The Frontiers of Platform Adoption

In the last two weeks we received two more data points which allow an update to the “race to a billion” platform growth trajectories. Android reached 130 million active users and iOS reached 200 million.

The updated picture looks like this:

Note again that this is a log scale graph. Every major horizontal gridline is an order of magnitude (10x) larger than the one below. It’s a busy graph. The linear version follows:

For the detail-minded, it makes for some interesting comparisons, but I want to create a more compelling visualization. One where each platform can be judged for potential and impact at a glance.

To that end I came up with something I call the platform “adoption frontier” view. Based loosely on the Pareto efficiency concept the chart is reduced to show the latest known figure of (users,time) and overlay concentric arcs centered on 10 million/10 years and radii of unit years (x-axis).

Note that the time axis is reversed from the chart above. This was to allow the best performers to be shown at the upper-right of the chart.

The way to read this is as follows:

  • Each platform is represented by a point which shows its currently known peak in users and time to reach that number of users.[1]
  • The arcs (frontiers) represent possible performance classes.
  • The lowest frontier spans “four years to reach 10 million” to “10 years to reach 200 million”. The highest frontier spans “<1 year to reach 10 million users” to “10 years to reach 1 billion users”. Each frontier can be read in a similar way.
  • The main assumption is that a platform can reach more users but it takes time. Better performance is when a platform moves toward a frontier further from the origin.[2]
  • I chose the limits of the chart specifically: 10 years is roughly the limit of most platforms[3] in the current cycle time of technology disruption; One billion users is an upper bound set by Windows.

One thing you can read from the chart is to say that “from a growth point of view, Blackberry is weaker than any of the other platforms.” Sitting below the lowest frontier with fewer users than Android which has been in the market for less than a third of the time it seems to be less impressive. It is, however, in a similar band of growth as Xbox 360, which, by some measures, is a success story. So the performance standard is relative.

One can also see the rough equivalence in growth between iPod and Symbian, both having crossed the second frontier. But iOS and Android are in a different league. They sit alone beyond the fourth frontier. The fifth frontier is the “billion user in a decade” potential and it seems within reach for both.

What this view also offers is an answer to the question of competitiveness. Although platforms can co-exist and don’t necessarily overlap, the question of becoming overwhelmed with “good enough” by widespread low end alternatives looms for the specialized platforms. For example, game consoles look very vulnerable because they simply do not have the potential to cross high frontiers and orders of magnitude of casual gamers (with potential TV connected devices) might orphan the consoles.

I’ve also included some platforms that have peaked and faded (AOL and i-Mode and Netscape) as a warning. The frontiers illustrate of how hard it is to reach the upper limits of growth. Each level is exponentially more difficult than the last and achieving it with paying customers is a remarkable story of value creation.


I add below the frontier chart with a linear vertical axis. Note that the frontier lines are not equivalent to the frontier arcs in terms of coverage.


  1. The data I have is for platforms where users have to pay something to participate. I exclude platforms where users are the only merchandise being sold (i.e. social networks or email provision).
  2. I only include the initial ramps not upgrades. Windows is anomalous because it is so old. During its first decade (1985 to 1995) it reached over 17 million users but placing it on this chart does not do it justice given the growth occurred in its second decade.
  3. See Symbian and Windows Mobile. However again, Windows which is nearly 30 years old, is an exception.


  • expandin

    Once a again, what a brilliant way of presenting data.

  • Is there any specific reason why you chose the arcs that you did? Why didn't you use straight lines when the acceleration needed to stay in a frontier zone is already implied by the log scale? Is it just because drawing it this way puts iOS and Android in their own class?

    • I guess it's because as the platforms approach 10 years old, they plateau.

    • asymco

      I chose arcs because they are more intuitive as equidistant from an origin. But the categorization is arbitrary. You can also use quadrants.

  • James

    Illuminating analysis as always Horace. Begs the question, what would / will Apple look like financially as a billion-user company?

  • I was listening (belatedly) to Marco Arment talk to Dan Benjamin about the state of tablets (in "Build and Analyze #31: Google Percent 2B") and given Marco's justifiably disparaging comments about the non-iPad contenders, i.e. "Why would anyone buy Tablet X instead of an iPad?", I wondered if we can assume:
    a) that many many people will buy tablets in the near future; and
    b) the vast majority of current iPhone users will opt for an iPad (instead of a competitor)

    If so what does that mean for iOS sales moving forward? It basically goes back to the ecosystem lock-in that Apple is creating.
    I don't foresee many iPhone users choosing an Android or other tablet. It would be interesting to figure out how many non-iPhone smartphone users (mainly Android it seems) opt for an iPad when they buy a tablet, and what effect that has on their subsequent smartphone purchase.

    It may be that the "halo effect" of the iPhone will be even more important in the post-PC world with respect to tablet purchases than it was with Macs.

  • Eric D.

    Apple is currently making a huge profit on its hardware, but inevitably these products will become commoditized and margins will shrink. Seeing this chart, I'm beginning to understand that Apple's real long-term goal may be to tie a few billion $150/year users into its moated ecosystem. This ties in with the recent Deutsche Bank report of iOs being "sticky" because users are more committed to a system they've already spend an average of $100 for apps.

    • Addicted44

      The flaw with this argument is that despite their margins, competitors can't compete on price. Let's look at non-subsidized markets (iPod touch and iPad) and there is no competitor that can undercut them (e.g. Rim is selling play books below cost)

    • asymco

      That's another way of saying that Apple sees iOS as a subscription service. One with a high level of loyalty. I believe that this is very likely the case.

    • Hamranhansenhansen

      Apple already commoditized its own hardware. That's why iPad undercuts everyone else in pricing. So your doomsday scenario is moot. The hardware is essentially free with your purchase of Apple software (Mac OS, iOS) and services (Apple Store, AppleCare, Software Update.) If you bought Snow Leopard on DVD, you were not paying for the plastic optical disc, you were paying for the software. Same is true today with an iPhone … the part you hold in your hand is just a transport medium for the software you are actually buying. That's why users want to know "does it run Angry Birds?" instead of "how much RAM does it have?" The amount of RAM in an iPhone is as unimportant as whether a DVD of Mac OS is a single-layer or dual-layer DVD. Who cares? Less than 1% of the user base.

      Will Apple software become commoditized? Not as long as Apple stays a full 10 years ahead of all of their competitors. Nobody else even has a native C API on a mobile yet … Android is rocking Java apps like a phone from 2005, which continues to hurt it as a developer platform and that will only get even worse on tablets, which are expected to run PC-class native C apps. Nobody but Apple and Microsoft are drawing their interfaces in the GPU yet, and Apple is way ahead of Microsoft. Like iTunes during the iPod years, Apple has simply built way, way, way more software infrastructure than anybody else. It's hard to make software, and very hard to make it quickly. Look at TouchPad and PlayBook, under development for a year at HP and RIM, respectively, after years of development work on WebOS and QNX before that, and still the devices have shipped with software bugs that have shocked reviewers. Look at Mac OS X: 5 years of development at Apple between 1997-2001 after years of Mac OS and NeXT development before that, and still, it took 2-3 years after release for Mac OS X to prove itself worthy just to succeed Mac OS Classic. Look at Microsoft taking 7 years to make Windows Vista, which sucked.

      So Apple is really, really safe going forward. They are on a road all by themselves. Their competitors have to make transitions away from being hardware companies to being software and services, and then they have to build software that is as good as Apple's, and deploy that software, and get people to actually use it, and get developers to develop for it. Nobody is even close. We are going to watch the generic tech industry die off in great numbers over the next couple of years because things have changed so radically over the past few years, but these companies haven't.

      • berult

        iOS is end-user centric. It stands ideologically neutral in enabling people's agendas, …in all deference.

        Google's set of axiomatic principles, embodied in 'self-servers' mathematical constructs, neuters unbridled agendas into a pattern of centrally dispatched afterthoughts, …in all defiance.

        In theory, software rules; mind over matter in the end….

        In practice, 'good enough' gets the nod. Who …apart from me of course… wouldn't match up with the clichéd blonde with the big boobs, wouldn't buy brand detergent at Walmart, fish life out of a pond, or vote 'nay' for the sheer disemboweling sound of it…? Grey matters, except when it does less …doesn't it?

      • Eric D.

        It appears that I don't understand the word commoditized. Still, I remember back in the '90s my employer leasing fully-loaded Macs (IIfx, Quadras, etc.) for upwards of $40K (monitor and RAM included). Today's iMacs (which are terrific workstations) go from a low $1,200 to top-of-line $3,000 because of wider user base, mass production and economies of scale. This process will continue.

        To further broaden its user base, Apple will no doubt one day offer $150 iPads and iPhones. These will serve as gateway devices into the Apple ecosphere. Look at the dividends that Apple is reaping from having seeded all those iPods to kids over the years. In upper education, the switch to Macs as the students' computer of choice has been rapid and dramatic. Steve Jobs underlined this strategy when he introduced the iPad: "If you have an iPhone, you already know how to use it."

        In the long run, the hardware is likely to produce thinner margins (vis the iPos), but profits will grow on the software side. (Which is your point too, Hamranhansenhansen, if I understand you correctly.) It's the old give away the razors and sell the blades strategy which we see driving today's inkjet market. Except that apps, stable operating systems and clouds are far more advanced and will keep evolving long after the technology curve has started to plateau. Yes, technology will keep advancing, but ultimately it will reach the "good enough" stage that toasters and microwave ovens have achieved. There will always be cutting edge offerings tech offerings, but it's software arena that holds the promise of sky-is-the-limit advances. Stable platforms are going to serve as enablers to an exploding range of applications.

        My view is that this will work in Apple's favor, because Apple is very strong in software innovation. Unlike companies beholden unto their legacy users cough-Microsoft-cough, Apple is willing to displease loyal customers in order to move the ball forward and reach broader markets. The furor over Final Cut Pro X is a perfect example: lose maybe 5000 pro users to Adobe and Avid; gain 5 million new aspiring filmmakers.

        Hardly a doomsday scenario, right?

      • In Econ 101 — tho perhaps not in the sense used here by some — a commodity is an item available from multiple sellers with no differentiation in quality. You will buy from to the person who will sell it to you for 1¢ less per ton.

        Shares of AAPL are commoditized in that sense: not only do you not care who's selling, you never find out. Only for large insider sales or purchases is identification necessary. Your 100 shares could have been the first ones issued or numbers 900,000,001 thru 900,000,100. They're all the same.

        So a commodity Android smartphone is identical regardless of whether it says "HTC" or "Moto" underneath the "Droid" label. For at least most purposes, it will have the same CPU, run the same software and hook you up pretty much exactly the same to Verizon.

        Just as no two snowflakes or kernels of corn are exactly identical, this definition of “commodity” takes some tolerance for poetic license.

      • asymco

        In Christensen's theory a commodity is a product for which improvements are not longer valued, in other words, over-serving.

  • Alan

    What determines whether the points on the first two plots have lines connecting them? Android, for example, has several connected points, then a non-connected point, then two connected points. iPod ends with a single disconnected point.

    • Lack of Data. Horace doesn't fudge data. If it's not available, he doesn't include a guess to make smoother or continuous lines.

    • asymco

      Points are connected when they are sequential. In other words if data exists in Q1 and Q2 and Q4 then Q1 and Q2 are connected but Q4 is not since Q3 is missing.

  • George Bailey

    A bit off topic, but I suggest using something like the amazingly handy "sharpshooter" so that your images get more meaningful names than "Screen shot 2011-07-16 at 7-16-10.53.14 PM"

  • This shows the criticality of iOS 5 and iCloud.

    If you look at the plot, we see that iTunes accounts are about 270 million accounts. This number puts an upper reach on iOS adoption since iOS versions below 5.0 are intrinsically tied to iTunes accounts and iTunes.

    With iOS 5.0, Apple is allowing the cutting of the cord so you can actually use an iOS device 100% free of iTunes. You will still have the iTunes account but it will be thought of as an iCloud account. This will allow future growth of the platform outside the iTunes "eco-system."

    I have little doubt that Apple recognizes this as well.

    • Hamranhansenhansen

      Although there are 270 million iTunes accounts, the iTunes app for Mac/PC is installed on something more like 750 million computers. I don't think iTunes was holding them back, it was PC's that were holding them back. If you didn't own a Mac/PC where you can run iTunes, you were a second class citizen on iOS before iCloud. Now, Apple is free to sell directly to mobile users who do not even have a PC, which is most of them.

    • Lava


      I believe you're wrong. You still need an Apple ID (and thus an iTunes account) to activate an iOS device. The difference is that you can now do it over the air without using a cable to a computer running iTunes.

      If anything this will accelerate adoption of iTunes because it reduces the friction needed to participate in the ecosystem.

  • Engineer

    I’m concerned that these numbers may not be accurate. Android phones are sold with BOGO model, replaced every couple months and the “activations” numbers are absurd.

    There’s always a delta between sales and installed base… Do you have any way to get at the real installed base? Or is my original iPhone still being counted as it rests out its final years in the garage?

    I know you have to work with the data you have but I think things are going to get distorted going forward.

    • Even in a BOGO(F) deal, the person who gets the handset for 'free' ends up paying for it subsequently through the contract they sign up for. Still a sale.

  • westech

    Here's a strange one. The graph with the arcs does not show up on my 27 inch iMac, but does on my iPad2.

  • westech

    Here's a strange one. The graph with the arcs does not show up on my 27 inch iMac

  • westech

    This is totally weird. It shows up fine on my 15 inch MacBook Pro. Anybody got any ideas?

    • asymco

      It could be the cache is loading a different page. Try accessing the graph image directly:

      • westech

        Thanks. A computer restart solved the problem.

        Great analysis. Wish you could post two points for each item: current and a year ago, with a line connecting them. Would show the recent rate of growth.

  • sscutchen

    I wonder about the affect of tragectory on the position in the chart. For example, Blackberry was a pioneer in smartphones, and perhaps they've gone through more of an S-curve of adoption that iOS or Android. Perhaps their start was much slower simply because the genre was not as pervasive. iOS and Android were introduced into a more mature market, and could grow quicker from the start. This would have the effect of shifting their time line to the left, and dropping Blackberries to a lower frontier.

  • Horace, thanks for the innovative ways of looking at the data.

    Is it your understanding that once a market becomes sufficiently large, it fragments? For instance, there is NOT a single market for “automobiles” everywhere I'm aware of; some are simply partitioned into chauffeur-driven versus private, others, like the US, into sporty/sedan/minivan/…

    Like your deodorant example a couple of weeks back, shoppers go looking for a particular item within that market. An iPhone might simply not be compared to an Android device; competition/choice between an iPad and an ASUS tablet might never happen. The question is largely how quickly each category grows, and when growth slows in maturity, profitability within a category.

    If you have retail / tech experience, it'd be interesting to know your thoughts about how hard-and-fast the boundaries are between categories; how disruptions in one (say, smartphones) bleed into another (feature phones); many more topics that we might ponder along with you.

  • Horace, thanks for the brilliant visualizations. As always, very insightful.

    2 modest suggestions:
    – Have you considered adding the communication platforms (email such as Hotmail/Gmail/Yahoo Mail, social networks such as Facebook/MySpace/Twitter, and now Google Plus, VoIP services such as Skype) on the chart as well?
    – Have you tried grouping the series by year in some ways, and looking at the relative success within "classes" (or cohorts)? One hypothesis would be that the speed at which top performers in each "class" are able to reach X million is increasing, in part thanks to the existence of previous platforms as a way to distribute the new ones. If you can show a correlation / trend (a la Moore's law?), it would be very interesting to know what the "expected best-in-class speed to X million" would be in the next few years.

    I imagine your databases are proprietary/paid, but if not, is there a place where we can play with (the public parts of) the dataset? I would be happy to make some visualizations and share them as well.

    Keep up the great analysis!

    • Hamranhansenhansen

      Hotmail/Gmail/Yahoo Mail are not platforms … they are just email clients. They are all part of the email platform. Facebook/MySpace/Twitter are not platforms … they are just Web apps. They are part of the World Wide Web platform.

    • From the Notes section:

      The data I have is for platforms where users have to pay something to participate.

      I agree that comparing this data to Facebook or Twitter would be very interesting. At the same time, those platforms are very different and comparing free-platforms to paid-platforms is like comparing apples to oranges.

  • anonymous

    Horace, I hadn’t heard of you or this site until your first episode of The Critical Path. The way you go about presenting data visually and how you’re able to string sentences together and express your thoughts audibly is remarkable. I’ve listened to all 3 episodes multiple times and it’s still mesmerizing. Thank you.

  • Symbian seems to be missing from the linear platform adoption ramps btw. It'd probably skew the graphs of course!

  • EWPellegrino

    It's also interesting to imagine how iPad would look on those graphs in its own right. With only 4 quarters of sales reported it's already at 19.5million users. As best as I can make it out that puts it substantially ahead of either iOS or Android at that point.

    I think it's worth asking whether tablets are actually part of the same platform as smart-phones or really a completely new platform that happens to runs a closely related OS. Arguably the two device families will follow their own independent adoption curves, much as Linux as a server platform has a very different adoption profile to Linux as a desktop platform.

    It's a minor quibble but currently the y-axis is labelled 'millions of users' but as we mix tablets and handsets together for iOS and Android those users will increasingly be double counted as they own multiple devices. Great visualizations nevertheless.

  • John

    thanks for showing these beautiful graphs. In comparing ramps from the 90's to ramps in 2011 it might be useful to scale with some notion of market size at the time or at the start of ramp, rather than an absolute number. Can one use "% of total cell phone users" to compare blackberry, android and iphone ramps? ie are their time to 5% or 10% of market ramps different? similar scaling could be done for AOL or netscape, scaling with pc users or game platforms scaling against pc market.

  • I would really like to see how Facebook fits in with this data. Of all the platforms out there, it seem to be at the latest stage of growing to 1 billion users in a decade. After roughly 28 quarters, Facebook has 750 million users and seems to be on track to a billion.

  • Lava


    Question, but if I'm reading the chart correctly, why does it seem like iOS is shown to be 4 quarters (1 year) to reach 200 million users? That seems a little off, doesn't it?

    • Lava

      Sorry, I think I see the mistake I made in reading the chart…I was following the orbit lines, but the chart shows iOS took slightly less than 4 years to hit 200 million.

  • Hi Horace,

    while the visualizations are beautiful, I am a bit sceptical about how meaningful those "frontiers" are. Please let me know your thoughts:

    The first problem about the model is that the horizontal axis will penalize players that have been in the industry before that industry actually exploded. Windows is a good example that is recognized in the article. Windows didn’t take off until the electronic components, processors, developers… tied together to boost the PC market.

    How about BlackBerry? It is shown as being born in 2003, the year where RIM added voice to the device. By that time, the network, standards, batteries, virtual machines… weren’t there for the mobile industry to explode. The model back then didn’t have the same uses and capabilities as of now, so is it fair to say it is the same platform? (shouldn’t then windows mobile be the same platform as windows? Doesn’t wii start with the Nintendo first devices?). Any date chosen as a start would be arbitrary, but undoubtedly the worst option has been taken for Blackberry.

    • asymco

      Of course, older platforms appear to be less powerful as they did not have the same underlying conditions to grow to billions of users. That implies that future platforms might eclipse the current leaders. It's a fact of life and the charts illustrate this phenomenon. One other thing to keep in mind is the implicit finite lifetime of platforms (about 10 years). Technology driven disruption does not let any company rest to reap rewards.

  • (continued)
    *** A second problem of the model is the way the “frontiers” are drawn. As they are, the platforms constantly move from one adoption frontier to another. An easy way to get from graph 1 to graph 3 is to flip graph one vertically and resize. You can easily see the how they move through frontiers: does that mean that they improve through time? Most probably it means the concept of adoption frontier is not very meaningful or reliable (to be fair it is mentioned in the article – “Based loosely on the Pareto efficiency concept” -, but it is surprising how from that “loose” step the article jumps to some conclusions).

    Furthermore, the center of the frontiers (44 quarters, 10 million users) is absolutely arbitrary, and it determines between what frontiers each platform falls…

  • @cnsh

    Can you control for US market penetration? Or at least weight for it? The closed nature of US cell phone markets is probably why Symbian (alas, now defunct, thanks Microsoft!) has relatively slow growth – it's mostly a non-US market.

  • bossjet

    I took the first platform adoption frontiers chart and applied it to social networks in the US.

  • Pingback: The ‘Age of Apple’ will end | memeburn()

  • Pingback: SEInsider – We share knowledge bringing to you news, phone news, apps, reviews, tips, android, rumors, products, services and the like of Sony Ericsson World » Blog Archive » Android not as open as previously thought()