that there’s much incentive, I believe) because you can duplicate the info that the program sends.
August 18th, 2008 at 2:10 am
Counting IP won’t work because of dynamic IPs
August 18th, 2008 at 9:29 am
Get the telefone, call 1000 people, ask them if they
use linux or what they use. You will have your answer
with less than 2% error. That’s how normal market
statistics are made, we don’t need guesses neither
huge technology.
August 18th, 2008 at 3:19 pm
I believe that the .8% that is so often quoted was the desktop market in ONLY the U.S. in 2005. Servers and other device were not included. Nobody has done a serous study since thene.
August 18th, 2008 at 3:57 pm
“Ask google, everyone uses google, and browsers leave their version and running os in the logs…”
Well I’m running firefox on linux, but I changed my user agent plenty long ago to stop retardedly coded pages from freaking out “server says: you run lin-what? meh…giving up now”
August 18th, 2008 at 6:45 pm
Total number of Computers - Windows - Mac = Linux (Approximately)
August 18th, 2008 at 6:54 pm
What does it matter how may are out. What matters is all the big players, ISV and Server/Desktop Manufactures all sell it now. 5-10 years ago nobody even knew what linux or opensource was. Now its a billion dollar industry.
Let the dollar be the barometer.
August 18th, 2008 at 9:23 pm
Leandro - The problem is that the average person probably does not know what an “operating system” is. I imagine a typical conversation might go like this:
- Hello. I am calling to ask you what operating system you use?
- What? What’s an operating system? I don’t use an operating system.
- Well, do you use a computer?
- Hey! Are you a telemarketer?
- No.. *click*
I guess you could just assume anyone who doesn’t know what an OS is uses Windows, but that might not be a correct assumption.
August 19th, 2008 at 4:36 am
The linux global pc share is not as important as specific shares of specific countries and markets.
Apple nearly disapeared in global numbers but succeded to resist because it maintained a share in EEUU and in multimedia manipulation markets.
Those markets are detected by analizing website logs. So the point is, is there any market segment now that starts to show numbers bigger than 3% on Linux users? Any country?
August 20th, 2008 at 8:31 pm
Macedonia, who’s whole school infrastructure (500,000 computers) switched to Ubuntu last year… definitely should hit a nice number heh
August 21st, 2008 at 10:34 am
http://royal.pingdom.com/?p=340
August 25th, 2008 at 4:43 pm
Dude, you need to educate yourself before you write stuff. Heres a suggestion, read about user agent strings, then go look at some apache logs. Then come back and write your article again, then perhaps you will be able to identify the magical application that _everyone_ runs on mac, windows and linux, that leaves a log message on every site it visits, in fact its probably the _most popular application_ on all three.
For extra points you can give a breakdown of what OS visitors to your site are running.
Newsflash! if you look at the webserver logs on a major site or aggregated logs of many sites you will have a _good_ snapshot of what os people are running. All the counter examples are _statistically insignificant_, that computer you heard about that’s running linux and not connected to the internet? Its insignificant.
August 25th, 2008 at 6:45 pm
Don - Actually, I am very well aware of that method. Unfortunately, it is highly flawed, since almost every website it more likely to be visited by users of one OS or another.
For example, if you take the logs from my website (which I have published sometimes, see “Site News”), you would find that somewhere around 40% of the world uses Linux. If you take any tech-related site you are likely to find an inflated number of Linux and Mac users. If you choose a cooking site, you are probably going to get mostly non-tech savvy users, so the Windows number will probably be inflated. The only websites you can really use are sites such as Google and Yahoo, who are not very likely to give up their numbers.
Even if you had numbers from Google or Yahoo, there is the problem that different search engines are used in different countries.
Additionally, user agents can -easily- be faked and many Linux users do fake them, since some websites are more friendly to browsers they believe are IE on Windows.
August 26th, 2008 at 11:35 am
A few points: -
1) One HP marketing officer from India said that 40% of Desktops they sale in India run Linux. I expressed my surprised and also when I said that HP does not sale Desktops with pre-installed Linux, he replied that the computers sold without OS are counted as Linux ones.
2) OEM’s install Linux into their PC’s when asked to (in India), although they do not openly disclose it. Govt of Assam, a state of India, gives PC’s installed with Linux to about 14000 students who pass the High School Leaving Certificate Exam in first Div every year. OEM’s install Linux into these PC’s. At least once it was Acer.
3) The number of people, mostly young, taking, using and showing expertise in Linux around my place (an Indian city) has increased drastically.
4) When introduced to some of my friends, adopted Linux easily although had fears earlier. Many said it is user friendly.
5) From the above observations and also the fact that Indian IT requirements are considerably different than that of US I believe that the percentage of user in India is likely to be more than 10%
6) Number of pirated softwares are perceived to have declined (in India). In govt institutions it is probably not there (in significant numbers).
7) As suggested by other, some form of scientific study (statistical, market survey, with common application usage, distro download etc factored in with due weightage) may give a much better picture and if the study is carried on for future may improve in accuracy.
August 29th, 2008 at 10:51 am
Well, we may not be able to exactly nail the total number of Linux users (I, myself, have been a dedicated Mandrake/Mandriva user since ‘97) but we can nail the exact amount of internet traffic and bandiwdth they account for. All we would need is a large search engine like Google to take a look at their logs. Not that they would do that. But even a “spoofed” OS (Like Konquerer saying it’s IE ) can been seen through java/flash apps that are specifically desgned to see such things (download the latest flash upgrade lately?).
Whoever says market share numbers aren’t important don’t know much about the business of business. Sure “most” of linux is open source and subject to the GPL, But a lot of apps, more and more everyday, are proprietary in nature and cost a nice chunk of money. If I’m writing software to make money, I’d sure like to know who my target consumer is.
November 17th, 2008 at 5:55 pm
I agree that Market Share is very important to businesses that are considering writing software for Linux. Why bother spending time and money on building software if hardly anyone can use it. If Market Share is high and reported so, then more companies will focus on Linux Development.