2010-09-10

Counter-Debunking the 1% Myth

Caitlyn Martin of O'Reilly Broadcast has another interesting article about why the figure of Linux market share is quite a bit more than the oft-quoted 1%. She starts out by doing a bit of math: (1 Linux netbook)/(3 total netbooks) * (18 total netbooks)/(100 (desktops + laptops + netbooks)) = (6 Linux netbooks)/(100 (desktops + laptops + netbooks)) — Linux netbook sales alone constitute 6% of the total desktop market. I can't argue with that. It's also a really impressive number; the number for total Linux desktop sales (that includes desktops, laptops, and netbooks) will obviously be higher — I don't know by how much, but the total number is certainly at least 6%. I think it's great that Linux desktop sales have come this far.
The bigger question regards the meaning of this number, and I believe the problems with this article occur before and after the calculations. For one, the author starts by saying that Linux market share in servers and embedded systems is significant. I don't think anyone ever doubted this, and several studies have shown that Linux distributions are in fact in the majority on these types of hardware. Therefore, I think it isn't quite right to start the article like this when the title clearly indicates a discussion about Linux desktop market share. That said, this is a fairly minor issue.
The article's argument gets more murky when Steve Ballmer is brought into the discussion. It has been written repeatedly that the presentation referenced clearly shows that the biggest threat to Microsoft Windows is illegally copied versions of Microsoft Windows (not Linux, Mac OS X, or BSD). Why? This is the case in many countries with poorer populations, as people simply can't afford to buy a licensed copy of Microsoft Windows. Furthermore, Linux adoption in these countries is low because the computers sold there often don't have the tools (Internet connection, working USB port, working disk writer drive) to create a live CD/USB, and there's no cost advantage to using Linux as people will just copy Microsoft Windows. A lot of commenters in the linked article also point out that in these countries where unlicensed use of Microsoft Windows is rampant, it is cheaper, in fact, to buy a machine with Linux preinstalled and then wipe the hard drive and install an unlicensed copy of Microsoft Windows. I don't doubt the verity of this. That said, the 6% figure is for the US, where unlicensed Microsoft Windows use is likely to be much lower; therein lies another problem, as statistics for US Linux market share is being compared to statistics for worldwide Microsoft Windows market share.
And why should Steve Ballmer know about precise statistics for Linux and Mac OS X usage? It certainly is easy to look at sales statistics and see what percentage of computers are sold with licensed copies of Microsoft Windows; also, as Microsoft "phones home" to see which copies of Windows are licensed and which are not, it should be pretty easy to tell how many unlicensed copies of Windows are being used. While one can safely conclude that the rest of the pie is occupied by Linux, BSD, and Mac OS X, it is not so clear how much each contender has of that slice of the pie. I guess one way of telling how many people use Mac OS X is by tracking sales of Microsoft Office for Mac, but this is problematic as many Mac OS X users stay away from Microsoft Office and instead use iWork, OpenOffice.org/NeoOffice, or a web-based productivity suite.
So why would Steve Ballmer put Linux market share at or above Mac OS X market share? Microsoft knows that it can compete with Apple's Mac OS X in terms of features (in some ways) and price, so Mac OS X isn't really considered a huge threat. On the other hand, Linux can do most of the things a typical computer user wants (e.g. web browsing, productivity, peripheral support) for free, which does technically pose a serious threat to Microsoft Windows. At the same time, Microsoft has repeatedly accused the free software community (and Linux in particular) of infringing upon its patents. Therefore, Microsoft is pegging Linux market share that high in order to justify those accusations — if Linux market share is that high, then of course Linux needs to be stopped before such patent infringement gets out of control! (This is what I would imagine a Microsoft executive saying at such a presentation.)
I won't argue that Linux desktop sales make up at least 6% of total desktop sales. Usage is a different matter entirely and should not be treated the same as sales figures.

32 comments:

  1. > Usage is a different matter entirely and should not be treated the same as sales figures.

    Unfortunately that cuts both ways. I've never bought a copy of Linux pre-installed on a PC, but instead just downloaded it for free and installed it myself. I'd guess that's how the majority of Linux systems come into being.

    I won't take either side of the 1% debate, because at this point in time it is an unknowable number.

    ReplyDelete
  2. I have fourteen machines at home- 6 servers, 5 laptops, 3 desktops. (Hardware only- have some running VMs, too.)

    All but one of these had OEM installations of Windows.

    2 servers are still Windows Server. (Only out of necessity, believe me.)

    One laptop is still Windows. (Same reason.)

    No desktops are still Windows.

    All distros were downloaded and installed manually, except one- Asus EeePC.

    So in terms of strict sales figures, my part of the example pool would be skewed.

    I do think that the NetApp-and-similar ratings systems are too easily given great credence, and the challenges to them are too easily dismissed. It's not insignificant that a majority of the sites that are rated actually pay to be counted. That does have an impact on figures.

    And at risk of sounding too idealogical about it, Steve Ballmer was right- Linux is a cancer. But not in the way he meant. It's a growing mass of people who are able to run it how they want it. As much as MS may hate that.

    ReplyDelete
  3. I keep asking the folks that's saying 1%. What's your ± %?

    How can you have a percentage at 1% without any margin of error or standard deviations. Those missing figures alone put it (well) over 1%.

    FYI: You CAN NOT have 1% as a hard percentage figure without any hard numbers to transpose. Any guesstimated figure will have a margin of error. Hence above 1%.

    ReplyDelete
  4. What was the point of this article? To tell us that YOU know S. Ballmer's motive? And to believe you?

    Believe you because ???

    You had no facts or figures. Just more hot air. Not helpful at all to the discussion.

    Even a waste of electrons.

    PS. Thanks Linux Today for your Linux traffic Today.

    ReplyDelete
  5. Hello Lady and Gentlemen:

    Now -- You see why Microsoft office is a waste. Case in point. He wrote this in MS Word and then cut and paste it here. Now, the $299 package with all the bells and whistle, can't maintain format outside of itself.

    PS. Dude. Learn HTML or something, that was (very) hard to read. They taught you better in school. I'm sure your teacher wouldn't except that -- but yet you want to challenge a profession writer on the web with this?

    ReplyDelete
  6. Thank you all very much for your comments. I'm sorry that this was not made clearer in the post itself: I am not trying to say that I know better than anyone else or that I have better data than the original article, just that the numbers in the original article themselves are flawed or questionable in some way or another (even the author herself calls some of them little more than "educated guesswork").
    @John and Anonymous 2: Yes, these are true statements. The sales figures seem accurate, but the usage statistic, while higher than 1% for sure (due to the whole margin of error thing), is probably not as high as the 8% posited in the original article.
    @Anonymous 1: Thanks for the information. I can certainly see how usage figures in many sites are skewed by the sites being paid to track such statistics. That said, I think Wikimedia pages' (e.g. Wikipedia) usage statistics are fairly good indicators of the usage reality, as I don't think users of one platform are more likely to use Wikipedia than users of another platform. I think those statistics peg Linux usage around 3-4%.
    @Anonymous 3: No, I don't ask that you believe me right off the bat. I also don't claim to know Steve Ballmer's motives; I am simply speculating why Microsoft could be quoting such high Linux usage figures, and speculation isn't the same as knowledge. If you have legitimate criticisms of this post, please lay them out here, and I will be more than happy to do a follow-up post with these corrections and comments taken into account. And yes, I have stated multiple times (and will state again) that I am very grateful for all of the readers coming from Linux Today.
    @Anonymous 4 (are you by any chance the same as Anonymous 3?): I do not own Microsoft Office. I type all of my posts in Blogger's built in WYSIWYG editor. The question I have for you is this: do you take issue with my writing style, formatting, or content? Please let me know, as I would appreciate any and all suggestions for improving my writing. That said, I feel like you're mixing up content issues with formatting issues. Please let me know which one is the problem.
    @Hans Bezemer: I do reference the figures Steve Ballmer gave in the talk referenced in the original article. In any case, the point I am trying to make is that the numbers leading to such high Linux usage statistics are probably flawed.
    Please understand that I would love for Linux usage statistics to be as high as 8-9%, but as long as we (as the Linux community) continue to quote such high figures for the sake of combating the very low figure of 1% that is commonly quoted, we will not be taken seriously as a desktop OS community. I'm not trying to bash Linux; I'm trying (obviously not especially well) to set the record straight (a little).

    ReplyDelete
  7. @PV:

    Sorry for assuming you used MS Word. Haven said that, I see it way to often...

    I had no problem with your writing style. Heck I can't write, admittedly, but I do try to make it legible (even in my comments).

    Why aren't you taking the time to hit return (twice)? Hence paragraph(s).

    Presentation is everything when you want to be taken seriously.

    PS. Yeah, yeah, I know about assumptions. Hence your article.

    ReplyDelete
  8. @PV:

    PV says:
    "... Wikipedia ... peg Linux usage around 3-4%."

    Okay. I'll buy that. Now what's the site ±% margin of error? That's all I'm asking. Because they do also record unknown OSes, ya know. What do you do about those? Disregard those as non-Linux?

    OK. So now add your ±%. We know Linux is not over represented so it's not under 1%. So what number did you come up with?
    A: It has to be greater than 1%.

    Let's say I have a co-worker (Windows user) that visits Wiki 10 times a day (That's what windows users do).
    I (Linux user w/ two VMs of Linux) may visit Wiki once a week.

    That's 10 to 1. And if I have two Windows co-workers, that can register as 20 to 1. But in actuality is 2 to 2 OSes.

    Now you see why web stats figures are flawed.

    Thanks for your reply

    ReplyDelete
  9. @Anonymous 1: Double line breaks is not a habit I have developed in my writing, but if you and others find that it'll make the content much more readable, then I'll try to make it a habit. Thanks for the input.
    @Anonymous 2: Your hypothetical example is valid, but I'm not sure that's exactly how it'll be counted especially if the site is viewing IP addresses (so the same computer shouldn't be double-counted on multiple visits). It'll still be flawed, to be sure; I think the only thing we can take away is that usage of Linux is more than 1% but less than 10%, and that more accurate numbers (or methods) cannot be found yet.

    ReplyDelete
  10. @PV

    PV Says:

    " ... Linux is more than 1% but less than 10%".

    I can buy that.

    Thanks for your reply

    ReplyDelete
  11. @Anonymous: Thank you for your comment!

    ReplyDelete
  12. PV Says:

    " ... Linux is more than 1% but less than 10%".

    That would make sense, for North America.

    Brazil, for example, is a very passionate user of GNU/Linux (I've heard around 40-50%) and ~89% of the top 500 super computers in the world run it as well.

    So depending on the demograph being observed, "Linux" usage is so high in variance that it, by its very nature, becomes difficult to measure.

    I, for one, believe that the entire world contains about 20-30% of GNU/Linux users, though I have nothing to back it up.

    I am quite grateful for your efforts in uncovering actual hard facts about total usage. Thank you.

    ReplyDelete
  13. My South Africa relative said they never heard of Ubuntu not any of their friends use linux. among the hundred of computer users I known, none of them used linux in their daily work.

    My first experience with linux (RH 7)is wifi card not working. I could not watch DVD. Even until now (Ubuntu 10.04) after installing linux, I still need great effort to download codec to play mp3, dvd etc (which is a standard usage for notebooks). How can someone less PC savvy or not good internet access ever try linux?

    1% may well over exaggerate.

    ReplyDelete
  14. @anonymous: use linuxmint!

    ReplyDelete
  15. @Dann: I mentioned at the beginning of the article that server, supercomputer, and embedded Linux usage should not be mixed in with desktop Linux usage. Also, I'm not trying to hurt you personally (so please don't interpret this as such), but putting out numbers without evidence one way or the other isn't much better than some of the usage statistics presented in the original article. I honestly haven't done anything in terms of finding better statistics; I'm simply trying to reason out why some of the existing statistics may be flawed.
    @Anonymous 1: The fact that your South African relatives have never heard of Ubuntu is surprising, considering that Mark Shuttleworth is South African and Canonical is based in South Africa. I guess they're so focused on the US, UK, and EU markets that they haven't put much thought into making their presence known in their home country. I'm guessing another reason for this is that in many poorer countries like South Africa, as I stated in my post (that others in the comments section of the original article pointed out), the biggest threat to licensed Microsoft Windows use is unlicensed Microsoft Windows use, not Apple Mac OS X or Linux/BSD use. Also, with regard to your codec troubles, as Anonymous 2 points out, please do try Linux Mint, as it supports almost all hardware drivers (including my NVidia graphics card) and proprietary codecs (e.g. Flash, libdvdcss (the codec to play encrypted DVDs)) out-of-the-box. It should greatly reduce your troubles, and recently a US court ruled that software like libdvdcss is not illegal for the sole purpose of playing encrypted DVDs (as long as it's not used for actual copyright infringement). Finally, if you look at some of the above commenters, it's clear that the usage statistic is higher than 1% (despite, in all frankness, what your anecdotes may lead you to believe (and I'm not doubting the veracity of these anecdotes)) as 1% falls within the margin of error (+/- 3%), so it can't possibly be a concrete number.
    @Anonymous 2: True that! I've used it every day since 2009 May 1, and I've loved pretty much every minute of it!
    Thank you all so much for your comments!

    ReplyDelete
  16. All i can tell you honestly is that 75% of my family is using Linux.

    Me : only Linux
    My wife : only Linux
    Daugher, 21 : Windows (silly girl, i warned her)
    Son, 20 : 50% Linux - 50% Windows (He promised to browse internet mostly by Ubuntu)
    Daughter, 18 - only Linux
    Son, 16 - only Linux

    ReplyDelete
  17. My guess:

    Linux in USA and UK - more than 1%, less than 2%
    Linux globally: more than 3%, less than 6%

    Not including smartphones. You know more than 200 000 new Android-users every day.

    ReplyDelete
  18. @Antton: It's great that almost your entire family is using Linux. (If you want, tell your Microsoft Windows-using daughter that she has to either pay for any software repairs out of her own pocket or she has to learn how to fix them herself. I figure eventually she'll gravitate towards Linux, unless there are some applications that have no substitutes in Linux, but for the typical user, these are few and far between.)
    I also purposefully only focused on the desktop market; I know that the Android market is exploding with new customers. That said, Linux usage even in the US has to be at least 3% (but probably not significantly more than that) as less than that is within the margin of error.
    Thanks for your comments!

    ReplyDelete
  19. It amazes me that the global usage of Linux remains a hotly debated question. And every article I've read that tries to settle the issue, either for or against the 1% figure, is full of conjecture.

    Aren't we all wasting our time that could be better spent elsewhere?

    ReplyDelete
  20. I believe the prime source of the 1% MYTH was Microsoft, as revealed in the combs v Microsoft case file:
    http://antitrust.slated.org/www.iowaconsumercase.org/011607/9000/PX09445.pdf

    It is what MS titled the "Linux Heat Map". It shows that between August and October of 2003 the US "MidMarket" organizations already using Linux as their PRIMAY OS was 1%, and that 12% were evaluating Linux, and that 8.2% were planning on changing to Linux.

    Interestingly, it explains why MS made such an effort in Brazil. In 2003, 3.4% of Brazil's MidMarket organizations were ALREADY using Linux as their PRIMARY OS, 28% were planing to change, and 18% were evaluating or deploying Linux. Russia, India, China, France and other countries were already showing almost equally as large numbers.

    The Combs v Microsoft trial was over before Steve Ballmer gave his Feb, 2009 talk in which he showed a graphic giving Linux at least 10%, right up there with Apple at 10%. http://www.osnews.com/story/21035/Ballmer_Linux_Bigger_Competitor_than_Apple. ALL of the 1% Myth preachers ignore or conveniently explain away his graphic, or even call him a liar!

    Four months later NetApplications, which had obediently published web statistics that purported to show that the Linux desktop was only 1% (SIX years after the "Linux Heat Map" -- can anything remain that static that long in the computer business, especially after the failure of VISTA and the slow uptake of Win7?), SUDDENLY cut Apple's percentage in half! http://tech.fortune.cnn.com/2009/08/02/net-applications-apple-just-lost-half-its-market-share

    Now why would NetApplications suddenly want to "add weight" to China's desktop usage? Could it be that Mac's price tag is too big for most of the Chinese market, AND, that a LOT of China's computers are running PIRATED copies of Windows? Just another way to tilt the playing field ... cook the statistical books.

    And, it goes without saying that nearly EVERY computer running Linux came with Windows pre-installed and is either dual booting with Windows or has REPLACED Windows, a reversal which never appears in any bean counters metrics.

    Microsoft wasn't lying to themselves with their "Linux Heat Map". In 2003 ZDNET cited IDC research that revealed that the Linux desktop market share was already at 3.2% and that it would reach 6% by 2007. http://www.zdnet.com/blog/itfacts/linux-desktop-market-share-to-reach-6-in-2007/5334 The 2009 Ballmer talk which presented the Linux desktop market share at 10% is right in line with those previous figures.

    Finally, IF Linux was REALLY at a measly 1% of the desktop market share, and OBVIOUSLY had to have remained there, UNCHANGING, for the last 7 years, (despite the IDC report), it boggles the mind that Microsoft would waste so much money, energy, PR and DIRTY TRICKS fighting such trivial competition, or that DELL would offer 3% of its laptops and 6% of its servers with Ubuntu pre-installed.

    ReplyDelete
  21. I think that other factors are at work. I had to pay the windows tax, for what used to be my machine (Upgraded now) and had the store wipe the disk (no OS)

    That was a 'windows' sale, even though no windows was technically sold.

    I am running Ubuntu, and have installed it on a total of 4 of my computers (all would have been listed as 'windows' computers. Some friends had windows computers, but with all the no security of windows they are Ubuntu Users. that is 10 of your windows computers not running windows... But, who is counting?? You are, it seems.

    Offer me a Fully rigged Windows 7 pro with All the software Microsoft could load, paid for...

    I would have Ubuntu 10.04.1 and 10.10 both installed in less than an hour, including time to fully format and wipe the garbage.

    You have to count much better than you seem to. There are more than one percent out here, but we don't tend to rely on the 'dining at the Redmond table' reporters for information or patches but never fixes...

    Speaking of which. Buffer overflow and JavaScript and ALL of Microsoft's major software are being patched Tuesday... On one page I read about how many bugs are being patched, and in same column reading of how many are still coming up new, and know of many that have yet to be patch after an extended period of time. (Years sometimes). That doesn't happen in Ubuntu/Linux. Because the Community works to stop it, not eat at Microsoft's table saying the problems don't exist.

    I was rather good at making Windows secure, and tuning it up. When I dropped it for Ubuntu, one of my standard tweaks still worked. Allowing for 32 bit processes instead of 16 bit processes which was how (and may still be) Windows installed.

    I am running 64 Bit processes (dual core - native) in Ubuntu..

    ReplyDelete
  22. @Barista Uno, GreyGeek77, and dataodo: Let me make myself clear: I do not claim to have better numbers than the ones that exist. What I will say is that neither the 1% nor the 10% figures are flawless, each for their own reasons; the actual figure is probably between 3-10%, but we probably are wasting our time trying to figure out where exactly it lies. Thank you all for your comments!

    ReplyDelete
  23. GreyGeek: Ballmer's pie chart does NOT show Linux anywhere near 10%. If Linux had 10%, and Apple another 10%, together they would make for almost one quarter of the pie. They're clearly much less than that. My visual estimate is that Ballmer's chart shows Apple at 4% and Linux at 5%.

    ReplyDelete
  24. @Onan the Barbarian: Very true, and that's what other articles dealing with that slide have said as well. Thanks for the comment!

    ReplyDelete
  25. A number of comments, in random order:

    How does Ballmer know? Well, to begin with Microsoft must have a pretty fair grasp on how many IPs stop updating Windows entirely, vs how many switch to an "upgraded" version, etc. I'm sure their are other means available (lost enterprise deployments, commissioned surveys/market research, etc) that MS can rely on.

    The Wikimedia number is a little under 2% --based on sampling their own logs, so that number must be taken as accurate. The question is "why so low?". The honesty of Wikimedia's reports is not in question, but considering various other methods of estimating Linux usage, there appears to be some sort of inadvertent bias. For example, one possible candidates is here http://stats.wikimedia.org/archive/squid_reports/2010-06/SquidReportOrigins.htm
    (Per project language / subproject) 43% of requests are english-language (more than 5 times the next most popular language); since the MS influence is known to be especially pronounced in anglo countries, this is probably a significant factor -- there are probably other biases, if one can only root them out.

    The Dell netbook numbers can be reasonably considered a lower limit, because Dell only makes pre-loaded Linux an option on their bottom-tier systems. Trying to upgrade the Linux system hardware to the specifications of Dell's Windows offerings results in a system that costs significantly more than buying the very same hardware with Windows -- and often the options to upgrade the hardware are not even made available in the Linux product line (eg. upgrading a netbook to a 6-cell battery is just not even an option). Would-be purchasers have to accept material short-comings, plus go to extra effort to even find Dell's Linux systems, in order to acquire a pre-loaded Linux Dell system (I've heard stories of buyers arguing with sales-reps on the phone to get the hardware available for Windows systems, on their Linux purchase -- with mixed success). Meanwhile, Dell's advertising material (I just leafed through an 8-page insert this morning) never even mentions Linux as an option. But the sales figures are still over 30%!

    I have no idea what the correct figure for Linux desktop usage is. In light of my above remarks and the previous discussion, I would put the lower credible limit firmly at 2%, assume the actual figure is around 4-6%, and would not be astounded if it's double that (global, not N.America).

    Higher than 12% would astound me -- this year. Inertia and legacy systems will keep the numbers down for some time, until some critical threshold is reached, and the artificial restraints imposed on Linux popularity crumble. Whether desktop Linux ever surpasses 50% will depend on economic factors and human behavioural factors I don't pretend to understand -- but it should gain general recognition as a "standard" desktop OS, at least as unremarkable as Apple OS X.

    ReplyDelete
  26. @Bernard Swiss: That sounds like an entirely reasonable assertion, and thanks for your comment. That said, I would appreciate it if you didn't try posting the same comment 4 times. Please just keep that in mind when you post comments in the future. Thanks!

    ReplyDelete
  27. Sorry about the multiple posts -- I kept getting messages that my comment had failed to post.

    I tried posting with the option for LiveJournal profile, Name/URL and Anonymous -- the failed post messages were different for each option, but alleged failure for each one. Nor did my post show up even if I refreshed the page. In the future I shall know to disregard those "failed post" messages.

    I'm glad you found my remarks of interest. It would be nice if we could somehow get more trustworthy figures for desktop Linux "market share" and "market penetration", but unfortunately not only is this inherently difficult to work out, but there is a significant amount of deliberate obfuscation by powerful vested interests with motivation to discount and even suppress Linux as a credible contender.

    Even supposedly disinterested third-parties have been caught tilting the playing field (eg, certain "bricks and mortar", "big-box" outlets, some OEMs) so there are problems not only in reporting good numbers, but also in how well those number reflect actual interest or demand (You might want to review this topic separately?).

    Of course, how biased the "market" is, is largely a separate question from what the correct numbers are for market share and penetration.

    ReplyDelete
  28. Addendum:
    my last post went through just fine, the very first time -- I have no idea what the problem might have been.

    ReplyDelete
  29. @Bernard Swiss: Yes, when web hit counters are paid (usually by Microsoft or companies that support Microsoft) to calculate OS usage statistics, the playing field certainly does tilt towards Microsoft. On the other hand, sites like FSDaily, DistroWatch, TechDirt, and others will report Linux usage statistics that are higher than average simply because of who the audience is; similarly, Apple's website will be visited mostly by Apple users (Quicktime and iTunes notwithstanding). Thanks for the comment!

    ReplyDelete
  30. The 6% number is based on ABI Research which looked at the worldwide market, not the U.S. market. When you add desktops and laptops I believe the correct number is 8-10% worldwide.

    Usage? In developed countries all you can find at retailers is Windows. The common action is to wipe Windows and install Linux. I am aware that the developing world may do things differently in many places, wiping FreeDOS or Linux and installing pirated Windows. There are a lot more computers sold in the developed world than in the developing world. Usage, if anything, is significantly higher than sales figures.

    ReplyDelete
  31. @Caitlyn: Are you Caitlyn Martin, author of the original article? Sweet! (If not, I have nothing against you.) That aside, don't count out China, with thrice the population and computers that are super-cheap. This is a country where a lot of buying computers with Linux and replacing them with Windows goes on.

    ReplyDelete