New Android and iOS "fragmentation" charts are just as flawed as the term

This article may contain personal views and opinion from the author.
New Android and iOS "fragmentation" charts are just as flawed as the term
For quite a long time now, I've been fighting back against the term "fragmentation". I have never argued that problems don't exist, just that the term "fragmentation" has such a broad meaning as to be essentially meaningless. These days, rather than just give up on using the flawed term in favor of more accurate descriptors, it has become more popular to turn the word "fragmentation" back on its original creator, Apple, (I have even done this myself) but this just further muddies the water.

Here's the TL;DR recap of my argument against the word: it misdirects the blame for multiple issues, which each have multiple causes and multiple possible solutions. For example, Android updates are slow to push out mostly because of manufacturer custom software and skins (as the Nexus program and Motorola have proven), not because of a flaw in how Google created Android, and we really can't even blame carriers that much anymore after the Moto X update hit Verizon so quickly. Similarly, compatibility issues are a byproduct of an ecosystem built on such huge variety in choice, not a problem that can be thrown in under a blanket term, because the solutions to that range from better developer tools, more access to hardware, more open manufacturers, etc. 

Inevitably, when a new platform version distribution chart is released, there will be a rush to compare Android and iOS to see where each stands in terms of "fragmentation", but really it's an exercise in comparing which platform has an easier path to upgrades, and showing which software versions developers should target when creating a new app. That is honestly the only salient information that can be drawn from these charts, but that doesn't stop pundits from trying anyway. 

Playing with the data


Yesterday, Forbes used two different platform distribution charts to attempt to compare Android and iOS in terms of "fragmentation". Writer Philip Elmer-Dewitt chose the Android distribution chart that can be found on the Android Wikipedia page (based on Google-supplied data, which is fairly reliable), and the latest Chitika stats for iOS, which show that iOS 7 has been installed on about 74% of North American iOS devices so far. We have to be careful with Chitika stats because they are generated through ad impression data, and have a tendency to vary anywhere from 5-15% from Apple's official numbers. These were the charts shown:


Right off the bat, you may notice that the two charts don't use the same date range, which adds in quite a bit of color to Android's chart because that chart goes all the way back to 2009. Even without thinking too much about the charts, the initial impression that a viewer is to take away is that "fewer colors = better". So, the iOS chart looks great, because it has the big swath of iOS 6, and the sharp penetration of iOS 7. Android looks like a mess; but, if the Android distribution chart were to be remade to only cover the same dates as the iOS chart, it would look quite a bit different. Luckily, a reader did just that and submitted this to Forbes, which even Elmer-Dewitt admits gives a more "apple-to-apple" comparison:


Once you get rid of all those superfluous version changes from the past, the visual impact alone of Android's distribution chart changes quite a bit. It still doesn't have the fast adoption rates of iOS, but fewer colors means things are better, right? With the new layout, there are essentially three main versions dominating the chart, and two of them (Ice Cream Sandwich and Jelly Bean) share more similarities than differences, since they are both incremental updates on Android 4.x. Which gives me another idea: why don't we go one step further and show the Android chart the same way as the iOS chart with all of the incremental updates removed and only the full number versions shown.


The visual impact on this chart is huge. Obviously though, this last chart isn't all that useful, because it doesn't show how Google handles version numbering, but it does show how easy it is to manipulate data to give off the impression that you want, even if it means obscuring the entire point of Android platform charts, which is to show developers where to focus their efforts. Unlike iOS, each incremental update for Android moves up the platform's SDK level, so this last chart hides information that is extremely useful to developers. 

Of course, each of these charts is incomplete in its own way. For example, even the versions of the Android platform distribution that show each nickname version are hiding the incremental updates within each of those levels. Jelly Bean alone spanned Android 4.1, 4.2, and 4.3, each of which brought a new SDK level for developers, and features to users. If you're worried about the security implications of not updating, neither the iOS nor the Android charts give you the incremental update data to consider that problem properly. 

The simple truth is that distribution charts don't tell the whole story anymore.

Platform version distribution is an incomplete metric


Just like the word "fragmentation" is constantly misused, now the charts used to illustrate the supposed problem don't even provide the data that is often claimed. As has been covered plenty of times by now, just because an iOS device gets the latest version of the software doesn't guarantee all of the latest features. No matter what software updates come, the iPhone 4 and iPad 2 will never have Siri; similarly, despite the newest software, the iPhone 4 & 4S and iPad 2 & 3 will not get AirDrop or Camera Filters. These are consumer-facing features that do really matter to users who may want access to Siri, but don't want an iPad mini and don't have the extra cash to drop another $500 on a new iPad Air, which is the only full-size iPad available from Apple that supports Siri since the iPad 2 is still inexplicably holding the lower price point. 

Of course, the other side of that argument is that even if those devices aren't getting all of the features, they are still getting all of the bug fixes and security updates. So, is iOS "fragmented"? Well, it depends on 1) your definition of fragmentation, 2) the metrics you use to define fragmentation, and 3) the tools you use to measure those metrics. You could claim that iOS is fragmented because not all users get all features, but we would say that is just the business of progress, and eventually something gets left behind. Or, you could claim that iOS isn't fragmented, because all users get security updates (assuming they install the various updates that hit their device). But, even that argument ignores the work that developers have to do to build assets for the various iOS display sizes and resolutions, because the system doesn't employ responsive design. You can already see the usefulness of the term "fragmentation" failing. 


When you talk about Android, security updates are often considered the smoking gun to the argument that Android is being crippled by supposed "fragmentation" (although really in this case it is slow/inconsistent updates by manufacturers because of bloated skins and baked-in apps, as Motorola continues to prove.) But, even that argument is starting to disappear as Google is putting fewer and fewer updates in the full system releases and offers more updates via decoupled apps and automatic updates to Google Play services

Two of the biggest security updates to hit Android over the past year were Verified Apps, which scans sideloaded apps for malware; and, the Android Device Manager, which allows you to locate a lost phone, remotely lock the device and change the PIN, or remotely wipe the device. Neither of these updates came in a full system release, but rather in the form of an invisible Google Play services update that went out to 99% of all Google Android devices (if we're not counting non-Google Android devices in the charts above, there's no point in talking about them here). This means that security was improved for all Android users running Android 2.2 and higher without any need to worry about whether or not manufacturers would push out the update. 

There are still security updates to be had in the various system updates of Android; but moving forward, it wouldn't be surprising if more and more of the security and feature updates were part of Google Play services and app updates rather than system updates. The trend has already been moving that way, so it's not hard to imagine it continuing. And if it does, these Android platform distribution charts will lose even more relevance when it comes to using them as part of the "fragmentation" argument, because that argument itself will become even more fragmented than it already is. 

Conclusion


The conclusion may sound redundant at this point, but it is fairly simple: if you want to have a discussion about Android and iOS (and there are plenty of incredibly interesting discussions to be had), think about the issues you want to cover, and break each down on their own terms. Trying to bundle arguments under and umbrella term like "fragmentation" is just lazy and it holds very little meaning at this point. If you want to talk about the speed of software updates, then by all means, take a look at how quickly platform updates are adopted, but be sure to take a look at app updates as well, because that is a big part of that discussion. 

If you want to talk about security risks, that is a plenty valid discussion to have, but platform version charts don't give the whole story there either. For iOS, you have to drill down farther to see how many users bother to install the incremental updates, like iOS 7.0.4, which is the kind of data that you don't see often. And, on the Android side, you have to take into account Google Play services, as well as security issues that have seemingly been fixed by manufacturers

If you want to talk about app compatibility, that's fine too, but there are specific considerations there as well. All of these and more are wonderful discussions, and each has a lot of interesting interpretations to be made. But please, let's all stop believing that the only thing we need to consider is to look at a chart and see "more colors = bad".

FEATURED VIDEO

36 Comments

45. TechieXP1969

Posts: 14967; Member since: Sep 25, 2013

Android has no fragmentation. Most people are using Android 4.x period. There are more people using native 4.x devices vs people using native iOS7 devices. Also, why show device from 2009? Let's show all of Apple devices since 2007. Those devices cant even use iOS7 just like devices from 2009 cant use todays version of Android. As with all these facts, when you remove the fanboi element of thinking, thinks look less blurred.

41. Dadler22

Posts: 243; Member since: Dec 11, 2008

bottom line, doesn't matter. Specs range is the biggest problem for Android devs, and screen resolution/size (non-hd 3.5-4' to super high res 7-10') is the hardest challenge for iOS devs

37. blade19

Posts: 65; Member since: Apr 29, 2011

Always like reading articles by Micheal H. Keep it up. He also wrote another great article that explained fragmentation not too long ago on PA's site.

23. shadowbob

Posts: 1; Member since: Dec 15, 2013

Please do another graphic like Apple with the same dates difference that they used for Android, and put all their updates versions as separation, then compare it to your last Android graphic, once done we can spread the word that iOS is way more fragmented than Android, if they dare say something against it we just tell them that it's the same strategy they used against Android.

33. JakeLee

Posts: 1021; Member since: Nov 02, 2013

What a moronic comment. Fragmentation is all about the eco system. No new sdk ==> minor update, no count.

17. JakeLee

Posts: 1021; Member since: Nov 02, 2013

An interesting article, but many devs won't agree with some points. Each time a new SDK is released with a new version of Android, it's considered a major update with new/enhanced APIs. Just as Michael mentioned, Jelly Bean alone consists of three major updates (4.1-4.3) Is the modified chart ignoring all theses major updates then misleading? No, not in the slightest. And that's exactly the problem. If Apple releases a new version of SDK, the devs look at the release note thoroughly for new/enhanced APIs that might be useful for their current/upcoming apps. It's not the same for Android. The only decision the devs have to make is : Do I target 2 or 4? The modified chart is fully valid considering this. And it's not a good thing. Ironically enough, the modified chart highlights the seriousness of fragmentation instead of airbrushing it. Vast majority of those major updates are almost all in vain since they don't get utilized at all by third party devs. If you ask an Android app dev which version to target : 4.4 : off the table 4.3 : no way 4.2 : nope 4.1 : no 4.0 : Ummm..... I have to ask my boss And an iOS dev's answer would be : 7 : I'd love to, but my boss won't approve yet. Maybe next time 6 : My boss explicitly ordered me to. It's mandatory for official 64-bit support anyway. 5 : Already done. We also did the extra job supporting 64-bit via workarounds upon boss' order. From a dev's point of view, Michael seems to be talking about fragmentation without knowing why and how it matters. Why mention extra features like Siri when talking about fragmentation in first place? For app devs, they mean just as much as the bundled pinball game and/or Aero feature on Windows - nothing. Looking at the chart, it's finally a reasonable decision targeting 4.0 ICS. And if it takes 2+years until a new set of APIs gets finally utilized, it's a problem IMO.

18. nasznjoka

Posts: 418; Member since: Oct 05, 2012

nope i disagree on some points you mentioned, in android there's target, complilation and lowest sdk api. So when you develope a new app it will be compatible with the highest api depending on whether you have updated your sdk or not but if its up to date, when you say target 2.3 but compiling against the highest sdk that you have( if you are up to date) then that app is gonna be compatible with both the targeted and the highest.api

20. JakeLee

Posts: 1021; Member since: Nov 02, 2013

True, you can (and shall) always put the most recent SDK/NDK as a base, but if you target 2.3 as the minimum for example, you are not allowed to use the newer APIs not available in 2.3

22. nasznjoka

Posts: 418; Member since: Oct 05, 2012

nope you are allowed but it is a matter of using control statements. For example the action bar is not on 2.3 devices but you can create an app that is targeting 2.3 and yet for other upper revs still it is there.

25. JakeLee

Posts: 1021; Member since: Nov 02, 2013

A big question here : Do you do this extra work with control statements? or What do you think how many of the devs do this? We already know the answer.

26. Finalflash

Posts: 4063; Member since: Jul 23, 2013

No there is minor extra work. All you have to do minimum is create an annotation over the method that states which API level it is targeting and then point your code to the proper method based on a simple "if" check of version. Other than that you can create complete objects in separate classes to be used or complete apks whatever you are comfortable with.

39. bluemeansgo

Posts: 3; Member since: Dec 16, 2013

It's not just about control statements. It's about building fallbacks when an API is not available and developing alternatives for older systems. On the web, they call this graceful degradation and it's a LOT of work in many cases.

19. Reality_Check

Posts: 277; Member since: Aug 15, 2013

I believe you just proved him right. It all depends upon the viewer to decide which parameters to look for in a discussion on fragmentation. Different people consider different parameters.

21. JakeLee

Posts: 1021; Member since: Nov 02, 2013

I have to admit that I somehow misinterpreted that part. Michael was very polite in what he is saying. Ya know, he has to be careful not to enrage certain people. Here is a blunt translation : "Taking Siri into parameter on fragmentation is utterly bull****. There are people who like to do this, but they should start with all the carrier bloatwares on Android phones." Wow, there won't be enough colors left drawing the chart then.

24. darioq

Posts: 1; Member since: Dec 15, 2013

I also think it's misleading to put all effort into reducing the colours within the Android fragmentation chart, just to state it's not true. Every dev will just slap your face. Could we get charts for Android about the distribution of currently deployed apps? Like how many dev are now ONLY deploying for 4.4 in contrast to how many devs are only deploying to iOS7? In my naive world that would show the "dev adoption rate".

31. nasznjoka

Posts: 418; Member since: Oct 05, 2012

on my side 2.3 leads the way followed by jellybean on my appls with 15,000+ installations. So if you see it correctly the chats are correct cause its a battle between 2.3 and 4.x.x

16. SuperNexus

Posts: 127; Member since: Jan 18, 2013

Learn something new here. as always one of the best article.

15. TwentyWendy

Posts: 65; Member since: Aug 12, 2013

My iPhone 3gs doesn't have siri, but it has Google now because of iOS 6 :)

12. Dowson unregistered

Exceptional article. Well done.

10. nasznjoka

Posts: 418; Member since: Oct 05, 2012

Well written and nailed aritcle from who else than ...................?? MH keep on writing sh**t like this to bring them guys back on boad

7. N-fanboy

Posts: 543; Member since: Jan 12, 2013

You know you're a nerd when you finish an article this long. Lol

8. Hatshipuh

Posts: 163; Member since: Aug 09, 2012

Since when do nerds think this little bit of text is a long article?

9. Commentator

Posts: 3723; Member since: Aug 16, 2011

Since the average attention span has been whittled down to the

11. g2a5b0e unregistered

Nice. I see what you did th

14. elitewolverine

Posts: 5192; Member since: Oct 28, 2013

since nerds have tried to write less and less code and do more and more with said co

29. remixfa

Posts: 14605; Member since: Dec 19, 2008

u gys r fny :)

6. Chris_Bakke

Posts: 247; Member since: Jan 23, 2013

A fantastic read no matter what side of the iOS/Android fence you sit on. Spectacularly written Michael.

4. xperiance

Posts: 44; Member since: Oct 22, 2013

Thumps Up

2. hmd74

Posts: 546; Member since: Jan 31, 2013

Sooo long article... Read it later :]

3. PapaSmurf

Posts: 10457; Member since: May 14, 2012

You're missing out on great journalism.

Latest Stories

This copy is for your personal, non-commercial use only. You can order presentation-ready copies for distribution to your colleagues, clients or customers at https://www.parsintl.com/phonearena or use the Reprints & Permissions tool that appears at the bottom of each web page. Visit https://www.parsintl.com/ for samples and additional information.
FCC OKs Cingular's purchase of AT&T Wireless