Last Updated: April 30th, 2012

Pop quiz: How long does it take for a new version of Android to be widely adopted? A new version of Android comes out, AOSP updates, OEMs adapt it to a myriad of devices, and carriers test the updates. That process. How long does it take?

It's a tough question to answer, mostly because Google doesn't provide data like that. The official site shows a 6 month version history, and that's it. Anyone looking for a decent amount of data is out of luck. There’s no way to view the long journey older Android versions have taken, and no way to see the bigger picture of how the update process eventually works out.

Until now.

After a good amount of digging and some pointed swear words aimed in the Wayback Machine’s direction, I’ve compiled a complete history of the last two years of Android updates.

This isn't just the old Google charts poorly stitched together in Photoshop (that was version one), this is a seriously accurate piece of chart-work made from the actual data used to make the official history graphs. With a little Google Chart API data-extraction magic, you can get numbers accurate to 1 tenth of a percentage point in 15 day increments. In other words, we're going for "crazy accurate."

Ok, enough exposition, I present to you the definitive history of the last two years of Android updates, which shall henceforth be known as "The Big Android Chart":


Neat huh? Towards the bottom you’ll notice pixel-accurate tick marks for SDK releases and AOSP drops. Two years of data is the oldest you're going to get short of an official Android history section (please?).

The biggest thing that sticks out as me is just how poorly Android tablets have been doing. The Xoom came out February 24, 2011, right around the time of the 3.0 SDK release, and Honeycomb flat lined for the next 7 months. It didn’t even register a pixel on the chart until the holiday season of that year.

One strange aberration I can’t quite figure out: What the heck happened at the top of the graph from 5/01/11 to 11/15/11? There’s a gap that isn't claimed by any version number. The official Google charts look the same way, so don't blame the chart. How did this happen? Poorly identified custom ROMs? Google ICS testing? I have no idea. Any commenters want to take a swing at an explanation?

One more fun chart. Android platform support aligned by months after SDK release:


That's right, it’s not your imagination, Ice Cream Sandwich adoption is going very, very slowly. You’ll notice update percentage gets progressively slower with each new version, but keep in mind the Android ecosystem is also getting progressively larger. Ice Cream Sandwich has to deal with many, many more models than Éclair did.

Our more astute readers will have noticed that Honeycomb is missing from this one. This chart is meant to be a measurement of upgrade ubiquity, and most devices (phones) are not eligible for a Honeycomb upgrade, so it doesn’t belong here. It’s also important to realize this is chart is about “Platform Support,” not “Version Adoption.” Version adoption would be that individual version only, platform support is a combination of that version and all versions after it. So Éclair being at 97.8% doesn’t mean almost all phones run Éclair, it means almost all phones run Éclair or greater. An install of Froyo also counts for Éclair.

So there you have it, Android history data that isn't available anywhere else. It's been a long road, hasn't it? I'll be keeping the massive spreadsheet generated from this and updating it every few months or so. Here's hoping that ICS line takes a huge upswing by the time that next update rolls around.

There's a ton of data to take in, I'd love to hear your observations in the comments.

Ron Amadeo
Ron loves everything related to technology, design, and Google. He always wants to talk about "the big picture" and what's next for Android, and he's not afraid to get knee-deep in an APK for some details. Expect a good eye for detail, lots of research, and some lamenting about how something isn't designed well enough.
  • Nathan Wong

    Great work to put this together!

    I think it's worth noting that because we're looking at the % of installed units rather than absolute numbers, that later releases will always be much harder pushed to see a wider rollout simply because the rollout will inherently be to a much larger installed base of a much larger variety of devices. Eclair / Froyo adoption was really quick because there were only a half dozen existing devices on the market, that could (almost) all be updated, and it was at this time that Android had a huge surge in popularity.

  • Pburns

    Great Work! I've always been curious, I've been using Android since 1.5 on the Tmobile HTC G1. I've been running CM since his first release. :) Thank You!

  • carnegie0107

    Awesome. I've used every single AOSP release, starting with Android 1.1 on the ADP1/Dream/G1. Remember when there was no on-screen keyboard? No auto-rotation? No video playback? Only 3 widgets, and no ability to add more? How far we've come...

    • http://dmonzel.com dmonzel

      Oh yeah, those were the days. And remember not being able to save pictures from MMS messages? And I remember rubbing it in my iFriends' faces when I flashed CM3. It's great to see how much we've grown up.

    • http://www.theandroidsite.com benmarvin

      ...And before there were paid apps in the Market

      • Aeires

        When you could actually browse the entire market in a few hours.

    • Jeff B

      I remember when the HTC magic Android 1.6 rom leaked. I haven't looked back on custom roms since then. I used CM from day 1.

  • Peter

    ICS should look a  lot better when the next numbers come out. It is at about 12% for my app. 5 fold increase over the last month. Then again the ICS roleout in the UK sounds a lot better than the US so who knows.

    • Fifarunnerr

      Same, it's 13% on my app(Students in The Netherlands only).
      Global, it reports that 7% of the Android devices who are using Productivity apps are running ICS.

  • http://meatcastle.com/ Youre My Boy Bloo

    Haha, this is reminiscent of my last company when it came to computers. We were upgrading our DHS and we needed to make sure everything onsite had windows XP or greater. When running the diagnostics I began finding all kinds of crazy things. There were a half dozen Win 2000s smattered around our plant floor, and even an old Windows 95 machine locked in a closet running homebrew software to manage our phone pre recorded message for our PBX.

    Those were dark days.

  • Jeffrey Feely

    Is the weird dip caused by 2.3.4 - 2.3.7 not being represented?

    • Ron Amadeo

      "2.3.3" is actually "2.3.3 - 2.3.7" I'll update the chart.

  • Ruben Alvarez

    Looks like fragmentation is getting much better. It is neat to see donut and cupcake. Have not thought about that in a while. Miss the old G1 and MyTouch. I was so amazed with cupcake and to look at ICS it just makes me shake my head. Sweet info.

  • http://twitter.com/zk0sn1 Nee Austin

    I'm curious as to how semi-retired phones should be counted.  These aren't actively using voice services, but still use wifi data and apps.  They are kind of off the grid from a practical sense.  

    My OG Droid is still plugging away on 2.2.   It's disconnected from Verizon but it still hits the market, and probably always will until apps stop supporting 2.2 and it's given a full honors burial (in the green bin at work).

    • Tim

      I have an OG Droid that I just retired for the Razr MAXX.  Did you do anything special to your OG Droid to disconnect it from Verizon?  I would love to use my OG Droid as a music box to connect to a set of speakers or some other task I haven't thought of yet.

      Have WIFI available in my home at all times, I'm just afraid that when I turn the phone on it will attempt to connect to VZW's network and cause problems with my Razr.

      • Yggugkuyg

        If your razor is working and connected, the your droid isn't. Turning on your old droid won't do anything because it would have to be reactivated by dialing a special number. Don't worry. Just turn it on and use it.

        • Tim

          Thanks for the reply.  I had a feeling this was the case, but wanted to confirm it prior to trying anything.  Very much appreciate the reply!

      • http://twitter.com/zk0sn1 Nee Austin

        Wifi works fine.  No 3G once it was disconnected from VzW. 

        Sorry that wasn't clear.

  • rj5555

    You state HC flat lined... while true one should keep in mind that the graph is in % of the total and that in the "flat rate period" the total number of Android devices in use exploded, so the few % on the right are a lot more devices than the same few % on the left

  • Tom Stieger

    Can we get the first graph in number of active devices rather then percents, so that we can see the total Android growth as well?

  • http://profiles.google.com/pbooker117 Phillip Booker

    I would love to see this list represented by "flagship device" updates. How many, and how quickly were supposed "flagship" or "high-end" devices updated? Anyone with Sprint remember the wait to go from 1.5 to 2.1 on our HTC Hero?

  • daengbo

    Android tablets have done poorly, but most of the early ones came with 2.3.

    • http://codytoombs.wordpress.com/ Cody Toombs

      It was only the OG Galaxy Tab and B&N Nook that came with 2.3 before 3.0 hit.  The Nook wasn't official so only a percent were rooted and given market access (I'm suspect there were some off-brand devices too, but they didn't get market access either).  It was after Honeycomb came out that a ton of no-name companies started pushing out sub-$250 tablets running 2.3, and most of them didn't have market access either.  That's kind of the key detail, if it didn't have market access, it won't show up on this graph (a few rooted devices are exceptions, but not enough to change the percentages).

  • guard her

    Well just got my hands on a jellybean test build for my SIII. Haaaa ha

  • Michael Arcement II

    Well we should see an adoption spike over the summer months of ICS as the new phones roll out. There already is a small uptick at the tail for ICS

    • Josh Brown

      That's really the big concern I'm having.  There's always a point where there's a sharp up-tick, and adoption goes pretty quickly from there.  Gingerbread's wasn't as sharp, but it wasn't as big of an upgrade (from a user's standpoint) from Froyo, so adoption was more linear.  What disturbs me is that the time until up-tick is getting much longer.  This can't be explained by the larger pool of devices, and could be indicative of a much more serious problem.

      • http://codytoombs.wordpress.com/ Cody Toombs

        I think two things are to blame.  The first, as you and others mentioned, a lot more devices require updates thanks to the gadget vomit from the biggest 4-5 manufacturers of last year.

        The second, is how many devices are being abandoned by manufacturers that people still have on contract.  We aren't going to see that drastic upswing until contracts start running out on people who bought abandoned devices later in their lifecycle.  The Galaxy S series alone probably accounts for about 10% of that list, and since everybody who bought that phone with 2 year contract (the very first was released in June 2010) is still under contract, that single product line ensures there is going to be a much slower rise.  Considering the variants of the HTC Desire HD (not abandoned, but close) and Galaxy S amount to more than 15%, and possibly as high as 20%, of the active Android phones out there, there's just no way for a massive upswing like there was in earlier versions.

  • http://twitter.com/squidlr squidlr

    ICS is notoriously difficult for skilled hackers to get working on older devices, never mind the amateurs who work for the handset makers.

  • Jasper

    Is there any chance you can do a version with absolute number scale instead of percentages? With a logarithmic scale, if necessary. I would like to see if that gives any insight to upgrade-a-device versus buy-a-newer-device.

  • colormedisappointed

    If I use the same "trick" that the global climate change alarmists use to extrapolate into the future I see ICS dominating in two months.

  • http://twitter.com/usefulDesign Useful Design

    There's no label on your Y-Axis for the Big Android Chart. Scale of 100 suggests percentages but of what? Downloads, installs, phones shipped with said OS, etc etc?

    • http://www.androidpolice.com/ Artem Russakovskii

      Devices running said OS.

      • Josh Brown

        If you're pulling from the developer site, it's percent of devices that have checked into Android Market/Google Play in the last two weeks.  Inactive devices are not included. Neither are non-gapps devices like the Kindle Fire or the Nook.  This doesn't mean the user has opened Market/Play, just that the device has been able to check in.

    • Ron Amadeo

      It percent, and it's a big version of the chart Google shows, so "Each dataset in the timeline is based on the number of Android devices that accessed Google Play within a 14-day period ending on the date indicated on the x-axis." In other words, percentage of a active devices.

      I should add that... somewhere.

  • http://twitter.com/davidchu David Chu

    Another data point that might be useful is the number of phones that have ICS available.  For example, the Lucid by LG on Verizon is still on Gingerbread.  I think a reason why Android releases are taking longer to update is because OEMs do not want to spend the money to update a their OS.

  • stevenjklein

    I'd love to see a comparison about how iOS 5 adoption is coming along.  Is it just as bad?

    • http://twitter.com/smitty97 Tom Smith
    • http://www.lazyprogrammers.com Eugene Kim

      I believe iOS adoption rates are measured in days and weeks. 

      • Josh Brown

        Just until they get majority usage.  As far as when they go away, that takes about the same amount of time.  That means most developers still have to worry about legacy devices for about the same time period.  To users this matters, to developers not so much.

        • stevenjklein

          But Gregory Wild-Smith (below) says that iOS 5 hit 82% in just 15 days. Meaning developers can ignore than 18% running older versions of iOS.  And since all current model iOS devices come with 5.1, the percentage using older versions declines daily.

          Believe me, I know. I have a 4-year-old iPod touch that can't run iOS 5, and there are lots of apps that don't support it.

    • http://twitter.com/abritinthebay Gregory Wild-Smith

      Not even close. iOS5 adoption is at 82% (maybe more now) and it took iOS just 15 days to reach adoption levels greater than any one single version of Android.

  • Woody

    Absolutely excellent work.

  • Thoughtlesskyle

    i think the reason tablets looked like they were doing so poorly was that fact that the source was never released to AOSP, those results are based on the version of android accessing the play store. Even with the SDK ports of Honeycomb none of them were functional enough to use as a daily, I don't know what the exact # of users who root and rom their tablets are, but i know the nook color, and the HTC flyer both would have had 3rd party support for honeycomb since they didn't launch with it, and lest we forget how the HP Touchpad would have affected that number with an AOSP release of Honeycomb. just my 2 cents

  • Blow

    the version is largely adopted once people upgrade from their 2 year contracts so every 2 years a substantial bump is seen in the graph!!!

  • http://www.innotechive.com/ K Kool Swapnil

    Heres an infographic of ANDROID versions
    would be useful
    Infographic Of Android Versions

  • Samvith V Rao
  • Samvith V Rao