The (fixable) problem with Steve Ballmer’s USAFacts site: its lack of transparency

Those of us who have been in or around the technology space since the 1990s when Microsoft used to bulldozer all in front of it (for those younger than that: like Google in the 00s, and Facebook now) are completely unused to Steve Ballmer doing things that we can uncomplicatedly see as good.

But his new site, USAFacts, is one of those things. Like Bill Gates pouring money into vaccinations for developing countries, it’s a good thing to have done.

USAFacts, in case you haven’t read the New York Times piece about it (where the site is described as a “fascinating data trove”), the precis is that Ballmer has spent $10m hiring economists and others to put all the spending and other data that the US government products – particularly its budget, where it spends $5.4 trillion and gets $5.2trn (it’s left as an exercise to the reader to figure out how one bridges that $0.2trn gap) – into a site which you can query, to find out just where your money goes, if you’re an American.

Neatly, it divides expenditure into the four “missions” of the US government, which apparently are “establish justice and ensure domestic tranquility; provide for the common defence; promote the general welfare; secure the blessings of liberty to ourselves and our posterity”.

There may be trouble ahead

The problem starts once you begin wandering into it and wondering: ok, how does that number come together? Take the “Crime and police” tab under the first mission: click through and you get some data showing how arrests, violent crime, and “public safety officer” numbers are moving. The prisoner data actually has a link to a set of slides – 291 PDFs – drawn from the Bureau of Justice Statistics and reproduced as PDFs by USAFacts, which sticks its own copyright onto them (down in the bottom right).

Screenshot 2017 04 19 17 54 04

(Come on. That’s crazy. Repurposing open-licensed data as PDFs and sticking your own copyright note on them? Is this the 1990s?) In general, you’re left having to trust the site to have got it right. Plus: in almost every situation, we don’t know where the data has actually come from. The prison data is an exception. There’s no transparency in a site which is trying to make government transparent.

Unless I’ve missed something staringly obvious, there are no places where you click on a link and it says “we got this from the Bureau of Labor Statistics, and this from the Treasury, and this from the Department of Defense”. The methodologies (here’s one – PDF) will give you a number of the methods and sources from which all the numbers are collated, but not how they’re balanced out against each other.

This is a huge omission. I can understand that $10m only goes so far, but if you compare it to a site like Our World In Data, which has been built on a budget probably comparable to one day of USAFacts’s, you see the contrast. Not only are the datasets open and downloadable, you get pointers back to the originals.

Coins: adding up to something

Nor is it as though trying to analyse government spending is a new thing. In the UK, HM Treasury made its COINS database into open data after pressure from newspapers – well, particularly the Guardian, where we championed free data.

From COINS you could generate visualisations like these:

And you could do all sorts of visualisations of what was going on.

This was back in 2010; it’s not as though this is some groundbreaking piece of work that has only just been released to the public and might not have crossed the radar of, say, a multibillionaire who has enlisted a lot of economists and statisticians and programmers to do some work analysing a country’s budget.

Action points

In conclusion: Steve Ballmer has made a start. It’s an adequate start, and given the size of what he’s trying to contend with, it’s laudable that he’s got this far. But there are many more things that remain to be done.

Just in case he’s vanity-searching, here is my list of suggestions for how to turn USAFacts into a truly useful site that will let people explore what the US government and its states are doing with their money:

• link back to the original data. People need to be able to trust it.
• offer inflation-adjusted views of the data. It’s not hard to find inflation figures for past years.
• add trade figures. Imports and exports are relevant data for understanding your country’s performance, and some part of government definitely collects them.
• add comparative figures with other countries. It means nothing to tell people how much they’re spending on health care if they don’t know how that compares to other countries.
• find ways to let people drill down to their state, country and district, and compare them with other states, counties and districts. It’s just data and databases, after all.
• let people see how employment and other elements have changed. Show them how racial factors have changed. Show them how things have changed with different politicians. Everyone wants to apply these prisms to these data, and while you might not want this data to become partisan, the reality is that these numbers will be used anyway by people of whatever tinge to prove whatever they want. You’re already in that battle, so give people more weapons to fight it.

It’s a good start, Steve. Please don’t make us wait for version 3 for it to become a must-use. (Yes, kids, that was a Microsoft joke. Ask your parents.)

You think it’s a Muslim laptop ban? This picture suggests it’s really a terrorist ban

Airside at Baghdad Airport
The photo is apparently taken airside at Baghdad Airport; the paper says “The Islamic State is right in your home”. Source: Twitter.

The decision by the US and UK to ban carry-on electronics of various sorts from a number of countries in the Middle East has brought out all sorts of unthinking reactions. Trump is driving a lot of people stupid, which is a pity.

An example of the unthought-through reaction is this article at PC Mag, where Sascha Segan says

The DHS notice doesn’t give any evidence of specific threats leading to this new ban, which will go on indefinitely. It doesn’t explain why a bomb can explode in the cabin but not the cargo hold, or why travelers but not airline employees are affected. While it has a 30-question FAQ about the ban, most of it is meaningless weasel words adding up to “trust us.” The more you think about any aspect of this ban, the less it makes sense from a security perspective.

Not to pick on Segan particularly; variations of this article, in hot-take form, were all over the web when the news broke on Monday (US ban) and Tuesday (UK ban).

However, it’s worth remembering – as if you hadn’t had plenty of reminders recently? – that the intelligence services have access to more information than you do.

Liquid memories

Remember the liquids ban of summer 2006? It was imposed out of the blue, and threw airports, airlines and security into near-chaos. Wikipedia has a good summary of why it was introduced: British police (perhaps helped by intelligence services) had uncovered a plot to blow up a plane in mid-air, using liquid explosives in soft drink bottles. In all, more than 20 people were arrested; nine were eventually tried, and seven found guilty of conspiracy to murder.

Now, with that in mind, when the governments of not one but two countries impose sudden bans on the transport of potentially explosive things, you might think that people would take it seriously. There was one occasion when a would-be mass murderer ignited a bomb on the passenger deck of a plane out of Somalia – after apparently being handed the explosives by a ground worker. In a fabulous demonstration of karma, he was sucked out of the hole he’d made in the fuselage, and the plane landed safely. Subsequently, 20 ground staff in Somalia were arrested.

There are suggestions that this latest ban has been under discussion for a couple of weeks, in fact. That’s how intelligence works: gather data, consider risks, act.

The number of people complaining that “it’s just another version of the [Trump] Muslim ban” can’t be thinking clearly. The original “Muslim ban”, as a reminder, included Syria, Yemen, Iran, Iraq, Sudan, Somalia and Yemen.

It didn’t include the ones in the US ban: UAE (which includes Dubai), Turkey, Egypt, Jordan, Saudi Arabia, Qatar, Kuwait, or Morocco. The UK ban includes Tunisia too.

It should be pretty obvious, even if you think Trump is a fool, that this isn’t his idea. It has come from intelligence agencies who are worried about the possibility of a mid-air explosion.

You can see why Islamic State and similar terror groups might want to do something now. IS is being gradually crushed in Mosul, which means that its fighting force is dwindling fast. Its revenues are dwindling too, because its sources – illicit oil sales, “taxes” on the populations it oppresses, and ransoms – are all being squeezed. Lower revenues means less money for weapons and less opportunity to control territory, and the caliphate suddenly looks a lot less attractive.

(None of this is Trump’s doing either. He doesn’t have a “30-day plan to get rid of Isis”. He has no plan, other than “ask someone else to do it and bluster”.)

IS’s oil income has been plummeting as Turkey in particular has cracked down on illicit sales, and the price of oil itself has fallen considerably since IS got its big break.)

Losing the fight

Which brings us to terror groups wanting to make a splash. Simple way: get airside, put a bomb on board. Or whatever. The photo at the top was sent to me by a source on Twitter who watches this stuff. It was originally tweeted by a Twitter account @poihhp – since suspended. I can’t find any data on the account (age or followers) though the lack of responses to it suggests it is pretty new. As the photo caption above says, the paper seems to say “The Islamic State is right in your home”, and claims to have been taken at Baghdad International Airport.

I’ll admit that my ability to read Arabic is nonexistent (I relied on Bing Translate and my source’s slightly better translation). But that looks like a form of the IS flag scribbled on the right-hand side of the paper. They’re holding it in their right hand. I can’t identify the airlines that the two aircraft are from – they don’t seem to be Turkish Airlines or Iraq Airlines. There’s a list of airlines which go through Baghdad International. I can’t identify them from that either.

It’s possible this is a fake, or a jape. But it feels like there’s something authentic there. And remember: you didn’t know why the liquids ban was introduced in 2006, and you probably thought that was stupid too. (The arrests weren’t announced.)

But it turned out not to be. The reasons behind the carry-on ban are likely to be the same.

How many Apple Music users are on Android?

Apple Music: also on Android. Photo by tualamac on Flickr.

Apple announced on Wednesday that it now has 20 million subscribers to Apple Music after just 18 months – which feels like pretty good progress. Apple Music is also the only meaningful Apple service that’s also available on Android, as Apple tries to spread itself cross-platform.

Why is Music on Android as well as iOS? Because it’s not a distinguishing feature. Unlike Apple’s iMessage or its App Store, both of which are exclusive to Apple, and whose features are unique to it, you can already get lots of music services on both iOS and Android – Spotify, Tidal, Deezer, Google Play Music, Amazon Music and so on. (Not many, if any, of them are making money, but leave that aside.)

For Apple, every additional subscriber to Music is a bonus; it’s all money. The ones on Android are potential converts to iOS, where there’s more money to be made selling them iPhones, iPads and Macs, but time spent on Music is time not spent on Spotify, its principal rival, so that’s a benefit to Apple. App development costs aside, which are comparatively small, Android is a benefit. Additionally, if someone with an iPhone gets a Family Plan (allowing five people to use the same account; a Family Plan only counts as one subscription in Apple’s total), those five could include Android users.

So how many of those 20m are on Android? I’ve been tracking the stats on Google’s Play store for Apple Music downloads since its launch, including the download range and the number of 1-, 2-, 3-, 4- and 5-star reviews.

There are two things we need to work out: how many downloads there have been, and how many of those have resulted in subscriptions.

Download power

To estimate the number of downloads, I track the definite waypoints – when it passed 5,000, 50,000, 500,000, 1m, 5m – and the number of reviews against them.

Apple Music: how many downloads per comment?

The process is very standard across all apps: to begin with comments come quickly, so that almost every download prompts someone to review (about every 27 downloads, someone left a comment up to the 500,000 mark) which then tails off (only one in every 100 downloads prompted a comment by the 10m download mark). You can thus model how the number will change; and so even when you’re in the vague space between waypoints, you can estimate the number of downloads. (I’ve used this technique to calculate Android Wear downloads; it was accurate about when it would hit the 5m mark by a few weeks in a two-year timeframe.) The R-squared value is the correlation, which runs between 0 and 1: closer to 1 is nearer exact correlation.

Based on that, I calculate that this week Apple Music on Android hit 12.75m downloads. (It passed 10m around September 6.) It’s adding about 230,000 downloads per week, at a rate that seems to be holding pretty steady.

But downloads aren’t subscriptions. Apple Music offers a three-month trial period, after which you have to pony up. Clearly, some people will drop out. But what proportion of those who download it stay on and subscribe?

One subtlety here: the three-month trial means that strictly, we should ignore the downloads made in the past three months, since none of them will have qualified to become subscribers yet. (Downloads made to join a Family Plan don’t count as subscriptions.)

So the figure we’re looking at is from 6 September – which just happens to be when Apple Music hit the 10m download mark.

Star quality

Here’s an obvious way to estimate that: look at the high-quality reviews. It seems logical (Captain) that five-star reviews indicate people who are really happy with the service. Four-star reviews are people who are pretty happy, but find some hassles with it.

Apple Music reviews on Android

By the time you’re down to three-star and below, I think you’re talking about people who aren’t impressed and won’t be staying.

Although people don’t become subscribers until three months have elapsed, I think you can include recent ratings, since those could come from people who have become subscribers. (We don’t know what prompts people to review an app.)

So what do the ratings show? At present, five-star reviews are about 44% of all reviews; four-star ones, 11%. That ratio has been pretty consistent; five-star reviews have been at least 38% of all, and average 42% over the life of the app. Four-star reviews go down to 10%, and average 11%. (I don’t know what the average is across all apps on Google Play.) In fact, the data shows a gradual improvement in how the app is perceived, according to the reviews.

Apple Music on Android reviews

Based on this information, we can get some useful minima and maxima.

10m downloads, five-star users subscribe: 4.4m Apple Music subscribers on Android

10m downloads, four- and five-star users subscribe: 5.5m subscribers

12.75m downloads, five-star users subscribe: 5.61m subscribers

12.75m downloads, four- and five-star users subscribe: 7m subscribers.

I’d suggest the useful range is probably the 4.4m-5.5m one.

(One confounding caveat: we don’t know how many of the eager reviews come from people who downloaded it because they’re part of a Family Plan. I can’t think of a simple way to evaluate this unknown.)

Phoning home

If we accept those numbers, it suggests there are 14.5m-15.6m Apple Music subscribers on iOS.

What does that mean in the context of how many phones are out there?

There are about 550m iPhones in use, according to Neil Cybart. And there are around 1.2bn Android phones in use. (Apple Music is available in China, so it can run on phones there.)

This implies quite low penetration for Apple Music: about 2.5% on iPhones, and about 0.3%-0.4% on Android.

Then again, given that Spotify’s last published figure (in September) is 40m subscribers, and it is also available on both platforms, it’s clear that it’s just difficult to get people to sign up to these services. Given how many people Spotify has been available to in multiple countries, it has only been able to convert about 1% of the total available internet population during its life. It seems like getting people to sign up to music streaming really isn’t easy at all. So if you’re an Apple Music or Spotify subscriber, you’re very much in the minority.

Even though Apple’s progress in a short time looks strong compared to Spotify’s over the same period, it’s an open question how big the total addressable market is here. Are we just crossing over from the early adopters to the broader audience who will jump on streaming? Or is it going to struggle to break through? These are still open questions.

Fear of a USB-C planet: why we’re going to miss Apple’s MagSafe

MagSafe 2: you’re gonna miss it. Photo by yum9me on Flickr.

Want to know my pick for the best technology Apple ever introduced? MagSafe. That’s the magnetic attachment system for power cords which it unveiled in January 2006 with the new MacBook Pro. Six years later, it followed that up with the MagSafe 2 connector.

From 2006 until the USB-C MacBook in mid-2015, every portable had a MagSafe connector of one sort or another. (The desktop machines didn’t.) As a technology, it’s so clever and so convenient, being reversible, and an inherent safety feature – you (or your dog/significant other) couldn’t destroy your laptop by accidentally walking into the power cord and pulling it off your desk, or destroying the power socket on the machine, either of which would be expensive to get fixed.

And now, I suspect Apple is going to get rid of MagSafe all across the laptop line, and replace it with USB-C. That’s what the “Hello again” launch looks like to me. (The MacBook Air might be kept on, unimproved and unadvertised, to anchor a lower price point.)

This is going to be a hell of a thing. From April 2006 to the end of December 2012, Apple sold 57m laptops, with – it’s a safe assumption – MagSafe adaptors on all of them. It used to split out the numbers, but stopped at the end of 2012, when the desktop:laptop ratio was running at 20:80.

Since then it has sold a total of 88.3m laptops and desktops; at that 20:80 ratio, one could reckon it has sold another 70.6m laptops.

That means that since the first MagSafe was introduced, Apple has sold a total of more than 120m laptops with MagSafe chargers.

And now it’s dumping them for USB-C.

Oh man. This is going to be a hell of a thing. If you’re like me and my family, then you have a ton of MagSafe chargers, possibly of different generations with the little magnetic doohickey which converts a MagSafe 1 to a MagSafe 2, distributed around your home and office. It’s so convenient not to have to take your charger from home to work and vice-versa; no forgetting. If you upgraded your laptop during the past ten years, you didn’t have to worry about getting a different charger; you could use the same one, or upgrade it. MagSafe is great.

Now? USB-C is coming. But you can’t make it compatible with MagSafe. No doohickey is going to make it right.

I can see that while a lot of people have been waiting for the new Mac laptops, this is going to be a real wrench. It’s as bad as the original iMac in 1998, which ripped up peoples’ investments in ADB plugs and forced them to USB. Except that at least then you could make converters which would get ADB to talk to USB; you won’t be able to get your old MagSafe charger to work with USB-C. (If you could, Apple would have done that with the new USB-C MacBook.)

There’s no doubt Apple is going to go all-out on USB-C with the new models; it wouldn’t surprise me if it revises the entire desktop and laptop line to introduce the new ports. When it does this stuff, it does it for keeps.

I’d love to be wrong on this; I’d love for Apple to have realised how wonderful MagSafe is, and determined to implement it on USB-C. But as noted above, if it could, then it would have done this on the MacBook.

Man, you’re going to miss those chargers. It’s going to take a long time for 10 years’ worth of chargers to go away. And a little part of me hates USB-C that tiny bit more; even though I’m in no hurry to upgrade, I know that when it happens all those chargers will be useless to me. Damn.

(On Twitter, Ken Tindell points out that Griffin – which picked up on the gap left when the first iMac introduced USB – has done a magnetic USB-C connector. Looks nifty.)

Adieu Note 7, advantage Huawei, maybe Apple, maybe not so much Google Pixel

The Note 7 is toast. Photo by strategeme on Flickr.

So the Galaxy Note 7 is no more. Samsung has, belatedly, said that it has halted manufacturing of the device, and that anyone who has one should power it down and return it for an exchange.

Huawei’s going to be an obvious winner; the Google Pixel not so much, because of its limited volumes (I’ve got an estimate below). It’s very hard to see whether Apple will really benefit.

This is fine

US carriers had over the weekend all decided to stop selling it after multiple replacement devices began overheating and catching fire. Having initially moved somewhat slowly to a recall when 2.5m of the devices had already gone out to carriers and end users, Samsung seemed bamboozled by the second stage; for a couple of days it was a deer in the headlights, when it seemed clear to everyone else (independent brand advisors, analysts, etc) that the simplest move was the Tylenol solution – get every damn one off the shelves.

The case study is enlightening:

By withdrawing all Tylenol, even though there was little chance of discovering more cyanide laced tablets, Johnson & Johnson showed that they were not willing to take a risk with the public’s safety, even if it cost the company millions of dollars. The end result was the public viewing Tylenol as the unfortunate victim of a malicious crime (Broom, 1994).

The difference here of course being that Samsung wasn’t the victim of a malicious crime; the fault seems to lie in the combination of the phone and the battery – perhaps there’s something about the fast charging via USB-C plus tiny variations in battery manufacture which makes some batteries prone to thermal runaway.

Ben Thompson points out in his Stratechery newsletter that the US consumer regulator suggested that the battery compartment is too small, which pinches some batteries. That’s probably enough to spark a runaway:

“The dimensions of the materials they put into the pouch were a little bigger than the pouch itself,” [Consumer Product Safety Council chairman Elliot] Kaye said. “By putting that all together and squeezing it into the compartment, it caused some pinching.”

(There’s a clue in the subtly different dimensions: Note 7 dimensions: 153.5 x 73.9 x 7.9 mm, v the
Note 5 (the 2015 precedessor): 153.2 x 76.1 x 7.6 mm. Very slightly narrower, yet with a 3,500mAh battery compared to the 3,000mAh of the Note 5. Bump it about a bit, perhaps a microscopic part of the battery begins thermal runaway, that spreads slowly to another part, pretty soon the whole thing is getting hot… and blam.

Bad karma

The Note 7 will go down in history as the phone which Samsung tried to rush out ahead of the similarly-numbered iPhone (and bizarrely abandoned its own numbering, following 2015’s Note 5 with 2016’s Note 7.). It was announced in August, part of Samsung’s effort in 2016 to get ahead of everyone else by announcing its flagship much earlier than rivals, rather like glossy magazines publishing their Christmas issue some time around October.

Even though no smartphone is truly designed in a hurry, the Note 7 seems to have been the victim of a rush to market. That tiny bit less QA (quality assurance) testing, that insistence on a form factor… while the engineers succeeded in making it waterproof, they failed to make it fireproof.

So now it’s gone, and Samsung’s reputation is effectively being trashed all over the world by the fact of news reports about the end of production, the details of the fires, and so on.

Estimates vary on the damage in terms of lost sales. There were 2.5m already out there, but some reports suggest analysts thought it could have sold up to 19m units over its lifetime. Reset that to zero now.

Lost sales?

Who wins? Or, to put it another way, where will those sales go?

I don’t think any of those initial buyers will go anywhere but Samsung. These are people who wanted the Note 7, and it offers them something (a stylus, split screen, the Samsung name) that they really wanted. It was exceedingly well received by reviewers, though of course that’s barely half the battle; getting it into peoples’ hands is a far more difficult task.

But for the rest? One of the “replacement in flames” people said he was getting an iPhone instead, but he seems like an outlier (he’s American, after all). Far more likely that someone who’s looking to buy a premium Samsung phone is embedded in the Android ecosystem, and so will be looking for something similar – big, perhaps a little flashy, premium.

Their first port of call is likely to be Samsung, but after this mishap it’s going to be tricky. These are people who have clearly chosen not to buy the S7 or S7 Edge, since those have been available for months.

In which case they could be looking for something quite different. Apple can tempt (the iPhone 7 Plus does have that New Phone look, sorta – especially if you get it in Unobtainable Black). But there’s a switching cost in moving from Android to iOS, in time spent finding equivalent apps if not in pure cost. (Though the latter may also apply.)

So who or what else might they choose?

Advantage Huawei

Huawei is the third biggest phone maker in the world, and it is very ambitious indeed. It is going to seek to take advantage of this everywhere outside the US – where it is weak – by pushing its budget. That’s certainly what I’d do if I were them, and they’re a deal more capable at this than me.

Huawei also has attractive hardware at the high end which will certainly appeal to Chinese and Asian customers, and it has been stepping up its efforts in Europe with some success too. It’s just possible that we’ll look back in a few years on this moment and realise that this was when Huawei took the chance to overtake Samsung at the top of the phone tree.

It could happen: if the reputational hit on Samsung is bad enough (and it is already sunk in China, the world’s biggest smartphone market) then it could be a long decline.

Hey, what about the Pixel?

But what about the Google Pixel? The search company launched two new phones recently. Couldn’t they take advantage of this new hole in the market, especially in the US where Huawei is weak (partly because of Google) and Google is strong?

Well, not on the numbers. The Pixel is built for Google by HTC. Thus HTC’s revenues are going to fluctuate partly through the number of Pixels that it has been asked to make. A number of those would have been made ahead of the launch.

Can we estimate how many? Sure we can.

In the first six months of 2016, from January to June, HTC’s revenue drifted downwards compared to 2015 by an average of 50% year-on-year.

Then in July, the drift reversed; it fell only 14%. In August, 4.5%. In September, it was actually up compared to 2015 by 31.3%.

Given that the HTC Vive VR headset hasn’t set the world aflame, I’m going to hypothesise that all the extra revenue (ie everything above a 50% decline y-o-y) in July-September comes from making Pixel handsets for Google.

HTC revenues changes year-on-year by month

(Yes, you could argue that the revenues seem to suggest a rebound after February – but the Vive went on sale in April, with preorders opening in February, which would explain that too.)

On that basis, HTC has received NT$11.5bn, or about $363m, from Google to make the HTC Pixel – sorry, the Google Pixel.

How many handsets does that translate to? It depends on how much you think it costs to make them. If you take the IHS estimate for the manufacturing cost of the iPhone 7, of $220, then HTC has made 363m/220 = 1.65m Pixels.

However, that seems like an absurdly high manufacturing cost. So let’s try $110 instead – given that Google is selling them for the same price as an iPhone, that would certainly help profits.

At $110, our $363m translates into 3.3m handsets already made and filtering into sales channels as you read this. I’d expect the true figure lies somewhere between these two – I could have overestimated the money put into this manufacturing effort, and if Google really is charging $600+ for a $110 handset, it’s bringing chutzpah to a whole new level.

So let’s say it might have around 2m to 3m handsets available over the next few months. Is that going to sate those seeking an alternative to the Note 7? Probably not – as a handset it’s OK, but it’s not YUGE like the Note 7, and it doesn’t have a cool stylus or all those other things.

Also, 2-3m handsets (built over three months) isn’t going to be enough to fulfil all those would-be Note 7 buyers, who are reckoned to number 19m over the next 12 months, remember. Google can keep the HTC factories working, churning out a million or so a month, but it would be very surprising if all of the 17m or so Note 7 buyers go for the Pixel instead.Google may well get a good slice of those buyers – more, in fact, than it was ever expecting, given the market it thought it was going to be launching into – but I don’t think it’ll capture all of them by any means.

(In passing, notice that that $363m only gets you to first base in this project: having some phones built. It doesn’t cover the R+D, sales, marketing or administration. You’re easily talking half a billion dollars here to do this thoroughly. And that’s for a phone which is only going to make a marginal impact on this quarter’s sales.)

And back to Samsung

In the end, most of those early Note 7 buyers are probably going to buy another Samsung. But that’s not what its management should worry about – nor the $2bn exceptional item that’s going to be sitting on the balance sheet for the next quarter.

It’s this: what happens when it announces the Galaxy S8 next January or February? People are going to ask whether it explodes, and Samsung will say of course not, and then inevitably there will be a case where it does, because manufacturing defects are inevitable. Hell, there’s even been a claimed instance with an iPhone 7, and Apple doesn’t use fast charging or USB-C and doesn’t squeeze high-capacity batteries into small spaces. It’s pretty much certain that a Galaxy S8 will go off. Then what?

And this is even before the media has gotten onto other flaming Samsung appliances, such as its washing machines in Australia. (Dedicated Overspill readers knew about this a while back.)

The reason why this is a particular problem now is that the smartphone market has pretty much peaked; in the premium market, replacement cycles are lengthening and sales are shrinking, so anyone that Samsung doesn’t capture now might be lost (as a potential sale) for another two or three years. That’s a substantial amount of foregone profit

The upside for Huawei looks big. The upside for Google (Pixel) looks like hitting sales targets. The upside for Apple looks like some more iPhone 7 sales, though I wouldn’t hang your hat on it.

The downside for Samsung, though, looks big. Events like this can make and break a company, and this is going to be its more serious test ever.

How to recover when Apple Numbers says a file “can’t be opened for some reason”

Do not use this as an error message
Ever wondered what the worst error message you could encounter might be? This ranks pretty highly.

Prologue: backup

First of all, remember how people always told you to take backups, rather as you were advised to wear sunscreen? Well, they were right. Bear that in mind as I take you through on a journey of mild technology pain.

The high: Sierra

Having not seen any reports of gigantic showstopping bugs in the upgrade to Mac OS Sierra, I took the plunge the other day. Things were going fine. Everything worked. Nothing had crashed. Then I updated Numbers from 3.6.2 to 4.0, whose “new” features are apparently collaboration – and nothing much else.

Having done that, I tried to open one of my most-used spreadsheets, into which I have poured years of experience and hours of analysis. I’d had it open before the update, but (I think) had closed it before updating. (Whether it was open or not is immaterial; some other spreadsheets were open before the update and opened fine afterwards; some were closed before the update and opened fine afterwards; some were closed before and wouldn’t open afterwards.)

I was met with this response:

Screenshot 2016 09 26 15 06 13

The low: Numbers

The spreadsheet can’t be opened “for some reason”?? What sort of error message is that??

But at least it offers the option to “Browse all versions”, which should be stored in iCloud, where the spreadsheet itself is stored. You then go into a Time Machine interface, and get this:

Screenshot 2016 09 26 15 07 23

It’s “unable to open version”. This happens no matter how far back you want to go. You can try with lots of “versions”. Or you can realise you’re onto a lost cause and give up. At which point the “Time Machine” interface resolves itself into a rectangle, in which you find this message:

Screenshot 2016 09 24 20 57 46

Well, thanks a lot. “For some reason.” How this ever got past any sort of quality assurance I cannot imagine. Did the engineer/s assign an out-of-bounds error code to the problem, and the operating system can’t decide what to say and so falls back to “for some reason”?

This is a giant screwup

Whichever; it’s a terrible, terrible experience for the user. You’re left unable to open the file, with no idea what has gone wrong, and no clues how to progress. If you had really valuable stuff in here, and no means of rolling back, you would be absolutely furious – justifiably so – with Apple.

Tracing the error

What has gone wrong here? You can dig into iWork files (Numbers, Pages, Keynote). They’re “packages”, which means that they’re folders disguised to look like files. Control-click on the file and you can “view package contents”, which in the case of this spreadsheet looks like this:

Screenshot 2016 09 26 15 34 49

Turns out that all the meat is in “”. I made a copy on my Desktop and unzipped it:
Screenshot 2016 09 26 15 36 43

That’s only a few files; the “Tables” folder contains 523 items. Which of these hundreds of items is at fault? One? Two? Two hundred? There’s no way of knowing. Given that none of the previous versions will load under this version of Numbers, it doesn’t matter how many of the files are screwed. You can’t get there from here.

Why you love your backups

I did try to get around this. Believe me. On an iPad (which hadn’t updated to the equivalent newer version of Numbers) I tried opening the original spreadsheet.

Opened fine.

Oh. So I tried AirDrop to send the can-be-opened-from-iCloud spreadsheet from the iPad to the Mac. The AirDrop worked, but the Mac wouldn’t open it – same message as before. On the iPad, you can also export the file: your options are Numbers, Excel, PDF, or CSV.

Export in Numbers and AirDrop? Didn’t work.
Export in Excel and AirDrop? Worked – except that the various tables that had been on a single sheet were split out into separate sheets. Non-ideal.

So the iPad route wasn’t quite right.

But I wasn’t finished yet. Did you notice how I mentioned backups? Before upgrading, I had made a backup of my hard drive using SuperDuper! (highly recommended).

So I plugged in my backup drive – I’m always careful not to overwrite it until I’m confident a big OS update hasn’t screwed anything – and dug around for the old version of Numbers (v 3.6.2), and put that back in.

Open Numbers 362, try to open spreadsheet.

It opens. No muss, no fuss.

Worse than error messages: no error messages

In many ways, this is even worse. What’s the situation here? We have a newer version of Numbers on the Mac which cannot open an untouched version of a spreadsheet that the older version can open.

Together with the colossal stupidity of “for some reason” as an error message, a new version that randomly can’t open an old spreadsheet (but is fine with many others), even while the old one can, makes one think that whoever is in charge of Numbers, or iWork, isn’t getting it right.

A lot of it is down to the error message. If it said “because two of the files are corrupt” you might begin to understand. But of course it can’t be that, because the old version can read it. “Some reason” sounds vague – is vague – but in a sense, it’s accurate. Whatever the reason for being unable to open this file is, it’s quite elusive. I had initially thought that it was something to do with picture embeds, but the problem persisted when I got rid of those. (There’s nothing in the Console app about it, so Numbers clearly doesn’t want to share whatever its discomfort is with the file/s.)

Anyway. Having got the old version of Numbers installed, I could now open the old spreadsheet. Fine. I’ll stick with that, I thought.

The morning after the night before

Problem over, you think? Not at all. On returning to the iPad the next day, I found it had updated to the newer version of Numbers – the one with collaboration.

Guess what? That’s right: the iPad version no longer opened the old spreadsheet.

Computing often has these moments – when you feel as though you’re standing on a very rickety rope ladder across a gigantic chasm, halfway from each side, with little prospect of reaching either side safely, yet obliged to go in one direction or the other. The previous day I could open the spreadsheet on the iPad, but I couldn’t get it safely back to the Mac. Now I could open it fine on the Mac, but I couldn’t get the iPad to read it. Not really the world of device-independent operation that one dreams of.

But but but! There is a solution on the Mac. You can load the file on the old version of Numbers, and then in the File menu there’s the option to export it to a Numbers ’09 format. (No idea what’s so great/terrible about that.) Notice that that export option wasn’t available on the iPad.

Here’s what it looks like:

Screenshot 2016 09 26 16 14 52

Worth a try, I thought. And indeed it was. I named the files created that way with an “09” suffix, and suddenly they opened on the iPad – with all the tables and charts intact.

Update: another tactic which I didn’t try, but which might work (I haven’t had the same problem again) is to log in to and try to open or upload or similarly wrangle the file there. Make sure FIRST you have a backup of it, on a USB key or other cloud service; the greatest mistake is working on the only copy of an essential file.

Teachable moments

This is one of the biggest WTF moments in an episode replete with them. I’ve reinstalled an older version of Numbers, and exported to an even older file version, in order to open the file on the newest version. It’s beyond bizarre.

Thankfully, it seems that there aren’t too many people having this problem; my own searches on the phrase “can’t be opened for some reason” turned up pretty much nothing. If we’re all lucky, then nobody will land on this page via a web search; you’ll all just be reading it for abstract interest, wondering how an operating system and a QA team can ever let “can’t be opened for some reason” be signed off as “OK for public consumption”. Apple puts a premium on its products and prides itself on its user interface; this, though, is one that got away, badly.

But what if you haven’t kept that backup of the Numbers app? In that case, I’m not able to offer any help. Perhaps you can find a friend who has a copy of the older version. Perhaps there’s a trustworthy download site. Perhaps you can get one by finding a Mac that hasn’t been updated and sending the version there. Perhaps you can rummage around in your Time Machine backup and reanimate the old version. Maybe you have a CD in your house with an older version. (Clutching at straws here, but I recognise that spreadsheets carry a lot of our lives nowadays.)

The simplest solution is not to update Numbers, which of course always feels like admitting defeat. The pragmatic solution is to export all your spreadsheets to the 09 format. The belt-and-braces solution (since this might be an iCloud problem) is to duplicate your spreadsheets on your hard drive, and export each into the 09 format – then you have three copies of them.

Whichever – I hope it goes well. And I hope never ever to run into “some reason” as the explanation for why an essential piece of content can’t be accessed. Fix it, Apple.

Is Android Wear going to make Google create the next Zune?

This isn’t an Android Wear watch, but there are similarities. Photo by David Lee King on Flickr.

Once upon a time there was a Big Software Company, and it wrote some Software for a Task, and that Software could run on Devices made by Other Companies. However, the Big Software Company didn’t want to get into the Devices business in any serious way; it much preferred to let the Other Companies deal with the (low-margin) hardware side of business, and stay principally on the high-earning Software side. (The Big Software Company did make some hardware, but it really didn’t make much money on it. In fact it even made a loss at times.)

Meanwhile Apple also began offering a thing for the Task, competing with both the Big Software Company and the Other Companies’ Devices. Except that Apple wrote its own software and built its own devices, which meant that it could extract profit from whichever part of the value chain it could, and subsidise the bits that weren’t profitable by its end pricing. If writing Software was a cost (which it is), it could make up for that by screwing down suppliers by buying up in volume, or raising prices subtly. The relationship between the Big Software Company and the Other Companies didn’t allow for that dynamic.

Pretty soon, Apple dominated sales of devices dedicated to the Task. This didn’t please the Big Software Company, which decided that the Other Companies needed someone to show them how to do it properly, and set about building its own Devices using its own Software to do the Task.

At this, the Other Companies looked askance at the Big Software Company, and essentially told it to go screw itself.

Not a moment too Zune

Reading this, you may – if your memory is long enough – have thought “oh, it’s the story of Microsoft, the MP3 makers, the iPod and the Zune.” And that story certainly fits. Microsoft wrote the Windows Media Player and controller software, which then had to synchronise with the music stores and with the music players made by OEMs, all with the intention of giving the customer the most delightful experience.

The reality was rather different – it was so hard to keep all the software and hardware synchronised that you never knew who to blame. Did your player not work because the OEM had been lazy in implementing Microsoft’s software, or because Microsoft had somehow screwed things up?

There was also the problem of marketing and advertising: despite having a powerful brand, Microsoft couldn’t really favour one OEM over all the others, so they all had to fight for themselves. You got different devices, different sizes, and no clear story about what to do.

Apple, meanwhile, came through the middle, launching into the MP3 player market just as it was ready to expand (in 2001, when MP3 was big on the desktop but not in the pocket) with a clearly identifiable design and really easy-to-use software. (Yeah, yeah, I know, you hate iTunes. Believe me, there was a time when it wasn’t overloaded.)

The rise of the iPod left other MP3 makers floundering; they just couldn’t get traction, because they couldn’t have a single message (they needed to differentiate themselves from each other) and they often didn’t update in sync. Apple, meanwhile, ruthlessly exploited advertising gaps, celebrity endorsement and an interested media (which ran endless “iPod killer” stories, the first in October 2002 – it was attached to a Creative Technologies player).

This went on from 2001-2006, by which time Microsoft decided that it had had enough of these messups – Steve Ballmer, then CEO, was particularly annoyed by Apple’s dominance – and so it made its own, the Zune. The project started in March 2006 and launched in November that year. You can guess that the timing wasn’t fortuitous; the Zune arrived as the MP3 player market was going into decline. It’s often overlooked that in the same January 2007 video where Steve Ballmer dissed the iPhone, he also talked up the prospects of the Zune:

(Jump to 1’00” for the iPod discussion)

(You can read more about the whole Microsoft/Windows Media/iPod/Zune debacle in my book Digital Wars. Also has chapters about search, smartphones, tablets and China. UK Amazon paperback, UK Amazon Kindle, US Amazon paperback, US Amazon Kindle, iBooks.)

Not only but also: Android Wear

However, you could read the above story and replace “Microsoft” with “Google”, the Task with “smartwatches”, and all the OEMs with… the OEMs who have struggled to sell Android Wear watches.

And it is a struggle. Last week, Android Wear finally passed 5m downloads on Google Play. As I understand it (from sources), each download is an activation of an Android Wear device.

Here’s how the download figure, using the waypoints on Play, has progressed since the release at Google I/O 2014, when all the attendees got one:

Android Wear downloads/activations by week
Android Wear downloads, definite waypoints; the polynomial fit is pretty good, according to the correlation coefficient (R-squared: 1 = perfect correlation, 0 = none).

That’s actually quite encouraging: it seems to suggest that after a slow start, sales are just beginning to take off.

And now here’s how it looks if you also include the number of reviews for the app on Google Play. The trend there suggests that reviews led downloads – that is, the ratio of reviews to downloads wasn’t steady, and has shifted down, so that over time fewer people who download the app leave a review. (I use that approach in my modelling to estimate download growth.)

Android Wear downloads v reviews
Android Wear downloads, and reviews left on the app page. (Some data for reviews missing.)

So, 5m sold in 115 weeks (just under 30 months). Here’s the problem, though: those 5m have been divided up between LG, Motorola, Huawei, Asus, Acer and the various other bit players in the Android Wear space.

Enter the Watch

But that’s before we consider Apple’s offering here. Apple doesn’t release official figures for individual sales of the Watch (“competitive reasons” is the line, and they’re sticking to it). According to Canalys though in its first two quarters of availability – April-October 2015 – it sold 7m. That was probably split 4m + 3m, given the estimates for a 4m first quarter.

And for the Christmas quarter, estimates (again, necessarily) put sales at 5.1m. And then for the first quarter of 2016 they were estimated (again) at 1.5m – a long way down, for certain, but not surprising for something which isn’t a necessity (like, say, a phone in the modern age) but more of a gift. A bit like iPods used to be, in fact.

So that takes us to 13.6m by March. Since then there have been two more quarters – well, we’re nearly through September. In the second, Apple was reckoned to have sold another 1.6m amid a widespread decline in smartwatch sales. Probably sales during this current about-to-end quarter have been similar, or even lower.

That’s a total so far of around 16.5m in five quarters. Those aren’t iPad-style numbers – that achieved 28.7m sales in its first five quarters – though it’s better than the iPhone (6.1m) over the same time from launch. (The iPod did 0.6m in its first five quarters, but they were different times.)

Here though is where Apple Watch/Android Wear starts to look like iPod/MP3 players. Apple gets all the revenue from the Watch, and it controls the software, and it decides how it’s going to move the business along. That’s why Apple’s trajectory looks a lot stronger at present:

Apple Watch sales v Android Wear activations
On cumulative sales, Apple Watch is outpacing Android Wear by a substantial margin.

By contrast the Android Wear OEMS don’t even get to personalise Android Wear; they just make the product, and still have the software support costs (those driver updates don’t write themselves) without getting any sort of side benefit. It’s not even as if using one model ties people to that brand of phone; interchangeability is the model with Android, unlike Apple. For the consumer who wants to pick and mix, it’s fine, but as a business model it brings some pain.

That’s clearly why, as Roger Cheng found out at CNet (by the shocking tactic of Asking People), the big Android Wear OEMs are going to hold fire for now:

LG, Huawei and Lenovo’s Motorola unit will not release a smartwatch in the waning months of the year, the companies confirmed to CNET. While LG launched a watch in the first half, it’ll have been more than a year since Huawei and Motorola offered an update on their wearables.

That marks a reversal from last year, when all three companies launched Android Wear smartwatches at the early September IFA trade show in Berlin in what was supposed to be a resurgence of the platform. At this year’s show, Chinese maker Asus was the only major tech company to return with a new Android Wear watch.

That’s perhaps a bit discouraging, since aside from chip companies, you’d struggle to see other OEMs in the big list on the Android Wear site:

Android Wear partners
Not many of these are selling Android Wear watches as their primary business. Or even secondary business.

According to Wareable, there are somewhere between 10 and 21, perhaps 24, Android Wear watches to choose from. (Its guide is pretty handy.) You can also get an idea of pricing, which tends to be around £200; again, that’s quite a bit less than Apple’s Watch used to be, though since the introduction of the new version the price for last year’s model has fallen to £269.

Altogether, Android Wear vendors have taken in a total of 5m x £200 = £1bn ($1.3bn). That’s spread across multiple vendors – let’s assume four of them – over 30 months. It’s not a great return on investment; it’s about $10m per month each, or $120m per year. Apple, in the meantime, has probably taken in $350 x 16.5m = $5.7bn.

(Sanity check: Tim Cook said Apple was second only to Rolex in watch revenues for 2015, and Rolex did $4.5bn in sales in 2015, and the sales estimate above is 12.1m; at $350 each that’s $4.2bn. That’s also ahead of Fossil, which did $3.2bn. So this all fits with what we know.)

This is the inherent weakness, for a device that needs to be personal, in the modular business model: OEMs can’t control enough of the story around it. Samsung’s decision to go its own way with Tizen looks sensible; its use of the watch bezel for a control is an inspired little bit of UX.

See you zune?

In the face of indifference by the OEMs, though, what’s a Big Software Company to do? Well, perhaps what Big Software Companies do: roll its own. Google is said to be working on not one but two Android Wear watches of its own, set for release very soon. One is big and one is smaller, and it’s reported (by Android Police) that the larger one will have LTE connectivity as well as GPS and a heart rate sensor, and be a “do-all device that will allow Google to demonstrate Wear’s most robust capabilities, including the announcement at Google I/O that Android Wear 2.0 will support standalone apps”.

The smaller one, meanwhile, won’t have LTE or GPS and maybe not even a heartrate sensor. Price unknown.

If they’re going to happen, then they’ll come with the October 4 announcement of the new Pixel phones. Google has already indicated that it’s going to try to move beyond the Nexus line (something which has occasionally annoyed its Android OEMs; that’s part of why the Nexus phone line has never been big in sales terms). Perhaps it has decided the same for smartwatches. I guess we’ll find out zune enough.

The 2Q 2016 smartphone scorecard: players searching for an exit

Exit. Who’s next? Photo by Today is a good day on Flickr.

There comes a time in every former top-ranking sports player’s career when they have to accept reality: they’re not up to it any more. They keep getting beaten by people whom they once would have trampled; what should have been easy wins are now struggles, or upsets. Eventually, they accept the reality everyone else has already seen: it’s time to exit.

And now we’re seeing that happen in the smartphone market. This isn’t really about sales of iPhones being down year-on-year – though they are, for the second quarter in a row, and though in the previous quarter Apple managed to keep its handset ASP (average selling price – calculated by total handset revenue divided by the number of handsets) up, in this quarter it was substantially down, below $600 for the first time since 2Q 2014.

But more generally, this is the quarter where China really began to muscle into the top ranks of Android OEMs – and all the players who used to be the big names there are inching towards the exit. The problem for the big-name Android OEMs is that, because it’s Android, they’re replaceable. Android on one handset is quite a lot like Android on another. But an Apple device, and its integrated software, is sui generis.

Numbers for all

So here are the numbers showing how that replacement is going. The list below is all in diminishing size of handset shipment volume. Other data sometimes has to be estimated, and in the case of Huawei, OPPO and vivo you’d have to be in one of the big analyst camps to know what their ASPs and hence revenues are, and you might have to be at the companies to know whether they’re profitable.

Standout elements from the quarter: Sony made a profit! (Even as it dwindled.) Lenovo kept shrinking; Apple’s ASP fell; Samsung trundled on; LG made more losses (the G5 flagship essentially sank); Microsoft barely turned up.

Q2 2016: the smartphone scorecard

* denotes estimate: explanations below

Company Handsets
Revenues Handset
% profit/loss
Samsung 77.0 $22.61bn $275.64* $3.75bn $48.66* 16.59%
Apple 40.4 $24.05bn $595.26 $6.71bn* $166.09* 27.9%*
Huawei 32.1 $7.06bn* $220 positive? positive? positive?
OPPO 22.6 $4bn?? $177* positive? positive? positive?
vivo 16.4 $3.7bn?? $225.60* positive? positive? positive?
ZTE 14.7 $2.5bn?? $170* ?? ?? ??
Xiaomi 14.5 $2.28bn* $150 negative? negative? negative?
LG 13.9 $2.88bn $207.52 –$177m –$12.73 –6.15%
11.3 $1.71bn $150.97 –$163m –$14.42 -9.53%
Sony 3.1 (not a misprint) $3.64bn $582.26 $4.03m $1.30 0.11%
HTC 2.3* $0.5bn* $217.39* –$128.50m –$55.87* -25.7%
1.2 (not a misprint either) $0.23bn* $190.80* –$45m* –$38* –19.56%
Everyone else 135.4m

Samsung: 6m tablets sold for $175 ASP at zero profit; 11.4m featurephones sold for $15 ASP at zero profit. (For every $1 fall in featurephone price, smartphone ASPs rise by $0.14 – so with zero featurephones and 6m $175 tablets, smartphone ASP would be $277.84. For tablets, every $5 rise in ASP lowers smartphone ASPs by $0.38 – so if tablets were free and there were no featurephones, smartphone ASPs would be $291.37. It isn’t a huge difference; tablets and featurephones are together generate about $880m, or less than 5% of overall mobile revenues.)

Apple: operating profit calculated at the historic figure of 27.9% (derived from multiple financial analysts). Might have been lower or higher – the 6S range maybe costs more to make than the 6 range, but there’s the SE range which might be cheaper because less retooling needed.

Huawei, OPPO, vivo, ZTE, Xiaomi: ASP figures all estimated, based on their perceived market power

How do I calculate the revenue figures (and hence ASPs) for OPPO, vivo, Xiaomi? According to According to Strategy Analytics,

Global Smartphone Industry revenues declined by -5% YoY in Q2 2016, due to softening of volumes. Apple was followed by Samsung, Huawei, Oppo and vivo from a revenue perspective. The report also captures the Wholesale Average Selling ASP’s for all major vendors across six regions. ASP’s in the quarter declined by -6% globally.

So if Oppo and vivo were bigger than Sony, they must have done more than 3.64bn. (Xiaomi must have been less than them too.) I’m guessing they weren’t that much bigger. For Huawei, which like those two doesn’t release revenue figures, I’ve estimated an ASP (up from the previous quarter) and generated the revenue figure from that.

LG: assume tablet sales were minimal, and had zero profit.

HTC: given that it now sells the Vive headset too, though not in large numbers (certainly not millions), it only takes a small adjustment from the overall revenue.

Microsoft Mobile: Microsoft gave figures for featurephone sales, of 9m; assuming an ASP of $15 for those and gross margin of $5 each (as before) gives the featurephone revenue. Assume the same manufacture cost as before, and you get zero gross margin; even with zero sales/marketing and R+D, you get a negative margin.

Rampant deflation

Everyone’s seeing price declines, which is what you’d expect in a growing market where you also have Moore’s Law and scale coming into play. But this is barely a growth market. Smartphone shipments were up just 0.26% year-on-year. When you look at the trend over the past nine years, we’ve really hit a wall here:

Smartphone growth year-on-year.png

The red line shows the four-quarter moving average, and that’s clearly down. What that suggestion of slowdown doesn’t quite tell is how the market is diverging. The premium end was long ago saturated: people who could buy expensive phones did so, but now there’s no new market to sell into in the developed countries – and consequently the US, China and western Europe are expected to see slowdowns, and even reductions in volume, this year (per IDC). The action, such as it is, will be in emerging markets such as the Middle East, Africa and Latin America – though even they will only see growth of about 5.6%.

In such a world, the companies which initially made Android a Huge Thing are beginning to head for the exit. HTC built the first Android phone. Sony had to go Android (as Sony Ericsson) because it was losing money hand over fist. LG had to figure out how to make smartphones quickly, because its featurephone business was being destroyed.

Now though they’re seeing those be destroyed all over again. You can see the numbers above. And here’s a graph of how pretty much everyone is seeing sales growth compared to the smartphone market turn negative (so if the market grows 10% and they grow 5%, they’re falling behind):

Smartphone OEMs: growth against the overall market

Year-on-year shipment growth measured against the overall market

But I’ve been collecting the revenue and profit/loss numbers too (and publishing them) going back to Q4 2014. That’s seven quarters. What if you add that up?

Seven quarters of hurt

Here’s the lineup when you calculate it over seven quarters:

Seven-quarter smartphone scorecard covering Q4 2014 to Q2 2016 inclusive

(all estimate elements as above)

Company Handsets
Revenues Handset
Total operating
% profit/loss
Samsung 555.4 $158.70bn $285.74 $17.95bn $32.32 11.31%
Apple 401.07 $263.59bn $657.22 $73.62bn* $183.56* 27.92%
Xiaomi 116.92 $18.62bn* $159.25 ? ? ?
LG 102.75 $21.58bn $210.02 –$428.39m –$4.17 –1.98%
121 $17.44bn $144.13 –$1,114m –$9.26 –6.39%
Sony 47.8 $17.13bn $358.37 –$908.33m –19.00 –5.30%
HTC 26.1 $6.45bn* $247.13 –$717.51m –$27.49 –11.12%
41.3 $5.76bn $139.47 –$2,621m –$63.46 –45.50%
(Huawei, OPPO, vivo and ZTE aren’t included because I don’t have figures for them over the period; and there aren’t any financials for any of them.)

This bears out a truth that is borne out again and again by analyst reports into best-selling handsets, brand loyalty, and customer satisfaction: these days it’s a two-horse race, Apple and Samsung.

Xiaomi is an unknown, financially. But all the rest are losing money hand over fist, and as Vlad Savov wrote in a terrific piece entitled “Android OEM death watch: Sony, HTC and LG edition“, you do wonder why they soldier on:

The Android ecosystem has never been more diverse than it is today, but I suspect that what we’re witnessing now is a peak from which the basic economics of a maturing smartphone market will rapidly drag us down. Niche players like Nextbit, Vertu, and BlackBerry might survive thanks to their low volume of sales and correspondingly limited costs. But the big names we’ve known for so long, the Sonys and HTCs of this world, seem fated to fade from view.

I think this is absolutely right. Look at those numbers: why is LG putting up with a division that has lost money, and shows no sign of stopping? Although Sony made money this quarter, it’s fading from view. Lenovo’s ASP is so woefully low that it’s an obvious target for every up-and-coming Chinese OEM. (I was recently contacted by Meizu, which is launching into the Asian market: yet another rival for the uncommitted phone buyer.)

It isn’t even as if these struggling companies have scale: Sony has only sold 12% as many phones as Apple over the period (and 8.6% as many as Samsung, which might be the better comparison); LG has managed a more respectable 18.5% of Samsung’s number, but it’s losing money on them, over seven quarters.

Sure these companies have a lot invested in this business; you can’t just shut down a smartphone business like closing a corner shop. There are contracts, staff, distribution deals. But you can edge out, which is what Sony seems to be doing as its range and distribution shrinks. Will LG follow, or is its rivalry with Samsung in Korea just too strong to let it ever let go?

I’m honestly puzzled by companies which tot up millions in red ink and decide it’s fine to carry on. Microsoft is clearly getting out (who wouldn’t, looking at those margins) but how can Sony or Lenovo look at their returns and feel they’re OK? That’s the puzzle here.

Sure, there’s lots else going on: Apple’s falling ASPs and falling sales point to the saturation of the markets. Equally, the cheap hardware is getting really good – the Shenzhen effect, as volume of production means that the only distinguishing thing is software and, to a lesser extent, chip design ability. (Apple, Samsung and Huawei stand alone here.) I’m certainly impressed by Huawei, which offered a dual-lens camera on the new P9 which has a neat refocus/re-aperture effect, well ahead of Apple.

(Huawei’s problem is it doesn’t have a coherent strategy: it offered “3D Touch” before Apple too – as did ZTE – but hasn’t followed through; only the latest P9 still has it. Will the dual lens offering spread to the rest of its offerings, or fall by the wayside as happened with HTC’s dual system on the M8 in 2014?)

In search of the lost profits

What then happened to all the profits that HTC, Lenovo, and Sony used to earn? Simple: eaten by Samsung, Apple, and Chinese rivals. The growth of companies like OnePlus, Meizu, and of course Huawei, vivo and OPPO and (less so) Xiaomi means the potential for scale falls away from those already in the market.

However it can take a while for these effects to become visible. HTC’s sales peaked in 2011; LG’s, Sony’s and Microsoft’s in the second half of 2014. From around that time, all the Chinese OEMs began growing rapidly, first in their home market, and then India; and in Huawei’s case, Africa, Europe and the US.

Late exit

Apple looks to have peaked in 2015 – but it has a solid ecosystem and so many users that any erosion would take a long, long time. That’s in stark contrast to every Android OEM, which (as even Xiaomi is finding out) is disposable and replaceable.

But it can take a long time. BlackBerry’s handset sales peaked in 2010, and yet it’s still going. (Though will John Chen finally announce the company is getting out of hardware at the quarterly results on September 28? One to watch.) HTC has been ebbing for a while, for example. Sony has begun withdrawing to Asia. LG is being pushed aside in Europe by Huawei.

The only question is when some of the executives at these companies will finally ask why they’re still trying to play a losing hand. There comes a time for the players to leave the game. When is it?

The iOS 10 changes that actually matter: ad tracking, camera changes, “press to unlock” and more

It’s that time of year! Photo by fldspierings on Flickr.

It’s iOS 10 release day, and everyone and their best friend is doing “10 [geddit??] things you need to know about iOS 10”. Most of them aren’t worth knowing, because

• you’ll discover them immediately when you update
• they’ve already been announced.
(Though I do love “how to update to iOS 10” stories. TL;DR: do an iCloud backup, or an iTunes backup, and then press the “software update” button in Settings → General → Software Update. Then wait while the internet falls to its knees.)

Let’s instead go a little deeper into the new OS, and point out the elements which you might not spot at first but which could potentially make a significant difference to your experience. I’ve been using iOS 10 through the betas on an iPad Pro and an iPhone SE, so that’s both the phone and the tablet experience.

Ad tracking

Remember how Apple introduced “Content Blockers” in Safari in iOS 9, and in parallel introduced “Safari Web View” for all apps – which meant simultaneously that you could install a mobile adblocker, and that that adblocker could be used in any app which opened web pages (such as Tweetbot, my weapon of choice for Twitter)?

The ad business had a collective fit over iOS adblocking, and it’s ready to have a second one now. Dean Murphy, who profited handsomely (and rightly so) from his Crystal adblocker, points out that with iOS 10, Apple is taking your ability to block targeted advertising one step further, even if you don’t want to install an adblocker.

On his blog, Murphy explains that “Apple is changing the way that the ‘Limit Ad Tracking’ setting works in Settings → Privacy → Advertising, and it seems to be causing a mini storm in a teacup among the adtech world.”

As he points out, while Apple got rid of the “UDID” (Unique Device IDentifier) for iPhones some time ago, in iOS 6 it provided the IDFA – ID For Advertisers. If you turned on “Limit Ad Tracking”, you’d be given a random new IDFA, plus a flag would be set telling advertisers you didn’t want to be tracked. But guess what! Advertisers don’t seem interested in saluting when that’s run up the flagpole.

So, says Murphy:

In iOS 10, when you enable “Limit Ad Tracking”, it now returns a string of zeroes. So for the estimated 15-20% of people who enable this feature, they will all have the same IDFA instead of unique ones. This makes the IDFA pretty much useless when “Limit Ad Tracking” is on, which is a bonus, as this is what users will expect when they enable the feature. These users will still be served ads, but its more likely they will not be targeted to them based on their behaviour.

This didn’t stop one guy over at Ad Exchanger wailing that Apple is “giving consumers a way to opt out of advertising altogether” (it’s not) and that people shouldn’t have the right to opt out of advertising. Which is quite a stretch. Murphy has some more figures on how much the adtech people aren’t losing by this move. But it’s still a good one by Apple, which fits well with its privacy story.

Open the camera, Hal

So you lift up your iPhone to wake it – did every other article mention it now has “lift to wake”? Yes they did (it’s triggered by the orientation sensors) – and now you have a screen with three little dots at the bottom. You’re in the middle; swipe right (that is, pull from left to right) and you get a ton of widgets.

But swipe left (pull right to left) from the home screen, and you now get the camera. This is such an obvious and timesaving move that it’s amazing it has taken four iterations of the “swipe” motif introduced with iOS 7 (7, 8, 9, 10 – that’s four) to get it right.

The Lock Screen in iOS 10 now shows you that the camera is off to the right (ie, swipe left). My arrows and text, obviously.

Having the camera a swipe left from the lock screen is quick, easy and a hell of a lot more convenient than having to swipe up, as has been the case since Apple introduced that route to the lockscreen camera in iOS 5.1 in March 2012.

You can understand why iOS 7 didn’t change that. People had had less than 18 months to get used to “swipe up” when iOS 7 was released in September 2013. Apple doesn’t do UI changes all at once. It taught people how to swipe, then a year later it introduced bigger screens where they’d need to swipe. So we’ve now had “swipe up for the camera” for just over four years. But it’s logical, and faster, to swipe left: it’s a shorter distance, it’s more natural for your thumb (I always found “swipe up” a struggle if I had the phone in one hand), and that screen on the right is unused virtual space.

So all hail the new way of getting to the camera. Though in iOS 10’s first few weeks you’re going to hear lots of people saying “how do you get the camera?” and probably swiping up to Control Centre – though the camera is there. But be the helpful one, and show them the side swipe.

Not quite better: Control Centre/r access

I don’t know about you, but if I’m typing something in Messages and need to bring up the Control Centre, it’s akin to an Olympic event to raise it first time. More often I hit a few random keys first, and have to retry.

Pulling up Control Centre is tricky
Pulling up Control Centre is hit-and-miss if you have a keyboard running

This doesn’t seem any better in iOS 10; I think it needs some sort of border below the keyboard. It’s a difficulty that seems to have come in with iOS 7, so perhaps in a couple of years..

The other change in Control Centre (I’m going to use the British spelling dammit) is that it’s now split into two panes, which you swipe between as needed: non-audio stuff in the left, audio stuff (such as music playback and audio output direction) in the right.

Control centre
The new Control Centre in iOS 10 is split across two screens – swipe between them. It remembers which one you last accessed.

Update: I’m told by Ravi Hiranand that the Home app gets its own Control Centre screen, if you have it functioning. As I’ll explain below, I didn’t so I didn’t. (End update.)

This is another thing that will have lots of people saying “hunh?” as they try to get used to it; since iOS 7 (when it came in) it had been all in one place, but with the introduction of Night Shift on the iPhone 5S and above, it was all getting a bit crowded. One pleasing little touch: when you touch the volume slider to change it, the speaker buttons at either end light up. (Update: Marc Blank-Settle says this was already in iOS 9, and he’s right, it was. This is what makes software reviewing tough: you notice something for the first time just when it has always been there.)

Press to unlock

The most subtle change is that it’s no longer enough to rest your thumb (or other finger) on the TouchID button to unlock the phone/tablet. It certainly used to be the case that it was, but on the 6S range in particular this could mean that if you picked the device up to see what was on the notification lock screen, and particularly if you used a phone, chances were high you’d unlock the thing and miss what you actually wanted to see.

Now you have to actively press on the button to both identify yourself and to open the lockscreen. This also fits in with the new Taptic buttons on the iPhone 7 range, which don’t actually move, so that you have to tell them you’re there by actively pressing.

This seems like a trivial point, but in the first few weeks you’re going to hear lots of people whose muscle memory is built around resting their fingers on that button who don’t understand why doing that doesn’t unlock it. On such small things are perceptions of ease of use built.

However you can turn this off, at least on TouchID devices. You have to go to Settings ▶️ Accessibility ▶️ Home button, and there you’ll find “Rest Finger to Open” as an option. Lots of things are hidden down there in “Accessibility”.

Home button: accessibility options

You can revert to the old TouchID behaviour via Accessibility.

Deleting apps

Sure, you can delete the stock apps. Don’t bother. You’re not really saving any space. And that app you downloaded to replace it? Takes up more room and doesn’t get system-wide benefits.

Mail, now with filters

Speaking of stock apps, iOS’s Mail is creeping towards a vague parity with what OSX’s Mail could do in about 2000, when the latter was still in beta. Though it is way easier to triage email with swipes on a touchscreen than a keyboard and mouse.

In iOS 10, you can filter email, via a little “filter” icon at the bottom of the screen: tap it to change between filter criteria.

iOS 10's mailbox filter

You can filter mailboxes by Unread, Flagged and a few other criteria: tap the icon

We’re still stuck, though, with a very limited number of ongoing filter systems: you can’t set up a “smart mailbox” based on a phrase, for example, even though OSX has had that forever. Here are the options for filters:


This “what does that do?” thing about the filter icon is something most people will probably come across by accident. It’s helpful, but Mail is still some way from being a powerful app. It’s still only useful.

Maps: you can get there from here

In iOS 9, Maps began getting public transport details, and that has quietly been enhanced over the past year. The key change is that it’s much more sensibly laid out: search is on the bottom, and location plus settings are in the top right.

Even better: search is coordinated among devices, so that if you do a search on your tablet, those searches will also be on your phone. (Finally.)

Ios9 10 maps
The Maps app is improved in iOS 10 (on right) over that on iOS 9 (left): it now puts search in a more accessible location at the bottom, remembers searches from other devices, and can offer ride-sharing app routes.

Notes, collaborate

Apple made something of collaborative editing coming to iWork at the iPhone introduction last week, but it’s offering exactly that in the new Notes: type up a note, and you can choose to share it with someone, who will see the changes that get made, and be able to edit it too.

Obvious use: shopping lists. As long as the person shopping (or suggesting shopping) doesn’t go out of range of data.

Under the hood: Siri and machine learning

The range of things that Siri can do hasn’t changed much in this update – at least, not visibly – but it is improving. And what’s really going to change is that it will be open to some developers, for a limited number of functions. I didn’t see any in the betas (you’ll have to see what developers do with it).

Photos are meant to get a tonne of machine learning. But it’s principally facial recognition, and the “Memories” function is – for me at least, having few photos with location tags – so-so. Yes, it’s nice to have photos collected together from particular days, but this isn’t Google Photos with its ability to find “photos of dogs” from an unlabelled corpus of pictures.

Update: Nick Heer points out that it does show you photos that match a keyword (singular is best). It hasn’t done this on the iPhone SE, but on checking my iPad and doing a search in the photos for “horse” I find that yes, he’s correct. iOS 10 calls them “categories”. You can discover what categories it has available by typing a single letter of the alphabet into the search box, and seeing what unravels. (Perhaps someone will make a list. What am I saying? For sure someone will make a list. And look – here it is.)

Photo search on iOS 10

Type a letter, get a list of categories

[end update]

Then again, the pictures sit on your phone, so possibly over time the capability will be there. (We simply don’t know how much processing power per photo is needed for Google Photos’ identification system, nor how many examples it has to see to hit its training targets.)

Finally: home screen widgets

Apple hasn’t gone as far as Google in Android, and nothing like as far Microsoft in Windows Phone, in terms of what widgets are able to do as a layer over the home screen. They don’t dynamically update while you’re not looking; they hurry to do it when you swipe across. Saves on background processing. But you can edit them, as before.

Home screen widgets on iOS 10

Yeah, that’s all

Sure, there’s a ton of other stuff. There’s:
• the update to Messages (annoy your iOS 10 friends by sending them “Happy Birthday” messages) which now means that it’s becoming something of a platform.
• Apple Pay on the web – possibly that should have been a feature above, but I never tried it out.
• Home. As an app. I couldn’t find any products that actually hooked into this, and I suspect it might be a while before I do. (Ravi Hiranand says Home found his Philips Hue light automatically, and “works better than the original app”.)
• Subtle thickening of fonts, so that text is easier to read. This is system-wide, and very noticeable in the re-thought Apple Music and in Maps.


So – should you upgrade to iOS 10? Don’t you love how this question is asked as if you might not? You’ve read a whole piece about it that you didn’t have to. You probably will. And yes, you should benefit. Some of the touches are clever, and some are overdue, and some are essential. But it’s all about getting the device out of the way.

The thing you’ll notice the most? Pressing the Home button. It’ll bug you gently for a couple of weeks. Then you’ll forget it. And after that, you’ll notice the Maps app’s improvements. And those you’ll probably forget; can you remember what it was like before? Hardly anyone can.

That’s the way with software: you change things wholesale, and within a few months nobody could draw what the old thing looked like. Believe me, though, if you came across a device running iOS 6 or earlier, you’d be amazed at how… primitive it looks. Pundits might have bitched about iOS 7, but it’s been a wholesale improvement in user interface.

One could wish for better, smarter AI, but that might have to wait a few years for more power on the device. Even so, the “Siripods” (aka AirPods) point towards Apple wanting us to have a closer verbal relationship with our devices.

Colours, names and numbers: why is it ‘iPhone 7’ but just ‘MacBook Air’?

Phil Schiller introducing the iPhone 7/Plus in San Francisco. Yes, but why “7”? Photo by tuaulamac on Flickr.

Yes, Apple launched new phones the other day. Yes, there isn’t a 3.5mm headphone jack – which will cause varying amounts of trouble for people and accessory makers. There are “AirPods” to buy ($159, or £159 – even with VAT that doesn’t make sense on the exchange rate front).

We know too that Apple faces challenges: the premium smartphone market is saturated, so that pretty much all sales now will be to people who are replacing an existing smartphone; and Apple only targets the premium end, since even the iPhone SE (which does have a 3.5mm headphone hack, and a 16GB model) starts at $350, which is the lower limit of what analysts call “premium”.

But there’s a subtler question around this iPhone. And it’s this: why does Apple call it “7”? Sure, they’ve moved things around, and redesigned this and tweaked that (oh, yeah, headphone jack has gone). This is known as a “refresh” in the business.

The MacBook. Who can tell which year they’re from? Only the colour indicates that the right-hand one isn’t from 2015. Photos by tuaulamac on Flickr.

So why do we have a number for this product, and yet when the MacBook line had a refresh earlier in 2016, that wasn’t called “MacBook 2” – or even “MacBook 3”, as there was one back in 2010?

Similarly, the expected and long overdue refresh of the Mac Pro isn’t going to be the “Mac Pro 2”; nor is the MacBook Pro line going to get a “MacBook Pro 9”, given that there have been eight iterations already since the line was introduced in February 2011.

On its face, this is puzzling. Apple wants people to upgrade their hardware – that’s where vast amounts of its revenue comes from – and with the iPhone, it signals to people that there are new models available through the numbering system, and also through the colours available. (In 2013, the 5S came in a “gold” colour; in 2015 there was the new “rose gold”.)

By contrast, the PCs come with very little variation in appearance – you have to be quite nerdy to spot the difference between a 2012-vintage model MacBook Pro and the 2015 one, say. Though the MacBook does come in multiple colours, and the “MacBook 2” added rose gold as a colour option. (Hence above we know the right-hand one is a 2016 model.)

Colour as a “novelty” signal has long been a favourite for Apple: remember the original iMac, in “Bondi blue”, where the colour range then expanded – signalling how much newer your model was – until it went mad with the Dalmatian range. Of course if the numbering system were used for the iMac, we’d be up to something like “iMac 20” by now, what with all the variations of the 1998 teardrop, the lampstand, and now the flat-panel version.

Yet the more you look across the product range, the more peculiar this choice of numbering only for phones becomes. It has even been dropped for the iPads; where we used to have iPad 2, 3, 4, and then “iPad mini”, and then Air and Air 2, we now just have iPad Pro and iPad Air. (And mini. Not sure how long that will survive. It’s something of an “iPod touch for the iPad range”.) There’s the Apple TV. Can you tell them apart? Only by size. They don’t get “Apple TV 4”. Airport Express? Airport Extreme?

And how about the iPod? That didn’t get numbers – though it did get descriptors as the range grew (“Classic”, “mini”, “nano”, “shuffle”, “touch”). The changing design was itself sufficient differentiation, perhaps.

Here’s another example from a saturated market: cars don’t get numbers. They get names. There’s the Ford Focus, the Vauxhall Zafira, the Chrysler Plymouth, and so on. Ah – except as John Dodds points out, BMW uses numbers for its ranges (BMW 3 series, 5 series). Hmm.

Note, by the way, that the Watch is getting a “Series 2” moniker, and also new materials and colours. But the descriptors (“Edition”, “Sport”) have been abandoned.

What’s in a name and a number?

What do we conclude? Perhaps those more steeped in marketing will provide a better analysis, but it feels to me as though there’s an urgency in using numbers to name a product: it immediately dates the old one and give the new one a sense of being right here, right now. (Except – odd detail – you won’t find the phone number on the phone itself. Nor is it even displayed in or on the phone. The back of the iPhone used to show the memory; now it doesn’t even do that.)

Why not do the same on the PCs then? It’s not that there’s suddenly a temptation to slide past a lengthening upgrade cycle; Apple used to upgrade them every six months or so like clockwork, but recently has become less interested in doing so. But it didn’t have a number or even descriptor for its PCs back when it was pushing new ones out.

Add in the fact that it’s almost impossible to differentiate between the iPhone 6/6S/7 (and 6+/6S+/7+ – except the latter has dual cameras) except if they’re in unusual colours, and you have a conundrum. Apple attaches numbers to these products in its marketing; yet there’s none on the devices themselves. Perhaps it’s to help people pretend that they have the latest when in fact they don’t; the “unashamedly plastic” 5C sold comparatively poorly (against expectations if Apple had simply continued the iPhone 5, which is what the 5C actually was inside) perhaps because owning one indicated to the world that you didn’t have enough money to buy the top-end new model, the 5S, and that you hadn’t had it the year before either (because you’d have had the 5).

So if the phones don’t have the numbers, why does the marketing? Possibly it’s just to do what marketing should: make you aware there’s a new product. Even if you can’t tell which one other people – or even you – are using. (Do you know which model of phone you have? This might be indicative of how susceptible you are to this marketing method.)

The only other question is: when will Apple stop numbering its iPhones? Will the 2017 version, being the tenth anniversary edition, be the “iPhone Edition” or some such? Once you’re on this thing, it seems unlikely to stop – but I can’t wrap my head around the concept of 2020 rolling around and the “iPhone 9” being unveiled while a few weeks later one of the PCs gets an update that makes it autonomously intelligent with VR, and yet it’s just called “the MacBook”.

(From an idea on Twitter by Joe Asbridge – thanks Joe. Told you I’d write it up eventually.)

Liked this? Each day there’s a Start Up post with links and commentary. You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.