Cellcom
|
Read the first paragraph here, and then see if you can spot the flaw in the sentence that follows. (It comes from here, but it just happens to be a convenient example.)
It’s been 20 months since we went hands-on with Nike’s first FuelBand, and despite assumptions at the time that the accompanying FuelBand app would land on Android in the not-too-distant future, the activity-tracking wristband remains compatible with iOS for now, even as we approach the brand new FuelBand launch next week.For many, this glaring omission in Nike’s technological armory is astounding given that, well, Android represents somewhere in the region of 80% of the smartphone market.
So what's wrong with the sentence? After all, in the third quarter of 2013 Android did have around 80% of worldwide market share. That's correct -here are the ABI research figures.
It's simply wrong, though, to extrapolate from that to think that four in five smartphones in peoples' hands are Android-powered. Here's the reality: at the time this was written, more than 40% of the smartphones in use in the US (a key market for Nike) were iPhones. Only about 51% of the smartphones in peoples' hands in the US are Android phones. The ratios are more in Android's favour elsewhere, but nowhere outside of China (and perhaps India) would you find four in five smartphone owners using an Android phone.
"Market share" and other phrases can lead you astray. So let's clarify those terms.
OK, so - what is market share?
It's a snapshot of how many items were sold (or maybe shipped) in a market over a specific period. So, for example, you might be told that in the most recent financial quarter (three-month period), sales data shows that the market share in country Z for the Widget variants was
It's a snapshot of how many items were sold (or maybe shipped) in a market over a specific period. So, for example, you might be told that in the most recent financial quarter (three-month period), sales data shows that the market share in country Z for the Widget variants was
So that means that 80% of people in country Z use Widget A?
No, it doesn't. In fact, you can't draw any conclusions at all about how many people in country Z use Widget A from that "market share" figure.
No, it doesn't. In fact, you can't draw any conclusions at all about how many people in country Z use Widget A from that "market share" figure.
Huh? But it says market share! It's a number and everything!
Yes, it does. But market share is a measure of sales. Only in the specific case where the market is saturated - that is, everyone who wants a Widget has one, so that now the market is essentially just replacements - does market share probably tally with "installed base".
Yes, it does. But market share is a measure of sales. Only in the specific case where the market is saturated - that is, everyone who wants a Widget has one, so that now the market is essentially just replacements - does market share probably tally with "installed base".
Installed base? What's that?
It's a measure of how many of something are already in use. So if 100 million people are using Widgets, there's an installed base of 100m.
It's a measure of how many of something are already in use. So if 100 million people are using Widgets, there's an installed base of 100m.
Now, let's suppose further that there are different kinds of Widgets - A, B, and C - and that the breakdown in installed base of Widgets is that
Ah - so now this market share data - A has 80%, B has 15% and C has 5% - means that A is starting to take over, right? Its market share is bigger!
It might be the case, but you can't draw that conclusion from this data. First, you don't know how many widgets were sold/shipped in that period.
It might be the case, but you can't draw that conclusion from this data. First, you don't know how many widgets were sold/shipped in that period.
What if only 10 were sold? That wouldn't make much impact on an installed base of a hundred million.
Second, you don't know if the market is expanding (or contracting, though few do in technology). If the market is expanding, then some are going to new users who didn't previously have Widgets.
Hang on. Why all this sold/shipped stuff?
Many technology transactions, for hardware at least, involve intermediaries. Handset manufacturers make handsets and then sell them to carriers, who sell them on to customers. The handset maker will record the number of handsets "shipped". But that doesn't mean they're in the hands of customers. If something goes wrong, then the carrier will send unsold handsets back. Different manufacturers have different rules on quite how they define "shipped", but generally it's when they receive money for them from someone else, who takes ownership.
Many technology transactions, for hardware at least, involve intermediaries. Handset manufacturers make handsets and then sell them to carriers, who sell them on to customers. The handset maker will record the number of handsets "shipped". But that doesn't mean they're in the hands of customers. If something goes wrong, then the carrier will send unsold handsets back. Different manufacturers have different rules on quite how they define "shipped", but generally it's when they receive money for them from someone else, who takes ownership.
That doesn't mean that they will reach customers, though. For instance, the HTC First – an Android phone built by HTC to have the Facebook Home profile built in – sold so poorly that AT&T dumped it; the unsold inventory went back to HTC. Shipped, but not sold, and definitely not in customers' hands.
In general, though, this distinction isn't important. Analysts call shipments to intermediaries "sell-in"; the total number that actually reaches customers is "sell-through".
Often they're pretty close, though sometimes (such as ahead of big launches) the "sell-in" can fall or rise as handset makers try to get the last of an old model out of the system, or try to pack lots of a new one in. BlackBerry indulged in a lot of "sell-in" of its Z10 handset after its launch in January 2013. The "sell-through" to customers was rather more limited.
Fine. So installed base and market share aren't the same. What can we deduce from market share then?
Not very much, initially. Consider all the variables, even if we only look at country Z - where, you'll recall, 20m use Widget A, 50m have B, and 30m have C.
Not very much, initially. Consider all the variables, even if we only look at country Z - where, you'll recall, 20m use Widget A, 50m have B, and 30m have C.
Let's also suppose that the market share numbers for the latest quarter - 80% A, 15% B, 5% C - was spread among 10m units. So sales (assuming no sell-in/sell-through shenanigans) were 8m for A, 1.5m for B, 0.5m for C.
I've got it! There are now 28 million A users, 51.5 million B users, 30.5 million C users. This is easy.
No, you can't say that. You don't have enough data.
No, you can't say that. You don't have enough data.
Why not? We have the market share, and we have the installed base. Add one to the other – there you are.
But you don't know whether the market is staying the same size or expanding; and you don't know what the replacement rate is for A, B and C. What if everyone who buys one of those widgets is just replacing an old one? Then the installed base stays the same size, and the ratio of users stays the same size.
But you don't know whether the market is staying the same size or expanding; and you don't know what the replacement rate is for A, B and C. What if everyone who buys one of those widgets is just replacing an old one? Then the installed base stays the same size, and the ratio of users stays the same size.
On the other hand, if the market is just expanding, with no replacement of existing Widgets, then you would see the 28/51.5/30.5 breakdown suggested above.
In reality, you'll see something in between: some Widgets go to new users, and some go to existing users. In general, the larger the installed base, the more units will go to replacing old ones. (For example, most people replace their phone every couple of years, so a large installed base of phones has constant "churn".)
The point is, without that extra data, you can't know how the installed base is changing.
Modelling how the installed base is changing for large groupings like the Apple and Android installed base is a challenge for analysts: you have to model how often people replace their device, and how many of them are out there, and how fast the overall market is growing. This is why they fall upon any tidbit of information that might hint at "installed base" eagerly.
There's a simple refutation of the idea that "80% market share for Android" means that four in five smartphone owners is using an Android device, though. And it comes from Google's Android developer dashboard.
Here's the image for the seven days to 1 November 2013, showing what version of Android was used by devices connecting to Google Play. Among them is "Froyo" (2.2) and "Honeycomb" (3.0). You can't actually buy new Android devices running either of those. Yet both make up a notable proportion. The newest software, "Jelly Bean" (which actually covers three different numbering versions), accounts for 52.1% of the devices. Yet Jelly Bean is the software powering all those new Android phones – the ones that were the 80% in the past quarter. Clearly, the installed base doesn't reflect the market share number.
Fine. But look, I saw some figures which said that last year Apple's market share in tablets was 50%, and now it's 30%. So Apple's selling fewer tablets, right?
No, that's not what that data tells you. What if the total number of tablets being sold has doubled? If last year there were 100m tablets sold in total, and this year 200m, then last year the figures would be 50m tablets and this year 60m. (Those aren't the numbers. They're just for illustration.)
No, that's not what that data tells you. What if the total number of tablets being sold has doubled? If last year there were 100m tablets sold in total, and this year 200m, then last year the figures would be 50m tablets and this year 60m. (Those aren't the numbers. They're just for illustration.)
So if you don't have the absolute numbers, you don't know what's happening. Those sort of year-to-year comparisons can be helpful to visualise changes in the market landscape, but in fast-changing markets it's not enough just to quote a single number. In some ways it obscures more than it reveals.
(Note that Google's diagram above doesn't have any numbers beyond the "share"; we don't know if more devices connected to Google Play, and if they did, how many more that was compared to a month or year ago.)
Here's an example, from YouGov in 2013, which showed Apple losing market share. Its share of the installed base had fallen, YouGov said, from 73% to 63% (note that unusually, this was an "installed base share", not a "sales market share").
And yet putting in the figures for how many units were bought showed that Apple had increased its installed base and increased its lead in that installed base. That's counterintuitive. Yet it emerges directly from the calculation: the iPad installed base had gone from 2m to 5.3m; it had gone from having 0.6m more tablets than all its rivals combined, to having 3.1m more.
This doesn't mean there won't be more non-iPad tablets than iPads at some point. But it does mean that you need to enquire more carefully about absolute numbers when you're presented with the word "share". It's absolute numbers that tend to matter.
But if the market share figure is so useless, why does everyone quote it all the time?
Now we get to the key point. Because it's easy to measure market share – much easier than measuring installed base, which requires large panels of people who you interview on a regular, repeated basis. (ComScore does this in the US, where it provides a picture of the installed base of smartphone users that is consistent back to the end of 2009. Its figures for the three months to September 2013 show a 51.8%installed base for Android - that's 76.6m - and 40.6% for iPhone - that's 60m. It's not 80% Android; not even close.)
Now we get to the key point. Because it's easy to measure market share – much easier than measuring installed base, which requires large panels of people who you interview on a regular, repeated basis. (ComScore does this in the US, where it provides a picture of the installed base of smartphone users that is consistent back to the end of 2009. Its figures for the three months to September 2013 show a 51.8%installed base for Android - that's 76.6m - and 40.6% for iPhone - that's 60m. It's not 80% Android; not even close.)
Plus "market share" gives journalists who like nothing better than a metaphorical horse race something to write about - look at the preponderance of polls, especially in the US presidential election. Trouble is, it doesn't necessarily give us useful information.
Market share does tell us who's shipping the most of something. If you put it together with other things that matter to businesses – such as profits and revenues – it can be illuminating about who is doing business most effectively. And it can tell you how a market is changing, if you have a time series.
Unfortunately some people don't understand how little market share on its own tells you. Which leads them to focus on those figures alone, and to misinterpret them, like the extract quoted at the top.
Does anything else make market share an unreliable predictor of installed base?
"Market share" tends to measure things that are sold new. But secondhand markets can also be significant. Some people buy a new TV and get rid of the old one at the local dump; others just move the old one to a different room. If the new TV is a "smart" TV and the old one wasn't, in the first case the installed base of smart TVs in that house is 100%; in the second, it's 50%.
However secondhand use is difficult to measure through any sales mechanism, and it can make calculations about the installed base harder. (In effect, it changes the replacement rate for a device.) If there's a significant difference in secondhand use/ownership between Widget A, Widget B and Widget C, your market share numbers are an even less good guide to the installed base.
Who cares about installed base, anyway?
App developers are a key market, as are accessory makers and users considering what their next purchase should be (because you're more likely to get repairs, support and linked services if the installed base is larger). The article linked at the top is all about that installed base/developer conundrum. The lack of an installed base (despite "market share" numbers) is typically why you can't get an app for a new platform: developers need to have reason to believe their effort won't be wasted.
Wait a minute, though. Windows runs 95% of PCs - that's the installed base - and it has 95% market share! You're trying to con me!
You're right - Windows has 95% of both installed base and market share. That's because it's effectively a saturated market: the number of people in the world using PCs (or Apple Macs) doesn't seem to be growing.
You're right - Windows has 95% of both installed base and market share. That's because it's effectively a saturated market: the number of people in the world using PCs (or Apple Macs) doesn't seem to be growing.
In turn, that means that if 10 PCs are shipped, they're almost certainly replacing 10 existing PCs. One in, one out. In fact, the problem that PC manufacturers presently face is that the installed base is pretty happy; PCs are living longer, which means that replacements aren't needed as often.
Windows achieved its dominance in the 1990s, and nothing has disrupted it for desktops since. In the context of the 1.4bn or so PCs out there, the few million that Apple sells each quarter isn't going to make a substantial impact on the installed base ratios (Windows:Mac).
So does that mean that market share eventually equals installed base?
Yes - eventually. But bear in mind all those points above about replacement rates.
What about search? If Google has a market share of 90% in search in Europe, is there something tricky behind that?
No, because searches aren't preserved. You search, it's done. For temporary items, "market share" describes use; there's no "installed base" of searches.
No, because searches aren't preserved. You search, it's done. For temporary items, "market share" describes use; there's no "installed base" of searches.
So what should I do next time I see a story about market share?
See whether it has the supporting numbers you need: the total size of the market, and what it tells us about the installed base. And what, if anything, it tells you about what's happening in the wider market. Samsung's colossal market share in smartphones and mobile phones, for instance, is reflected in installed base figures – and also in its profits and heft in the business world. Sometimes the numbers are useful. But on their own? Not really.
See whether it has the supporting numbers you need: the total size of the market, and what it tells us about the installed base. And what, if anything, it tells you about what's happening in the wider market. Samsung's colossal market share in smartphones and mobile phones, for instance, is reflected in installed base figures – and also in its profits and heft in the business world. Sometimes the numbers are useful. But on their own? Not really.
No comments:
Post a Comment