Updated Android version charts

Back in December I posted some thoughts on Google’s regularly updated Android developer dashboard, which provides data on the adoption of various flavors of Android as well as screen sizes and densities in use by devices hitting the Google Play store. As we now have several more months’ data to look at, I thought I’d update the charts from that post and revisit some of the trends I saw in evidence back then to see if they still hold.

Android versions in use

Here’s the overview of adoption rates for recent versions of Android, grouped by dessert name: Recent Android versions in use

Let’s test some of the observations I made last time around:

Major versions (i.e. those grouped together by dessert name) tend to take about 12-18 months from launch to hit their peak, usually at around 60-70% of the base

This seems to be holding true with Jelly Bean, which now appears to be starting to slow down in the low 60s about 19 months after launch in July 2012.

By this point, at least one and sometimes two new versions have been released, though recently often only one newer version.

At this point, we only have one newer major version – KitKat, though of course there were two sub-versions of Jelly Bean after the launch of the original version.

It is now typical for there to be at most three major versions at over 10% of the base, and on occasion just two. This means that developers can target a large chunk of the base by focusing on just two or three versions.

This is still the case – Gingerbread, Ice Cream Sandwich and Jelly Bean are the only versions over 10%.

Gingerbread represents both an outlier and a legacy which is unlikely to be repeated. It was a major release, installed on many devices which were not capable of supporting any of the subsequent versions, and as a result has had an unusually long lifecycle. Most other versions have dropped well below half their peaks a year later, and represent a very small share two years later. But Gingerbread breaks this pattern, hanging on as the second most used version three years after launch and 18 months after its peak. Both Google and its OEM partners have learned from Gingerbread and once it works its way out of the system the pattern going forward will look quite different.

Gingerbread continues to hang on at 19% of the total over the past week. The rest of the comments are longer-term in nature, so we’ll have to see how they bear up in the coming months, though I’m confident they continue to be on the mark.

Most versions peak and begin to decline about 3-6 months after the next version is released, reflecting the time it still takes most OEMs and carriers to approve a release for distribution to end users.

KitKat was released in October 2013, about 5 months ago. It looks like Jelly Bean will peak in the next month or two. So this looks on target again.

Screen sizes and densities

Here’s the chart for screen sizes. As a reminder, these are my interpretations of a somewhat vague chart from Google.

Android screen sizes corrected

Let’s revisit the earlier observations on screen size:

The “Normal” screen size remains massively dominant… Though it has fallen from 100% in 2009, it still represents around 80% of all Android devices hitting the Play store today. Given that almost all new high-end Android devices today have screen sizes in the high 4 inches range and often 5 inches or larger, this is a useful reminder that these devices are not representative of the Android base as a whole.

The “normal” screen size remains stubbornly at just under 80% and actually rose as a percentage of the total over the last couple of months.

Interestingly, the next most popular category is not larger phones but smaller ones, at around 3.5 inches and under.

This also held up in the last few months, as “small” screen sizes have held steady at just over 8%, while “large” screen sizes spiked from 6.9% in December to 8.4% in January but have since dropped back down to 7.7% over the next two months.

The two larger sizes, Large and Xlarge, represented 12% of the total as of the last figures, a significant increase from earlier years (they represented just over 4% together in 2011). The cutoff between the two categories is imprecise, at around 7 inches, but it seems likely that most Large devices are bigger smartphones, including most of the high-end devices such as the Galaxy S 4 and HTC One, while most Xlarge devices are tablets, both in the 7- and 10-inch ranges. Large devices actually represent a bigger chunk of the total – 6.9% – than Xlarge devices (4.9%).

The Xlarge bucket has shrunk back down under 5% this month after briefly popping over that mark in January and February. By contrast, iPads represent about 25-30% of iOS devices in use today. Of course, Google’s stats ignore Amazon’s Kindle Fire line as well as all the tablets sold in China which can’t access Google services. But it’s a reminder of how small the overall Google-supported Android tablets base is.

On to screen densities:

Android screen densities

Let’s revisit those conclusions:

Two categories have fallen somewhat rapidly over the period shown, namely the medium (130-190dpi) and high (190-270dpi) slices.

The high slice has now stabilized and remains the largest category. Together, the low, medium and high categories, which are all under Apple’s iPhone retina display density, make up two thirds of the base and that number doesn’t seem to be moving much. In fact, what’s striking about the chart above over the last nine months or so is how stable it is overall, with no lines moving obviously up or down and the relative rankings remaining exactly the same.

The decline was caused partly by the growth in the awkwardly named xhdpi and xxhdpi categories which have emerged over the past year or so, and which have suddenly risen to 22.3% and 9% respectively, together representing about a third of the total. These devices have displays of over 270 or so pixels per inch, so these are high-quality displays, many of which would qualify as retina or better by Apple’s definitions.

Again, the proportion of the total made up by these devices has actually stabilized, suggesting that although they are becoming more prevalent in mature markets, growth elsewhere in the world is offsetting that shift.

However, it’s also worth noting that the low dpi category has grown over time, having bumped along at 1 or 2% for a long time, but now rising to 9-11%. This suggests that there is a category of low-density devices which is emerging, which is actually about twice the size of the tablet category, so not insignificant. In the matrix Google provides, we can see clearly that 90% or so of these align with those small screens, so these are likely very low-end Android handsets sold in emerging markets.

The low dpi category has stabilized at just over 9% for the past several months, again just under twice the size of the Xlarge screen category.

At the same time, it’s interesting to see in that same matrix (see below) that  90% of the tablet-sized devices have only medium density screens of around 130-190dpi. Samsung’s latest 7-inch tablet has a 299dpi screen, and the latest Nexus 7 and Nexus 10 have 323dpi and 300dpi screens respectively, so these high-end, high-density devices are clearly in the minority in the Android tablet installed base, in marked contrast to the smaller devices.

This also hasn’t changed dramatically in the last few months – 4.4% out of the 4.9% Xlarge category are still using low or medium dpi screens, with just 0.3% using hdpi and 0.2% using xhdpi.