Corroborating the Android Platform Dashboards

Written by: on September 5

Should we really trust that the platform numbers published by Google represent the distribution of users in our applications?

[UPDATE 5-9-2014]: We’ve posted our 2014 data figures here.

We are fortunate to have a monthly release of Platform Dashboards by Google, detailing the breakdown, by platform version, screen size, etc., of users visiting the Google Play Store on their devices.  But perhaps we should not take this information at face value and use it as the best determination of what we developers should support in our applications.

To test this theory, we have chosen applications from three of our partners and analyzed their application-specific user data to determine if the numbers we see truly align with the graphs everyone is familiar with from Google.  All of the following application data collected represents installs as of 7/31/2013.

For context, here are Google’s Android Platform Dashboard numbers from that same timeframe (collection period ending 8/1/2013).

Version Codename API Distribution
1.6 Donut 4 0.10%
2.1 Eclair 7 1.20%
2.2 Froyo 8 2.50%
2.3 – 2.3.2 Gingerbread 9 0.10%
2.3.3 – 2.3.7 10 33.00%
3.2 Honeycomb 13 0.10%
4.0.3 – 4.0.4 ICS 15 22.50%
4.1.x Jelly Bean 16 34.00%
4.2.x 17 6.50%

Perhaps due to this effort by Google, few Android developers tend to release information about how the usage of their own applications breaks down for the same metrics.  Contrast this with Android’s largest competitor, iOS, where the issue is almost completely reversed.  Within the iOS ecosystem, Apple rarely provides distribution numbers of their platform except, perhaps, at the annual Worldwide Developers Conference (WWDC).  In response to this, many more iOS developers are releasing the analytics information they have collected about their own applications to help other developers make educated decisions about version support.

One problem with this approach is that information provided by companies like Apple and Google represents, as closely as possible, the statistics of the whole, while information from developers is representative of a more targeted subsection.  Even within those large-scale numbers, one company takes statistics from active store engagement while the other (we can only assume) uses numbers on raw activations… and yet, somehow, many of us take all of these elements to be equal and comparable directly.

I think we at Double Encore can provide some value through our diverse experience working with top-tier applications on the Android platform to help fill the hole that exists in the reported data available from shipping applications.  We hope that seeing Android platform usage statistics from real applications will help you make more informed decisions about the platform.

Data Collection

This data is direct from the Play Store Developer Console, not a third party analytics tool, so the data is coming from the same source.  We have aggregated the user information from these three applications to form our own snapshot of the ecosystem.  All of these applications have obtained 100,000-500,000 or more downloads, and each is in a separate store category, so we believe this data represents a good enough user sample to be considered relevant and valid. We will be displaying the following metrics:

  1. User breakdown by country (Where do the users live?)
  2. User breakdown by manufacturer/device (What hardware do our users own?)
  3. User breakdown by platform version (Are the users on an updated device?)

The numbers collected and published here are the “active user install” numbers from Google Play, which track unique accounts that are using the application on at least one “active” device and not individual device count.  How Google defines a device as active versus inactive isn’t really public, but, in general, it is intended to weed out devices that, for example, have been thrown away and haven’t accessed Google Play in a very long time.  This does not equate to active usage of the given application, but simply to active device use and an application still installed.  If you want more information about how Google defines these terms, have a look here.

Top Countries

First, we want to see where most of our users are coming from, as this will have a distinct impact on the devices they use.

While the top three countries are the United States, Canada, and the United Kingdom, the sum of these three countries only makes up 72.6% of the user base in this survey.  All ten countries discretely listed here held at least 1% of the user base, but the top 90% of users came from 25 different countries.

Top Devices

A primary concern for any Android developer is the effort required to test their application to ensure it runs well on the devices used in the market.  The tech press likes to popularize the tens of thousands of Android devices that exist in the wild as a developer problem, but what is considerably more relevant is the set of devices on which your application actually ends up.  First let’s look at the device manufacturers breakdown.

Samsung has the majority share by a wide margin here, and over 85% of the most-used devices come from just five manufacturers: Samsung, Motorola, HTC, Google (which includes the Nexus line), and LG.  This proportion follows with the country breakdown of users.  If our user base skewed more heavily to Europe or Asia, for example, we would probably see manufacturers like Sony or Huawei with a strong presence in the survey.  Let’s zoom in and take a look at these numbers by unique device.

This graph discretely calls out the 25 most popular devices of our aggregated users, which make up almost two-thirds of the active users.  The Samsung Galaxy line of devices alone (S, Tab, Note) commands 58% those top devices, with the Galaxy S III taking the top spot.  Overall, there were 2500 unique device signatures for the aggregated active user base, with the top 90% of users represented by 142 unique devices types.

It’s true, that’s a lot of devices; how on earth do we catch them all?  At DE, we focus on the fact that a significant portion of these users come from the same device family.  We also rely on the fact that, in general, devices made by the same company tend to behave the same under a given set of conditions, so we try to make sure we have testing coverage over the major manufacturers of the devices our users own, rather than worrying ourselves with testing on each one individually.

Users by Platform

Finally, we want to see what version of Android our users have on their devices.  This is, of course, the metric most of us use to determine what versions of the platform to support in our applications.

You can see that, in our survey, we have a distinctly smaller number of users running Gingerbread (2.3) and below.  In fact, that number is less than half of what Google publishes.

Concluding Thoughts

There is a movement amongst many Android developers to push applications to all support Android 4.0 as their minimum version; this is often accompanied by the tagline, minSdkVersion=”14″, which represents the manifest value an application would set to require this (and for which even bumper stickers have been made).

While our numbers for platform versions below that may be smaller here than what Google indicates for the market as a whole, Gingerbread use is still a significant portion in our apps as well.  Android 4.0 (Ice Cream Sandwich) was introduced by Google two years ago this October, so here in the U.S., as contracts expire and people go for their next device upgrade, perhaps we will see the number of active Gingerbread devices reduced significantly in the not too distant future; I would even predict that by the first quarter of 2014 this landscape will look very different.

However, one could definitely say that there is data to support the argument that the users of your application are significantly more up to date with the platform than Google’s published numbers may lead us to believe.  In many cases, this is a justifiable reason for raising the minimum Android platform version to a level that could greatly increase the performance, enhance the experience, and reduce the complexity of your application (the support library is a great thing, but that’s a lot of code to drag into your APK if you don’t need it).

Let’s keep this conversation going.  If you’re so inclined to post your own application’s data as well, let us know in the comments or send us a link via Twitter, Facebook, Google+, or Carrier Pigeon!

Dave Smith

Dave Smith is a Senior Engineer at Double Encore, Inc., a leading mobile development company. Dave is an expert in developing mobile applications that integrate with custom hardware and devices. His recent focus lies mainly in integrating the Android platform with embedded SoC hardware. He is a published author and speaks regularly at conferences on topics related to Android development. You can follow Dave on Twitter and Google+.

Article


Add your voice to the discussion: