One of the big puzzles in trying to understand the tablet market is data. Apple is pretty much the only large manufacturer giving unit sales, and not only are Samsung and Amazon saying nothing, but there are hundred of Chinese companies making Android tablets, and few if any say anything publicly. In the phone business the story is rather easier, since Google says roughly how many Android devices are being activated outside China, and give a screen size breakdown, and the Chinese mobile operators have a good sense of smartphone sales inside China. But in tablets there is no-one with an overall view of sales.
This means that analysts have to fall back on component makers. Though there are hundreds of companies making tablets, they almost all use more or less off-the-shelf chipsets from a handful of companies. These companies have a good sense of the overall number of tablet chips that are being sold into the manufacturers. So too do the screen markers - they have a good sense how many 7” panels are being sold. So you take an estimate of tablet chip sales and an estimate of tablet screen sales (and take into account people using ‘phone’ chips in tablets) and you estimate tablet sales, both inside and outside China.
The obvious problem with this is that these devices are being used in very different ways. It seems clear that most of the huge numbers of sub-$150 Android tablets now being sold do not have anything like the web or app usage that is seen on an iPad, Nexus or Galaxy Tab, and that many are mainly used as substitutes for TV sets, with maybe some gaming on the side.
The deeper issue, though, is that estimating tablet sales in this way is a little like trying to estimate global car sales by working out how many internal combustion engines are being made, and how many tyres, but not adjusting for motorbikes, cranes, outdoor generators or 18 wheelers. Lots of ‘tablet' chips and ‘tablet' screens do not actually end up in tablets.
Consider this device, one of thousands of similar products on offer on Alibaba - an in-car video player running Android, complete (if the screenshot is to be believed) with the phone app. This probably doesn’t activate with Google, but it certainly looks like a ‘tablet’ to LG or Rockchip.
Then there’s this TV dongle - no screen, but does it have a ‘tablet’ chip? A ‘phone’ chip? If you used it to watch Youtube, what would Google think it was? (Note also the memory card slot, used for side-loading pirate movies.)
Now what about this, from Steelcase? An meeting room door with a 7” capacitive touch screen. To a component maker, this is also a tablet. I have no idea if it runs Android today, but if it doesn’t, it probably should. And if Nest doesn’t, the copies of it will.
The important dynamic here is that a combination of very cheap off-the-shelf chips and free off-the-shelf software means that Android/ARM has become a new de facto platform for any piece of smart connected electronics. It might have a screen and it might connect to the internet, but it’s really a little computer doing something useful and specialised, and it probably has nothing to do with Google.
As should be obvious, this makes counting total ‘Android' devices as though they tell you something about Google or Apple’s competitive position increasingly problematic. But to me, pointing out that ‘Android’ doesn’t necessarily competed with iPad is rather boring - what’s really interesting are the possibilities that these new economics might unlock.
A good example is this - a 2G Android phone wholesaling for $35 (just one of hundreds).
Now, stop thinking about it as a phone. How do the economics of product design and consumer electronics change when you can deliver a real computer running a real Unix operating system with an internet connection and a colour touch screen for $35? How about when that price falls further? Today, anyone who can make a pocket calculator can make something like this, and for not far off the same cost. The cost of putting a real computer with an internet connection into a product is collapsing. What does that set of economics enable?
There are other interesting hardware trends that overlap with this as well. Bluetooth LE is the obvious one - you can make a widget that broadcasts a location ID for $50 or less, stick it on a wall, and the battery will last for years. EInk is also interesting here: it needs no power to show something, only to change, and you can pass enough power over an induction touch point to cycle the screen. The problem here is cost, but why doesn’t my Oyster card show the remaining balance? Why wouldn't Coin (a product I'm rather skeptical of, mind) show the cards loaded onto it on one side with an eink display? How do these trends interact with cheap Android computing?
Marc Andreessen famously coined the phrase 'software is eating the world’ to describe the way that functions that used to be served by dedicated hardware are now being subsumed by general purpose devices - mostly smartphones. But there’s also the beginnings of a trend in the other direction - devices that weren’t smart and didn’t get merged into the phone gaining a digital presence of their own, and creating a new set of opportunities.
Looking at the Chinese mobile market today reminds me a lot of looking at the Japanese model in 2000 or 2001 - lots of very interesting stuff is going on, but getting reliable data is very tough.
One triangulation point comes from app analytics platforms. You need to have some caution as to how representative they are, but the big ones give a good directional steer. Umeng (think Flurry for China) is one of the biggest. It puts out statistical reports every quarter or so - these are some of the key charts in the latest.
First, platform size - this is their estimate of active devices that are using apps (not total devices), including tablets.
Both iOS and Android are growing fast, and Android faster, as one would expect given the range of prices that Android devices are offered at. However, on this data there are probably more iOS devices in China than smartphones in the USA.
Second, handset brands. Apple is the largest single brand in this data set, but shrinking. This of course is users, not ongoing sales, so some recent suggestions that (for example) Xiaomi outsold iPhone in the run-up to the iPhone launch may be compatible with this.
Finally, and perhaps most interestingly, a window into the state of Android in China. As most people know, the great majority of Android devices sold in China are built on AOSP and have no Android services pre-installed (indeed, I've used a Motorola phone with no Google services present) - instead they have a range of apps from the local internet giants.
This means that most Android phones have no Gmail, Google Now, Google Maps or, of course, Google Play, and most apps are installed form third party app stores or side loaded, either pirated or downloaded directly from publishers' sites. Some handsets do have these, either because people added them afterwards (which is not easy) or because they're using grey market imports. Estimates of the total with Google apps on them are mostly in the 20-30% range. However, on Umeng's data, Google Play amounts to just 5.6% of Android app installs in China. OEM app stores are 8.5%. This, for example, is why Baidu paid $1.8bn to buy a couple of app stores earlier this year.
Apple, obviously, blew the doors off with opening weekend iPhone sales - 9m units in 3 days, up from 5m last year. Having day one launch in China and adding DoCoMo to distribution obviously helped, but there's clearly still strong underlying organic growth. And it appears that this is without substantial demand for the 5C (which is not an early-adopter/queue overnight sort of product).
However, the really interesting thing is that there are now 200m people using iOS7, where last year 'only' 100m people upgraded to iOS6 in the opening weekend. The chart below shows what this means, as compared to Android, the other platform.
Google, of course, is trying to address the fragmentation embodied in these charts with a shift to Google Play services, as neatly explained by Ars Technica here. But though this means Google itself is less subject to fragmentation, it isn't much help to a developer wondering whether to use APIs that are only in Android 4.2 or later - let alone one wondering why their app crashes on one Android 4.2 phone but not another.
This issue makes it hard for Google to drive the agenda for new mobile technologies within Android: it will take at least a year after announcement before a meaningful part of the base has access to anything new. Hence the focus on Google Play services and on the cloud with things like Google Now - moving everything several layers up the stack from the intractable fragmentation problem, and making the hardware OS less relevant. But of course, this reduces further the reasons to upgrade your OS, and makes it much less likely that third party apps will do anything on Android that they don't do on iOS (system utilities and other minority interests aside).
Conversely, a developer can use anything that Apple announced in iOS7 and be confident it will work on all their users' devices. So anything innovative Apple does takes effect right now. Apple does have some fragmentation issues - some of the coolest features in iOS, such as Airdrop or iBeacon, have chipset dependencies that rule out older hardware. But Apple's integration means that it can drive innovation on the device much faster than Google. It can do Airdrop - putting that in the next version of Android and expecting it to work predictable for a meaningful proportion of the base any time soon would be much harder.
Hence the paradox: the open platform is actually slower-moving in some ways than the closed one.
Incidentally, the fact that Google seems to be moving more and more innovation away from the OS poses interesting questions about future roadmaps. Will 'Android' still be the main platform in the future, or will it be Chrome, or something else, with Android buried underneath?
Moore's law in action.
This week Google announced that it has now passed 1bn cumulative activations. This is what that trend looks like.
Google generally releases these numbers at scheduled events, and they're pretty round, so we can't take them as exact. My model makes some allowances for the overall trend rather than trying to hit the exact number on the exact day of the announcement (since the figure might actually have been passed a week earlier anyway).
If we aggregate this by quarter and split out tablets (also provided by Google recently), we get this chart, showing rapid but slowing growth and tablets as a pretty small share of the total.
This base, as everyone knows, is somewhat fragmented, with a wide mix of versions in use.
The interesting thing is that if you apply these percentages to the active base (on the assumption that trailing 24m activations are active), it appears that there is very little upgrading of existing devices: the growth of new versions of the OS seems to come solely from new devices.
Of course, what's missing from this data are the devices that are used in China and do not have any Google services on them, and hence do not activate with Google. There is, obviously, no good data on this, but most people in the Chinese market think that only, say, 20% of Chinese Android devices have Google on them. That means the overall Android market is rather bigger than 1bn, with well over 200m extra devices in China, and run-rate 'unactivated' Android making up perhaps a quarter of total Android sales.
Meanwhile, as Ars Technica points out, Google is starting to get around fragmentation by putting all its services into a software layer on top of Android that it can update remotely. This solves the fragmentation problem for Google (their services are on every Android device outside Amazon and China), but doesn't help Android developers trying to work out why their app crashes on one Samsung model but not another.
The smartphone business tends to be cast in terms of a simple ‘platform war’ between Apple and Android, with market share and profit share tracked and compared. However, this does not really capture what Google or Apple are trying to achieve. It should be obvious Apple is uninterested in being the biggest manufacturer of phones per se, but the raw market share of Android is also several steps away from Google's objectives.
Google's fundamental strategic needs are to extend reach and gather data. Google needs to be on as many screens as possible, delivering as many searches* (and ads) as possible, and it also wants to have access to as much data as possible to index, understand and serve up in ever-improving search results.
The growth of the mobile internet intensifies these imperatives in two ways:
- Reach: search happens on more devices in more places with greater frequency
- Data: mobile devices can provide far more information about behaviour and intent that can help Google deliver more relevant search results
Hence, Google uses Android as a tool to extend its reach, both as a generic access platform that can go to Google.com and by embedding more and more ways to use Google Search within it. But the objective is reach itself, which it will take anywhere it can get it, including on the iPhone – and Google tries hard to put its services onto the iPhone.
In parallel, Google uses Google Plus to collect and understand usage. Talking about the low rate of social sharing on Plus misses the point. Rather, while the history of Google so far has been about understanding the web and the interactions and links within it, the future of Google is about understanding and learning from how people use the web. Plus is the mechanism to do that - it's PageRank for the users. This is why Plus is being stuffed into all manner of Google products - not to get you to share stuff with your friends, but to be able to draw conclusions from all of your activity in the same way that PageRank draws conclusions from all web links.
This means that a helpful way to look at Google is as a vast, decade-old machine learning project: mobile will feed the machine with far more data, making the barriers to entry in search and adjacent fields even higher, while Plus is the database that ties that data to individual behaviour. The combination of the two strategies is (hopefully) self-reinforcing – reach and data collection produces better results, more reach and more data collection. The more control Google has of the mobile device, the more it feeds the machine: on an Android phone you will always be signed into Plus, even if you never think you're using it. But all devices on all platforms (outside China) are feeding the machine.
(* Of course, things like Now and Glass point to changing future definitions of what 'search' means)
A good chart is worth a thousand words.
"The [GM] execs would fly into Phoenix’s Sky Harbor airport, limo out to the Desert Proving Grounds and drive the company’s latest models.
Our agent says that all the vehicles the execs drove were “ringers.” More specifically, the engineers would tweak the test vehicles to remove any hint of imperfection. “They use a rolling radius machine to choose the best tires, fix the headliner, tighten panel and interior gaps, remove shakes and rattles, repair bodywork—everything and anything.
Did the execs know this? “Nope. And nobody was going to tell them . . . As far as they knew, the cars were exactly as they would be coming off the line. That’s why Bob Lutz thinks GM’s products are world-class. The ones he’s driven are.”
I read this piece (from 2009, the time of the US auto bankruptcies) at the time and filed it in the back of my mind. But I was reminded of it this week, because I've been trying to use a Sony Xperia as my main phone. I won't bore you with my impressions - this isn't a gadget review blog - but what I found interesting was the gap between the experience offered and that on my Nexus 7, or indeed that described by Paul Stamatiou's great piece on the best aspects of Android.
This is hardly a new observation: the experience you get on most OEM Androids is not the same as that you get from a Nexus phone running 'stock Android', or one of the aftermarket ROMs (which however remain very fiddly to install). Google even sells phones like the Galaxy S4 in special 'Google' versions. It's rather like Microsoft selling 'stock Windows'.
It's unfair (or perhaps incomplete) to blame Google for this - this is all part of the fragmentation that's inherent in the open model that drives Android's 40% share of global phone sales. If you want to see a phone OS with none of these issues, look at how well Windows Phone is doing.
But the reason I'm reminded of the GM story is the old computing term about 'eating your own dog food'. Nexus devices and 'official Google' devices are not Google's dog food. They're not what actual customers use.
The Nexus 7 sold only 7m units in almost a year and was only 10% of 'activated' Android tablets, on Google's numbers. The Nexus 10 appears to have sold under 1m units. There's no strong indication that the Nexus handsets sell in large volumes, and rumours from Korea are for just 4m (global) unit sales of the most recent model in six months (it sold out at launch, but we don't know the order size). One can argue about why this is - distribution, branding, sales commissions, consumer awareness etc - but the Nexus phones aren't on anyone's list of top-selling or top-used Android phones that I've seen. This is device manufacturing as vanity publishing.
Hence, for a Google employee to use a Nexus is like a GM exec test-driving a ringer. They're using 'Android'. It's great. But that's not the phones the customers are getting. They're not eating their own dog food.
The irony is that if you buy a $125 generic Chinese Android you're actually more likely to get a 'stock Android' experience with no preinstalled 'crapware' than if you spend $400 or $500. Except, of course, that if you're in China there's a more than even chance you'll have no Google apps preloaded either.
Over the last 12 months, Google Android devices have outsold iOS by about 3 to 1. There are now perhaps 775m-800m 'official' Android devices in use, versus perhaps 415m iOS devices. This is without counting sales of the Amazon Kindle Fire or the (very) many Android devices sold in China that are not connected to Google services - these may be a further 150-200m active devices now (or more). So, the Android install base is more than double the size of iOS. If you look just at phones, there are maybe 250m iPhones in use and perhaps 700m 'official' Android phones alone.
Of course, Android has had a larger installed base than iOS since mid 2011, but engagement remained far behind. Until well into 2012 publishers and developers tended to see app download rates on Android of a half to a quarter of what they experienced on iOS, in absolute terms, while payment and purchase rates were a quarter or lower of iOS rates. This was due partly to weak execution by Google and partly to a vicious circle of self-selection: Android phones were cheaper (averaging $250-300 globally versus $600 for the iPhone) and had a poorer choice of apps, so they attracted people with less money and less interest in apps. The small and very vocal number of people who really want an Android are far outnumbered by people who don't particularly care.
This engagement gap is now starting to close:
- Google has significantly improved the execution of the ‘Google Play’ app store, and has made a strong effort to extend operator billing
- The sheer size of the Android base, at close to 800m active devices (outside China), versus around 400m for iOS (and around 250m iPhones), means that even with lower proportionate engagement rates the absolute numbers are starting to catch up
Hence, by the first half of 2013 Android cumulative downloads caught up with iOS (both at around 50bn), and both now see a run-rate of something around five apps downloaded per active device per month.
Google does not give any data on payment trends, but Apple gives enough to estimate that iOS users spend about a dollar a month on apps and in-app purchases. Android payment levels (both on Google play and with other methods) are significantly lower (perhaps half the value of iOS in total, despite having double the users, but there's very little data to go on), but again are rising (this Distimo report is one illustration).
- There are 2-3x more Android users than iOS users
- The average Android user, for a range of reasons, downloads fewer apps and spends less on the app store than does an iOS user
- However, there are so many more Android users that the total rate of app downloads has caught up with iOS and the total value of payments rate may also be starting to catch up
In other words, the average Android smartphone user is worth much less than the average iPhone user, but there are lots more of them.
Does install base matter? It may start to.
It is possible to argue that this does not matter much: if Apple’s ecosystem of 400m active devices is big enough to sustain developers already, does it matter that the Android opportunity may become even bigger?
When Apple launched the app store, a year after launching the original iPhone, it had sold just 6.1m iPhones (as well as a significant number of iPod Touches), yet this was enough to attract a flood of developers – even a year later it had only sold 26.1m, which is roughly the same as the 27m Windows Phone Lumias that Nokia has sold to date. Yet Windows Phone is an afterthought at best for developers. Though some independent developers may see an opportunity in an empty Windows Phone app store, for most the question is where best to allocate limited resources. Where is the opportunity cost, and where is the best return? The size of an ecosystem per se is not important – what matters is which should be the first priority.
Hence, developers are starting to move from creating new products on the basis ‘iPhone, then maybe Android’ to ‘iPhone and then Android’ or even ‘iPhone and Android at the same time’. Cool little apps from seed-funded companies are still iPhone-only, but most big well-funded businesses are doing both. Android is no longer optional for any publisher seeking real reach. This applies even more if your app is free, since Android download rates are much closer to iOS than are Android payment rates (and of course 'free' doesn't mean no revenue).
Of course, this comes with problems. It's pretty clear that for many developers supporting Android costs 2-3x what iOS costs (the BBC made this explicit, for example), which pushes it further down as a priority. In addition, lower individual engagement means lower CLV, and so the dynamics look different on (for example) customer acquisition. But the greater absolute numbers all act to push the other way.
If total Android engagement moves decisively above iOS, the fact that iOS will remain big will be beside the point – it will move from first to first-equal and then perhaps second place on the roadmap. And given the sales trajectories, that could start to happen in 2014. If you have 5-6x the users and a quarter of the engagement, you're still a more attractive market.
This is a major strategic threat for Apple. A key selling point for the iPhone (though not the only one) is that the best apps are on iPhone and are on iPhone first. If that does change then the virtuous circle of ‘best apps therefore best users therefore best apps’ will start to unwind and the wide array of Android devices at every price point will be much more likely to erode the iPhone base. Part of the reason for spending $600 on an iPhone instead of $300 on an Android is the apps – that cannot be allowed to change.
A new, cheaper, high-volume iPhone would have the potential to mitigate or even reverse this trend. Clearly, like current low-end Android, it would sell to a demographic with a lower average engagement and purchase rate and so the average iOS rates would drop. However, it would mean that iOS’s reach would expand significantly at the expense of Android. How would a $200 or $300 iPhone sell? Easily double digit millions, possible up to 50m units a quarter.
This means that the financial value of a cheaper iPhone cannot be considered in isolation. A large part of its purpose is to defend sales of the high-end model.
This is an excerpt from a longer report on the outlook for a cheap iPhone that I published for Enders Analysis.