The Real Threat to Apple: The Invisible Device

The Amazon Echo: what happens if this is all there is to a device? CC-licensed photo by Cryptik Merlin on Flickr

For years, Apple has made a name for itself through the design of its products – their combination of appearance, materials, and software functionality (which is part of the “design”, aka “how it works”). It has been able to command premium prices for desktops, laptops, phones, tablets, even routers by making things that not only work well, but look good.

What happens to that advantage and ability to command a premium, though, when there isn’t a product to hold? What happens if you don’t have a phone to pull out, a tablet to press, or a router to put in the corner of your room?

This thought struck me while listening to John Gruber and Ben Thompson discussing Amazon’s Echo, which we could roughly call a home automation device, and considering Google Home, which is going to be approximately the same thing. Both de-emphasise the physical product (there isn’t even a screen) in favour of an unobtrusive always-listening device which doesn’t need to be pressed or waved at; it just responds when spoken to.

It’s not hard to imagine future versions of Google Home or the Amazon Echo would have less and less physical hardware; essentially, they only need to power a microphone, a speaker and an internet connection. In which case, what would Apple’s version look like? It might look – might even be – the Apple TV. It’s nice, but many people would struggle to pick it out of a lineup. And once it’s underneath or behind your TV, you could forget it’s there.

Razing the playing field

But when you reach that point, the ground on which Apple used to fight – appearance, materials, “look and feel” – has suddenly vanished. The shift to systems which don’t need us to look at them directly and which feed information back to us by means other than an integrated item with a screen, doesn’t so much move the goalposts as set fire to them and terraform the field where they were standing.

In the same vein, I was asked a few years ago – when Siri had newly been announced, but Samsung was already making inroads to the premium market with the Galaxy Note – what I thought the phone of the future would look like. I suggested you wouldn’t actually look at it much. It would probably be Galaxy Note-sized but it would sit in your pocket and feed information to your headphones in response to questions you asked into the mic on the headset. Less need for the screen, less need for typing on the physical object.

Our obsession with photographs and cameras has forestalled that shift; Instagram and Snapchat demonstrate that, when it comes to social interaction, we love the visual. That suggests screens and devices – in other words, things we actually hold and carry around – remain important.

Even so, the invisible device does seem to me the biggest risk Apple faces. The advantage it has is Amazon’s Echo and Google Home are devices for, well, the home and, although we might talk a lot about it, the extent of our desire to have computing interaction with our home is surprisingly limited. Jan Dawson made this point well recently. The current “smart home” market is composed of early adopters because your fridge can’t ever be that smart and you’ll still have to load and unload your washing and put coffee into the coffee machine.

What’s more, most of us spend most of our time outside the home. And that’s where we need our devices. So far, we haven’t quite taken up the idea of chatting away to our headsets in the manner of Joaquin Phoenix in Her. But, bear in mind, social norms can shift; our parents’ generation would have been (and still are) appalled by the way teenagers today will ignore each other while across a table, or their elders, in favor of the glowing screen. And 30 years ago, walking along the street talking aloud to nobody was a sign of insanity. Now it just means you’re on a call. (Spot the difference: if you’re wearing earphones, nobody will turn a hair.) If intelligent assistants really take off, devices might shrink away too. Though every time I follow that reasoning, I arrive back at the need to digest visual information. We’ll still need screens, and hence housings for screens, and hence design.

These answers brought to you by…

That this potentially poses a threat to Apple doesn’t mean that everyone’s safe. Amazon’s pretty safe; if people order things via the Echo, it benefits. But Google relies on people looking at ads for 90% of its revenues and rather more of its profits. If we don’t look at a screen, how do we get the ads? Perhaps it will adopt the solution chosen in the UK by the “Speaking Clock”, a phone service you called to get the precise time read out to you. In 1986, the newly privatised British Telecom put it out to sponsorship, which was eagerly snapped up by Accurist – and so for 22 years, you would be told, “the time sponsored by Accurist is…”

Maybe that’s how Google will adapt if voice is the new interface. Equally, maybe that will open the door for companies like Apple to charge extra so we don’t hear the ads. The invisible device might still yield a premium. It’s just a question of what you’re paying for.

How LG’s add-on keyboard makes more profit than its phones

LG’s Rolly keyboard: strong enough to prop up a lossmaking mobile division? CC-licensed photo by Rob Pegoraro on Flickr.

LG released its first-quarter figures and, oh my, the news for its mobile division was bad again: shipments down to 13.5m (a 13% drop year-on-year), revenues of US$2.67bn at the prevailing exchange rate, and an operating loss of US$224m – that’s an average selling price of $197, and per-handset loss of $16.64.

But what caught my eye was the red ink on the mobile division seemed to go back further than before. Whereas I’d previously thought it lost money in Q3 and Q4, the latest results now seemed to suggest it had lost money in Q2 as well.

The reason why quickly became apparent: LG had restated the results for the mobile division. There in the Appendix where the divisional revenues were broken out was the footnote:

LG's footnote: IPD has moved

“1) The IPD (Innovative Personal Device) department within MC [Mobile Communications] division has been restated to ‘Others'”.

But what was the IPD department? What did it make?

A quick bit of searching turned up a few results, such as this Rolly Keyboard, introduced “in an effort to try to capture a larger share of the fast-growing mobile accessories market”. Or this Bluetooth headset (“designed for both professionals and avid music lovers”).

You get the idea: IPD is accessories. Happily, because LG had restated the results, it was quite easy to see how much the IPD department had been contributing to the mobile division’s revenues – and profits.

The result turned out to be surprising.

It figures

Here are the numbers (stated in Korean Wons), first for 2015:

LG results, end of 4Q 2015
In 4Q 15, LG suggested that the mobile division made profit for 6 months

And now the results published covering 1Q 2016 and the four quarters before it:

LG results, restated in 1Q 2016
Three months later, the mobile division apparently made a loss for nine, not six, months in 2015.

I put these into a spreadsheet to see how IPD was doing.

LG's IPD department financials

Pretty well, as it turns out. Although the revenues aren’t enormous, they point to healthy sales. The Rolly keyboard (which now has a v2) costs about $100; the Tone costs around the same. That suggests LG is shifting about a million units all told per quarter from this department.

But look at those operating profit margins: the worst is 16% and it has been nudging 20%. That’s miles better than for the phone division which, even though it turns out more than ten times as many devices that sell at twice the price, hasn’t turned a profit for a year. How come? Everyone’s heard of LG phones. Hardly anyone’s heard of the Rolly keyboard.

Accessory to the truth

But the reality is accessories have often been a far more profitable pursuit than making the original devices and they often produce a positive feedback loop. Part of the success of the iPod was the popularity of accessories. Shops liked selling them because the margins on the accessories was far better than on the iPods themselves. For instance, by the end of 2005, a US analyst company reckoned that, for every $3 spent on an iPod, another $1 was spent on an accessory such as a case, car charger or (briefly popular) in-car tuner – and retailers got about 15% margin on the iPod, against 25% for add-on electronics and 50% for a case. You didn’t even have to sell iPods to sell the add-ons or cases. And for Apple, the accessories served as an advert for the iPod.

You see the same dynamic playing out with iPhone accessories and especially cases. Look around and ask yourself how many cases you see for Motorola, HTC or – indeed – LG phones. Only Samsung is big enough now to merit its own line of accessories.

But it’s clear LG is hoping to help its profitability with IPDs and, of course, its new range of “Friends” for the LG G5. If you haven’t heard about Friends, they’re a set of add-on devices such as a camera and a high-resolution audio element: you slide them on and off the end of the phone to extend its functionality. You also buy them separately.

Part of the aim is to move towards more modular smartphones – as long as those are LG smartphones. Whether the Friends idea will come to other LG phones isn’t yet clear but the reason for doing it is a lot clearer: there’s gold, or at least profit, in accessories. If the idea takes off, it could even prop up the stumbling mobile business. If it doesn’t, well, IPD is still making a profit without phones.

China, Apple’s iOS source code and the pursuit of unfounded speculation

iPhone and code on screen
Source code isn’t trivial – and that for the iPhone is closely guarded. Photo by Jason Swaby on Flickr.

Let’s begin at the end – because news stories generally should start off with the most recent information. Speaking under oath to the US Congress, Apple’s general counsel Bruce Sewell said: “I want to be very clear on this. We have not provided source code to the Chinese government.”

It’s a strange thing to have to say. But the rumours have been swirling for a while. So I tried to track them down to their source. It turns out to be much stranger than you’d think.

Beginning with the briefs

A month ago, when Apple was fighting off the FBI’s demand that it unlock an iPhone 5C used by one of the San Bernadino killers, the US government seemed to suggest that Apple had already helped out the Chinese government.

In a sworn affidavit, Craig Federighi, Apple’s SVP of software engineering, flat-out denied it: “Apple uses the same security protocols everywhere in the world. Apple has never made user data, whether stored on the iPhone or in iCloud, more technologically accessible to any country’s government. We believe any such access is too dangerous to allow. Apple has also not provided any government with its proprietary iOS source code.”

In its “Reply Brief in Support of Apple’s Motion to Vacate” (relating to Apple’s legal tussle with the FBI over the creation of ‘GovtOS’ to hack into the iPhone 5C used by one of the San Bernadino shooters), on pages 20 and 21 Apple says:

“Finally, the government attempts to disclaim the obvious international implications of its demand, asserting that any pressure to hand over the same software to foreign agents ‘flows from [Apple’s] decision to do business in foreign countries…’. Contrary to the government’s misleading statistics, which had to do with lawful process and did not compel the creation of software that undermines the security of its users, Apple has never built a back door of any kind into iOS, or otherwise made data stored on the iPhone or in iCloud more technically accessible to any country’s government.

The government is wrong in asserting that Apple made ‘special accommodations’ for China, as Apple uses the same security protocols everyone in the world and follows the same standards for responding to law enforcement requests.”

At that point Apple’s brief cross-refers to Federighi’s sworn statement – which means if it’s untrue then he’s perjuring himself, and liable to imprisonment.

Why does this matter? Because there are quite a few people who are completely convinced that Apple has given the Chinese government access – at least for code review, perhaps for more – to the iOS source code. And the US government repeated that claim in its docket, apparently leaning on various claims about Apple made in the media, rather than supplying something a bit more, well, robust.

I think I’ve tracked down the basis for this meme. As with a lot of modern journalism, it starts with a pretty wild claim, and then it gets repeated – but nobody bothers to actually check the original.

Cutting into Quartz

As far as I can ascertain, the first time an English-language site (or indeed any site) suggested that Apple had provided source code to China’s government was in a January 23, 2015 article in Quartz, headlined “Apple is reportedly giving the Chinese government access to devices for ‘security checks’“.

That’s a pretty straightforward headline on an article which Quartz’s Asia correspondent Heather Timmons, an experienced journalist, wrote, following up on a tweet the day before from the state-run People’s Daily in China:

OK, but what are “security checks”? Tim Cook was quoted in the story with a preemptive denial:

“There were rumors that Apple built back doors in its devices, and let third parties have data and access those devices, but that was never true and that we would never do that in the future either,” Cook reportedly said.”

Why “reportedly” said? Because the quote is a translation from a story in the Chinese-language Beijing News, which said that Lu Wei, the director of the National Internet Information Office, met in early December (2014) with Cook, who “said China will cooperate with network security assessment of Apple products”.

The article continues (via Google Translate): “China is also willing to open to Apple and other technology giants, but only if the iPhone, iPad and Mac products must ensure information security and privacy of users, while maintaining national security.”

It then has the Cook quote about back doors, to which Lu Wei responded [in translation] “You said not, your new products to make our network security officer for evaluation. We need to draw conclusions, so that consumers must be assured.” Cook said Apple will fully cooperate with the Network Security Assessment China [of] Apple products, to ensure that the user can feel safe and secure during use of the product… which means that Apple is willing to accept China became the first official review of network security company.”

The Bing translation is pretty much identical, except that it uses “reliable” rather than secure. The tone is the same: no mention of source code. (Better translations welcomed.)

Timmons clearly wondered what a security check entails, and noted that Apple didn’t provide any information when contacted, and sought some outside opinions. But then things go a bit askew. Timmons continues:

“But analysts said the most likely interpretation is that the company is giving Beijing access to its operating system source code in return for being able to continue to do business in China”

Now that’s a pretty big claim to anyone who knows what source code is. So, which analysts? The first seems to be ‘Percy Alpha’, “a pseudonymous founder of the anti-censorship group”, which monitors connectivity and censorship topics around China’s “Great Firewall”.

‘Percy Alpha’ is quoted as saying:

“Handing over source code [would] mean that the Chinese government would know exactly how an Apple software works.”

Well, yes, it would. But there’s no evidence provided that “Percy Alpha” knows that is what happened. However, the story mentions “analysts”. Who else says Apple handed over the source code? The next quoted is Ben Cavender or China Market Research Group in Shanghai, saying that

if that is in fact what has been agreed, it’s a landmark deal, and Apple has not generally provided such information to other governments.

Clearly, Cavender doesn’t know if it has or not; my reading of the structure of the story is that Cavender was presented with the ‘Percy Alpha’ quote and asked to comment on it. Which leaves us with “Percy Alpha”. What, exactly, did he know?

Meme, aloft

Bear that question in mind, but put it to one side for a moment while we watch this “source code” story spread. (This isn’t, by the way, a criticism of Timmons; she asked people questions, and reported what they said. It’s notable too that the person who wrote the headline on the piece didn’t go with the “source code” element; instead, they went for the “access to devices” angle.)

On the same day, an article in iDigitalTimes on the same day references the first Quartz article – though it talks to someone different, at Quarks Labs [unrelated to Quartz, the news outlet] who says “Apple giving access to the source code of their product? Never!” (The quotes from the person at Quarks Labs are actually worth reading. One wonders what might have happened if they’d been in Timmons’s contacts book.)

Even so, the idea had begun running.

Next up, in terms of size, was Engadget, also on the same day, with an article saying “Apple lets China examine iOS code to assuage spying fears”. The basis for the claim? “According to the Beijing News, Chinese officials met with the [Apple] CEO in December to reach a deal which will allow the State Council Information Office to check for backdoors.” It’s the same Beijing News link, but there’s absolutely no supporting evidence – either in the story or the translation – to support the “source code” claim. And yet there it was.

The San Bernadino shootings in December, and the subsequent resistance by Apple to the FBI’s demands to hack into one of the shooters’ phones, brought this claim back to life again.

Among those who revived it was one of Quartz’s own writers in a February 17, 2016 piece headlined “Apple is openly defying US security orders, but in China it takes a very different approach”. “If Apple had indeed agreed to a Beijing security audit, it could have shared vital information with the Chinese government, such as its operating system’s source code, that could indirectly help government agents discover vulnerabilities on their own,” the writer observed.

The basis of the “source code” claim? The original Quartz story, the writer told me.

Not all outlets regurgitated the claim. A New York Times story on February 21 this year noted that “Apple sees value in its stand to protect security” and pointed out that

In China, for example, Apple — like any other foreign company selling smartphones — hands over devices for import checks by Chinese regulators. Apple also maintains server computers in China, but Apple has previously said that Beijing cannot view the data and that the keys to the servers are not stored in China. In practice and according to Chinese law, Beijing typically has access to any data stored in China.

No mention of source code. But Lawfare, a legal blog founded by a team including former advisers to Barack Obama, thought there should have been – so they created it. “Some reports have speculated that while Apple defies the US government it has no problem acquiescing to Beijing’s security demands”, including backdoors, said a piece on February 22. Which “reports”? That Quartz piece. In fact Lawfare, and others, have leant on that original article so many times it seemed worth swimming back upriver to the source.

No firewall, no smoke?

So we return to that earlier question: what exactly did ‘Percy Alpha’ know about Apple’s negotiation with China? I got in touch with ‘Percy’ has left, I was told by ‘Charlie Smith’ (another pseudonymous staffer, I’m guessing). Did he know where the “source code” suggestion had come from?

“Are we really the source of this Apple speculation on China?” he answered. “I know that we give Apple a hard time, but are we the only voices who are doing so? I have been looking over the Justice Department filings and have seen Apple’s response saying that there are three sources of information coming from three different newspapers – in your opinion, are we the source for that information in all three publications?”

Yes, I responded, ‘Percy Alpha’ at Greatfire really does seem to be the source.

“I don’t have the smoking gun, and Percy did not either when he gave that quote to Quartz,” said Charlie. The source code quote was “guesswork/speculation”, he said.

OK, but was it informed speculation based on contacts inside the “security review” teams, or people or companies that had gone through it, I asked?

“I would imagine that only a small number of people know exactly what the security review entails – the folks at CAC [Cyberspace Administration of China] and the reps from the companies that have gone through the process,” replied Charlie. “But, yes, I guess I have to agree that this is pure speculation on our part.”

Pure speculation. This didn’t stop Charlie making exactly the same “source code handover” claim in February 2015 to the Washington Post.

That’s because he thinks he’s right. “I believe that they would have had to do that [share the source code] because the authorities had concerns about certain things which would require that they look at the code. If you review the legislation (unofficial translation) it is also clear that an audit would not just be a rudimentary check.”

The translation doesn’t mention an audit, and it’s not obvious where Apple’s phones and the iOS source code would fit into the “Network Information Security” clauses there. More to the point: it’s a big allegation to make based on no direct evidence.

Source code reality

To be clear: source code audits do happen, in China and elsewhere. Notably, in 2012 the Chinese telecoms company Huawei submitted its router source code for review by the UK’s security agencies and the Australian government. (The US’s National Security Agency also targeted that source code for hacking, according to documents released by Edward Snowden.) The topic of Huawei and source code turns out to be a sensitive one, at least for Cisco, which accused Huawei of stealing code back in a lawsuit in 2003-4.

IBM is also reported to have let the Chinese government review “some product source code” in a secure room “without the ability to remove it from a room”, according to the Wall Street Journal in October 2015 – which quoted, at second hand, one of IBM’s own senior vice-presidents speaking in China. The WSJ noted that Microsoft said it had shared some Windows source code with the Chinese government. But Apple’s name is notably absent, either there or in a February 2015 story in the WSJ about US companies’ reluctance to share source code with China.

And of course the Chinese government can easily review the Android source code – famously, it’s open to anyone.

Even the US tries it. In March, Zack Whittaker at ZDNet reported that the US government has attempted to use the Foreign Intelligence Surveillance Act (FISA) to obtain source code from tech companies in numerous lawsuits. (It’s notable that IBM’s statement to Whittaker leaves lots of wiggle room: “the company does not provide ‘software source code or encryption keys to the NSA or any other government agency for the purpose of accessing client data.'” Which leaves other purposes open.) Many of the companies contacted didn’t comment. We have Federighi’s response for Apple from the court docket.

And how do you get at Apple’s source code? It’s very tightly locked away, according to evidence provided at the Apple-Samsung trial in 2012:

Apple source code is provided the highest level of protection and security within Apple. Physical access to the iOS source code is limited to select groups of authorized Apple employees, with access being provided only to portions of cod on a need-to-know basis. Acess is limited to employees directly involved in software development, management, and security. The employees with such access must be approved by management as authorized employees, their accounts must be specifically granted access.


So what access does Apple give in China? There’s a clue in Federighi’s affidavit: right after the part about not allowing access to source code, Federighi states: “While governmental agencies in various countries, including the United States, perform regulatory reviews of new iPhone releases, all that Apple provides in those circumstances is an unmodified iPhone device.” (Emphasis added.)

So in the end, it might come down to simply that: an iPhone for testing.

Critics of Apple – who are plentiful, and include Charlie at Greatfire – say that there’s wiggle room in Federighi’s statement. Regulatory agencies aren’t security agencies, points out Greatfire’s Charlie. To say “Apple has also not provided any government with its proprietary iOS source code” doesn’t rule out showing a government some part of its source code.

In that, he’s right: if you want, you can always find some formula of words that hasn’t been ruled out which might cover some way of interpreting a situation where someone sees some source code. It’s an endless corridor where every door just leads to another. All we have is Apple’s, and Federighi’s, and Sewell’s, denial. Those will satisfy some. They won’t satisfy others.

But it would be good if those claiming Apple has shown – or otherwise shared – its source code with foreign governments could put up some proof of it happening. The reality is that, despite my best efforts, I’ve found nothing solidly founded in anything published; only two denials, by two senior Apple people, on the record, under oath. Ranged against them: one pseudonymous staffer at a Chinese monitoring site.

At a time when the issue of access to Apple’s source code is one of the hottest topics in the technology/legal world, solid proof rather than unfounded speculation seems like the least we should demand from our discourse on the subject.

Is Firefox Search Worth $375M/Year to a Yahoo Buyer?

Marissa Mayer and Firefox: worth it?

Marissa Mayer and Firefox: can the marriage last? Photo of Marissa Mayer by Fortune Global Forum on Flickr.

Who stands to lose if Yahoo is sold — besides of course Marissa Mayer, who will probably lose her job along with a fair number of Yahoo staff? The surprising, and unobvious, answer is Mozilla and the Firefox browser.

That’s because Mozilla is highly dependent on a five-year contract with Yahoo, signed in December 2014, where it receives about $375m per year to make Yahoo the default search provider in the Firefox browser on the desktop. From 2004 to 2014, that contract was exclusively with Google; now it’s Yahoo in the US, Google in Europe, Yandex in Russia and Baidu in China.

How much is $375m per year compared to Mozilla’s spending? Most of it. Mozilla’s audited financials offer some useful details. They’re not as timely as a public company’s numbers; the most recent date to the end of 2014.

Mozilla’s numbers

In 2013, the Mozilla foundation recorded “royalties” (mainly, search income) of $306.1m out of total revenues of $314.1m; in 2014, 323.3m of $329.6m. Search income is about 97% of Mozilla’s total income.

It’s also clear Yahoo is paying Mozilla more than it got from Google. Marissa Mayer was reportedly so keen to secure the business, she made a preemptive bid that turned out to be far too high for the reality of a world where Firefox’s share on the desktop was falling and its position on mobile is minimal.

The question is, with Yahoo on the block, would a buyer of Yahoo want to continue with the Mozilla contract? It is a big drag on Yahoo’s spending. According to Yahoo’s financials, its Traffic Acquisition Costs (TAC) – the money it pays other companies to bring traffic to it – have rocketed.

Yahoo spending on traffic acquisition

Clearly, it’s spending a lot more both for display ads and for search. TAC can be a good thing: you pay a third-party site to bring people to you and then you make a profit by selling those people products or showing them ads.

Yahoo’s search TAC, in particular, has rocketed from a low of 0.7m in the first quarter of 2014 to $141m in the fourth quarter of 2015, just over a year after signing the deal with Mozilla.

That’s not all going to Mozilla. But digging into Yahoo’s financial statements, we can find out precisely how much it is paying.

In its annual report for 2015, Yahoo says: “Of the $350m increase in revenue and $660m increase in TAC for the year ended December 31, 2015, $394m and $375m were attributable to the agreement we entered into in November 2014 to compensate Mozilla for making us the default search provider on certain of Mozilla’s products in the United States (the “Mozilla Agreement”).”

(You might wonder: why is the increase in overall revenue smaller than the increase from Mozilla? It’s because Yahoo’s overall revenues fell.)

So Yahoo is paying $375m annually to Mozilla just to be the default search engine in Firefox on the desktop in the US. And it’s going to keep on paying. In the 3Q 15 report, it said: “The Company is obligated to make payments, which represent TAC, to its Affiliates. As of September 30, 2015, these commitments totaled $1,682 million, of which $100m will be payable in the [fourth quarter] of 2015, $401m will be payable in 2016, $400m will be payable in 2017, $375m will be payable in 2018, $375m will be payable in 2019, and $31 million will be payable thereafter.”

Given that $375m went to Mozilla in 2015, it seems likely the large part of those future sums are also bound for Mozilla.

But a future buyer might not want to stick with Mozilla because Yahoo’s TAC is beginning to get out of whack.

For comparison, Google’s TAC used to be between 23% and 25% of its ad and total revenues; more recently – since the end of the Mozilla contract – that has fallen below 20%.

As a proportion of search revenue, Yahoo’s search TAC has gone from a low of almost 1% of search revenue, to 27% in the fourth quarter of 2015. That’s bigger than Google’s TAC proportion. Yahoo’s problem is it doesn’t have the monopoly Google does and doesn’t monetise its advertising as well as Google. Google’s AdWords are a high-margin ad business. Yahoo offers display ads, which are a commodity.

The end of the search affair

So a Yahoo buyer would be very likely to look for a way to get out of the five-year Mozilla contract. How would that affect Mozilla?

Quite hard.

Mozilla’s expenses in 2013 were, mainly, $197.5m on “software development” (out of total costs of $295.4m); in 2014, that was $212.8m (of a total of $317.8m). “Software development” swallowed up about 65% of the royalty income in 2013; the same in 2014.

As Mozilla acknowledges, those “royalties” are payments from “various search engine and information providers”. What happens if one of those sources dries up?

Mozilla knows it’s at risk here. Under the “Concentrations of Risk” subheading, there’s this:

Mozilla entered into a contract with a search engine provider for royalties which expired in November 2014. In December 2014, Mozilla entered into a contract with another search engine provider for royalties which expires December 2019.

Approximately 90% of Mozilla’s royalty revenues were derived from these contracts for 2014 and 2013 with receivables from these contracts representing approximately 77% and 66% of the December 31, 2014 and 2013 outstanding receivables, respectively.

Yahoo, as we can see, is paying about $400m per year just for US search. How much did Google pay? In 2011, when it re-signed for three years, the estimate was that Google was paying just over $100m per year – for a worldwide deal. It seems likely the real figure was higher. But Mozilla relied on it. And the Yahoo money is even more needed as Mozilla tries to recover from the dead-end of Firefox OS on mobile.

Pulling the plug

Basically, if a new Yahoo owner pulls the plug on the search deal, Mozilla will have to seek a new contract in the US. But who’s going to be willing to step up? Microsoft, probably; but the price that Mozilla will be able to demand will be much lower than it got from Yahoo. Unless, of course, Google decides to step back in and push the bidding up. But its actions around the last auction suggest it wouldn’t be interested; Chrome is too dominant, and Firefox is dwindling.

So the next few weeks aren’t going to be tense just for Yahoo. There’s a whole team of software engineers working on Firefox and other products who will have to wonder about their future if Yahoo has a new owner.

The PC Decline: Four Key Points to Note

The PC market is in decline

The PC industry isn’t getting any healthier. Both IDC’s and Gartner’s figures for PC shipments in the first quarter are out and both show the same direction of travel: down 11.5% to 60.6m units for IDC, down 9.6% to 64.8m units for Gartner.

There are some variations between what the two count as a “PC”: IDC doesn’t include Windows tablets, 2-in-1s such as the Lenovo Yoga, but does include Chromebooks; Gartner doesn’t include Chromebooks, but does include 2-in-1s and Windows tablets. (You might think comparing the two datasets would make it easy to spot who is doing well in Chromebooks, and who isn’t doing well in 2-in-1s or Windows tablets. Sadly, that turns out not to be true; but we can say Chromebooks account for only a few million sales per year and there aren’t any clear signs of that changing.)

At the moment, PC shipments have receded below the point they were at in 3Q 2006 (by IDC’s data); that’s nearly a decade of progress wiped out. It’s a category in retreat. The peak was in 3Q 2011, at 96.1m (IDC) or 95.4m (Gartner).

There are a few points to note in what’s going on.

The Long Slow Goodbye

First, Windows PCs are in serious long-term decline, more so than Apple. Once you subtract Apple’s contribution, you find Windows shipments have been in a year-on-year decline for 15 straight quarters. That’s nearly four years.

Windows PC shipments are falling

Something Happened

Second, the trigger for that decline is entirely unlike the trigger for previous declines in PC shipments. If you look at the long-term view, there are three points since 1999 when Windows PC shipments have shrunk: near the end of 2001 (during a worldwide recession), the end of 2008 (during the global financial crash), and towards the end of 2012.

Three points where Windows shipments fell

What happened in 2012? Nothing external. Instead, the decline seems to have been prompted by the twin thrust of tablets and smartphones. Tablets became broadly available (the iPad 2 dropped in price) and smartphone screens grew larger (the first Samsung Galaxy Note, with a 5.3in screen, had been launched at the end of 2011, and the Note 2 was coming along).

Ben Thompson, writing at Stratechery ($subscription), suggests PCs are being disrupted. What we’re seeing is “less capable” and cheaper devices (tablets and smartphones) taking over the jobs that were being done by much more capable devices. But the users didn’t generally need that capability; browsing, email, writing text, organising and uploading photos and watching video, plus a few games, tended to fill the gamut of what most people need to do on a computer.

Certainly, all the signs are that Thompson is right. That 2012 decline points strongly to a change and, at Gartner, Mikako Kitagawa says, “The ongoing decline in US PC shipments showed that the installed base is still shrinking, a factor that played across developed economies.” A shrinking PC installed base? That must be quite a worry. But it’s what we’re seeing.

Linn Huang of IDC offers something of a hostage to fortune: “The PC market should experience a modest rebound in the coming months.”

The Others

Third, anyone in the “Other” category is probably getting pretty worried now. Even Acer has fallen out of the top five computer makers, displaced by Apple. (I continue to include it, estimating its volumes as slightly below Apple’s.)

"Other" PC suppliers, absolute shipments

"Other" PC players are being squeezed out

If you take “Other” to be companies such as Toshiba, Samsung, Fujitsu, and a myriad of others, then both their share and absolute number of shipments is going in a bad direction:
• Toshiba is struggling with an accounting scandal, and its PC shipments in the US fell by at least 25% in the first quarter, from 0.94m to less than 0.71m. The “Lifestyle” division, which includes the PC business, has seen revenues nearly halved since the calendar first quarter of 2014, and made an operating loss for the past 15 quarters.
• Samsung has withdrawn from Europe’s PC business and is hard to see elsewhere; its PC business is rolled into its IM division, which also includes its mobile division, for accounting. In Q4 2015 (the latest quarter for which there are figures) those revenues were US$780m – which would translate to 1m PCs sold at an ASP of $780, or more possibly 1.5m sold at an ASP of $520, or 2m at $390. (To make the top five for Q4 required shipment volumes of at least 5m.)
• For the fourth quarter, Fujitsu says revenues from PCs fell; it’s been saying that for at least a year.

There’s also a squeeze on Asian sub-scale PC makers, because the dollar’s strength hits them hard when they have to source components from the US (Microsoft Windows, perhaps?). That eats into profitability while the bigger players corner more of the market.

All this is going to lead to further consolidation. There’s already talk that Toshiba or Fujitsu will sell their PC businesses. Smaller companies might just shut up shop or seek a buyer.

Apple: Doing Fine, Thanks

Fourth, Apple is still solid. The USB-C Macbook is a year old and doesn’t show any sign of having set the PC world on fire but since the end of 2004 (a period stretching 46 quarters), Windows PCs have only seen faster than Apple growth in two quarters. Even while Apple’s total shipments (according to IDC and Gartner, ahead of Apple’s formal results later this month) fell in the most recent quarter, it was still nothing like the overall fall for Windows PC makers. Only Dell managed to stay upright better on the slippery slope, falling by 2.0% against Apple’s 2.1% (IDC); Gartner reckons Apple did better, showing 1.0% growth, but was bested by Asus with a 1.5% rise – perhaps through Windows tablet or convertible shipments.

What neither shows is that Apple commands the highest average prices in the industry. Once the figures for this quarter are in I’ll return to the topic but, for now, here are the figures for average selling prices (calculated from company financials and IDC shipment figures) for the big six PC companies:

Average selling prices for top PC makers

As you can see, Apple is miles above the rest there. That also means it can grab a healthy profit, which allows it to stay in business when others struggle.


For the longer term? PCs are contracting towards a core base of users who really want or need them. If people want to be able to plug in USB sticks or SD cards, there’s a PC there for them. But it turns out that lots of people don’t and they’re voting with their wallets. That’s creating a squeeze on the smaller players, but even the big players don’t have it easy – unless, like Apple, they can charge a premium.

Why Google needs an iOS Keyboard: Desktop Search is Dying

An iPhone, a keyboard: can Google replace Apple’s?

Casey Newton’s scoop at The Verge, revealing Google is developing a third-party keyboard for iOS, is one of those revelations which doesn’t surprise, it just slots in to make perfect sense. Of course Google would want a keyboard that would feed back the data about what users are searching for in return for fast searching. Especially on mobile and especially on iOS.

That’s because, as I pointed out via some straightforward calculations based on data Google gave out last year, on any given day more than 50% of people don’t do a single Google search on their mobile phone. By contrast, on any given day, more than 50% of people do at least one search on the desktop. (I did show my work, but in brief: if mobile search is bigger than desktop search, and there are 100bn searches per month, but more Google-capable smartphones than PCs, there are fewer searches per day on mobile than on desktop.)

We know, though, that mobile browser search ads don’t monetise as well for Google as desktop ones do. But there’s another problem for Google, at least in the US: people are doing fewer desktop searches.

That is the picture that emerges from studying data from ComScore, which provides monthly updates on browser share and number of searches carried out in the US. There’s an obvious, immediate criticism: this data doesn’t come from Google but instead from a third-party company which relies on people allowing it to monitor their web browsing habits. The key point though is even if the ComScore data is wrong (and it’s sure to be wrong in some way), it’s going to be consistently wrong, in a consistent way. It should be correct in the general trends it shows us.

With that in mind, take a look at the pattern of Google searches revealed by ComScore’s latest search engine rankings.

Search volume for Google on desktop Source: ComScore. Later data for YouTube and for ‘other’ sites (eg Blogger, shown in green) not available.

Explicit problems

The labeling does need some explanation. ComScore says it measures “explicit core search” which, I was told by Andrew Lipsman at ComScore, is “just searches on the five core engines – Google, Yahoo, Bing, AOL and Ask.” (ComScore also has an “expanded view” which looks at searches on sites such as eBay, Amazon and Craiglist, but that’s for subscribers only)

The reason for shifting to “explicit” search was, in 2010 or so, he says, “Yahoo and Microsoft both began auto-generating searches as people navigated throughout their portals. For example, if somebody clicked on a slideshow about zebras, it might auto-generate searches below about zebras. Since there wasn’t really any intent with these searches, they generated almost no clicks but they were having a significant impact on overall search query volume. So we created a definition that excluded these searches as a better representation of the monetizable search market.” That means it includes searches via “browser bar” searches on Safari and Chrome, of course.

There are also “other” searches within Google’s network of sites – YouTube being by far the biggest, followed by Blogger, Picasa, Photos, Google+ and so on. But Lipsman suggests all but YouTube are insignificant compared to the activity on the main search engine (and YouTube).

Growth: gone

Growth on the desktop isn’t promising. According to the ComScore data, core search alone (not including YouTube or others), peaked in February 2013, and has been shrinking since January 2014. Desktop search on Google is now at the same level, by ComScore’s measure, as at the end of 2010.

Google desktop search: in decline ComScore data says Google desktop search has been declining since January 2014

Two obvious questions: why is that? And what will Google do about it, given the pressure as a public company to keep delivering rising financial returns from its business when 90% of revenues (and perhaps more of profits) come from advertising around search?

The first – why – is pretty easy: mobile. ComScore’s other major public-facing dataset, on the installed base of smartphones in the US, shows that, since December 2010, ownership has risen from 63.2m, or 28% of adults, to 198.5m, or 79%. People who have a mobile phone in their hand are less likely to turn to a desktop to type a query. In fact, they’re not that likely to type a search query at all. Despite the presence of Google search in the Safari toolbar on iOS, and as a front page marquee feature on Android, most people go directly to apps such as Facebook, Twitter, Instagram or Amazon.

ComScore’s dataset tracking the use of the “Google search” app on smartphones in the US shows monthly use bumping along at between 53.5% and 47% of all iOS and Android users, respectively, with the average figure for January 2016 being 50.2%. On the desktop, it has around 64% search share in the US. (It’s much higher outside the US in countries where Google is the top-ranking search engine – up to 95% in some European countries.)

The second question – what will Google do about it, given the pressure to keep revenues rising – is pretty easy as well: pursue all possible avenues of extra monetization.

That’s why when you type pretty much any monetizable query into Google on desktop now you’ll be confronted with more ads directly above the search results than ever before: up to four, as well as the “OneBox” of Google-related subjects such as maps or YouTube videos. SEO analysis says that as a result, “click through rates for these ads should significantly rise” and “cost per click will rise”. Previously, when there were ads on the side of results, small advertisers could catch the searcher’s eye and get 5% of clicks despite having paid the lowest amount for their position.

On YouTube, Google has also become creative. After you watch one video, a “next” will play automatically and show an ad too. Google said YouTube revenue “continues to grow at a very significant rate” in its fourth-quarter results for 2015.

New avenues

What else can Google do? Given that desktop search is unlikely to come back, it has to explore all avenues on mobile and other formats. Which is why producing a keyboard for iOS makes perfect sense. It will be personalized, know what you type, and offer search results whenever you want – but the benefit to Google is any searches through the third-party keyboard won’t be part of its search deal with Apple. They’ll be pure profit.

Whether it will be popular is a different question. Even Google Maps isn’t that big a hit on iOS. If Apple’s report that only one-quarter of iPhone users use it is correct, that puts it at 13.4 million users on the iPhone in the US (because ComScore says that 53.8 million, or 28.1% of iOS users, used Apple Maps in January, consistent with a longer-term trend), against nearly 80 million on Android. We don’t have any data for takeup of third-party keyboards on iOS, but it’s doubtful they come close to that 7% Google Maps has on iOS.

In sum: it’s a nice idea – but may struggle to justify its development time. As Michael Love observed on Twitter after Newton’s scoop, “The people who want a keyboard that uses [Android font] Roboto and their favorite Material Design colors are already on Android.”

Apple v Samsung v Huawei and others: measuring processor efficiency

Whenever consumers are polled about what they want from their smartphone, the top answer is regularly “longer battery life”. That doesn’t necessarily translate into actual purchasing decisions; the number of plug-in power packs for phones is a testament to millions of people who didn’t, or couldn’t, evaluate battery life ahead of purchase or ended up pushing it down the stack of priorities when it came to buying.

A smartphone can, in essence, be boiled down to three elements: a battery, a screen and a processor. Yes, you need lots of other things too, but those are your essential building blocks.

With the release of the Samsung Galaxy S7 with Qualcomm’s new Snapdragon 820 processor, it seems like a good time to examine how the interplay of those elements shapes up. Is Apple ahead in processor design for battery life? Has the Snapdragon 820 created a new standard for the industry?

Benchmarks, benchmarks everywhere

I chose to use the benchmarks on battery life from Anandtech because they cover a wide range of handsets and they perform their own standardised tests. There is a wrinkle (there always is with benchmarks): they recently changed their methodology, which has substantially reduced the apparent battery life of devices being tested. For instance, the iPhone 6S Plus life was 13.1 hours on the old benchmark, but 9.05 on the new one. The Samsung Galaxy S6 went from 10.44 hours to 7.07.

The old benchmark dated from 2013, explains Joshua Ho at Anandtech: “[it] was relatively simple in the sense that it simply loaded a web page, then started a timer to wait a certain period of time before loading the next page. And after nearly 3 years it was time for it to evolve.” (Ho also confirmed to me “13.1 hours” means 13 hours and six minutes, rather than 13 hours and 10 minutes.)

You can find a discussion of why and what they changed on Anandtech’s Samsung Galaxy 7 review section on battery life.

In the new benchmark, Ho says, “we’ve added a major scrolling component to this battery life test. The use of scrolling should add an extra element of GPU compositing, in addition to providing a CPU workload that is not purely race to sleep from a UI perspective. Unfortunately, while it would be quite interesting to also test the power impact of touch-based CPU boost, due to issues with reproducing such a test reliably we’ve elected to avoid doing such things.” He cautions, “It’s important to emphasize that these results could change in the future as much of this data is preliminary”.

Noting that, let’s go to work. The three main elements of the phone – battery, screen, processor – all affect battery life. In theory, a bigger battery, smaller chip die, and fewer pixels on the screen will all lead to longer battery life.

I collected the recorded battery capacity and screen resolution from Phonearena for a range of phones benchmarked by Anandtech and put them into a spreadsheet.

The batteries

On the face of it, Apple’s battery capacities lag behind those of Android OEMs. The figure is boosted by Huawei in particular, so the Android average here is 3,131mAh against 2,297mAh for Apple.

Battery capacity for various phones

I’ve highlighted Apple, Samsung and Huawei because they’re the biggest players in the smartphone game. Also, those are the makers for which Anandtech has the most tests.

The first calculation: simple efficiency

The first obvious calculation to do is: divide the battery life (minutes) by the battery capacity (milliAmphours) to get “minutes per milliamp-hour”. Doing that gives the following graphs for the old and new benchmarks:

Apple, Samsung, Huawei: battery life divided by capacity

And for the new benchmark:

Longer is better: battery life divided by battery capacity

In both graphs, longer is better for this particular metric.

There are a couple of obvious points here. Apple seems to do very well on “bang for buck” on the old benchmark, well ahead of any Android OEM; only Samsung did well there, on last year’s phones.

On the new benchmark, Apple still comes out some distance ahead, with Samsung a lot closer with the Galaxy S7 using the Snapdragon 820. (The S7 wasn’t tested on the old benchmark; not all of the iPhones have yet been tested against the new benchmark.) Huawei, which uses its own Kirin processor for most of its phones, also shows respectable performance – above the average for non-Samsung Androids – except, strangely, in the Nexus 6P.

The second calculation: processor efficiency at pushing pixels

There’s only one problem with the calculations above: it doesn’t take into account how many pixels the battery has to light. The phones we’re looking at there have different numbers of pixels, so lighting each one must take battery power. So now we have to do a new calculation: manipulate the above calculation to account for the number of screen pixels. In other words, if one processor gets (say) 2 minutes per mAh to light 100 pixels, but another processor gets 2 minutes per mAh to light 1000 pixels, the second processor is clearly more efficient, by a factor of 10.

To get a view of processor efficiency, we multiply the above calculation by the number of pixels on the screens.

Let’s look at screen pixels:

Display pixels: big variation

As you can see, Apple is a long way behind most Android OEMs on this (except, notably, Huawei, and even then only for the phablet-sized Plus phones). Certainly you can argue the difference in pixel count actually makes no difference in the real world because, when held at a normal distance, the individual pixels can’t be discerned on an Apple device thus adding more to the screen makes no difference at all, except to put more load on the processor and the battery.

Now we’re ready to see how the processors perform when we break out battery life per pixel. This gives us some insight into processor efficiency.

Here are the results on the old Anandtech benchmark:

'Processor efficiency': Samsung leads

And the new:

'Processor efficiency': Samsung eroding its lead?

As above, longer is better on this metric.

(As a reminder, Anandtech hasn’t tested every device on the new benchmark that it did for the old.)

There are quite a few points to note here. The ones that stick out to me are:
• Samsung’s Exynos processor/display efficiency leads the pack
• The Snapdragon 820 is actually a slight regression from the S6’s Exynos XXX, though apparently better than the Snapdragon 810 (which powers the Xiaomi Mi Note Pro)
• Apple’s A9 does well under the new benchmark, though it’s behind Samsung’s implementation; on the old benchmark, it’s all over the place
• Huawei’s Kirin processor lags the rest of the pack. The Nexus 6P uses the Snapdragon 810; the Kirin seems to perform about as well as a second-tier Snapdragon processor
• there are variations between manufacturers, probably down to power management and other elements. For example, the Mi Note Pro and the Nexus 6P use the same processor and have the same number of pixels (though the Nexus has a larger battery) but the Xiaomi product comes out ahead on both versions of the “efficiency” benchmark
• Samsung’s clear advantage could be due to it making the screens as well as the processors, and so having much more control of manufacturing integration.

I’m sure there are plenty more observations to be made; they’re welcomed in the comments. One thing that’s definitely worth noting is these calculations don’t take into account any usability or user experience measurements. They don’t tell you about frame rates or what the phones or screens or user interface is like. That’s a far more complex question which likely remains beyond the province of benchmarks.

How AlphaGo Illustrates the “Warm Bath And Ice Bucket” View of Technology Progress

Positions from AlphaGo's first win against Fan Hui

Remember the last time you took a bath or shower and it started lukewarm but you gradually warmed it by adding more hot water, until it reached a temperature so hot you could never have got into it at the start? Isn’t it strange how we can be immune to subtle, slow changes all around us?

Then there’s the other extreme – the ice bucket experience, where you’re abruptly plunged into something so dramatically different you can’t think of anything else.

The warm bath and the ice bucket: that’s how technology progresses, too.

As an example of the warm bath, you could point to the improvements in computing power in PCs and smartphones. Every year, they’re faster. You don’t notice how much better until you have to use an old device. (Or, of course, upgrade from a years-old device to a brand new one. These days, the effect is less visible on PCs than smartphones.)

Warm enough yet?

Another, less familiar, example is the burgeoning field of artificial intelligence (also known as machine learning, deep learning, neural networks, expert systems and so on). AI has been the “next big thing” for decades; the burden of expectation was so great it couldn’t meet them. Where, in the year 2001, was HAL, the talking sentient computer from 1968’s film 2001?

And yet, bit by bit, AI has been improving. I realised something was going on two years ago when I wrote a story about an app called Jetpac which could examine Instagram photos and determine whether the people in it were happy, grumpy, and so on.

To do that, Jetpac analysed 100 million photos and was able to determine whether those in it were wearing lipstick (so must be “glammed up”), had moustaches, etc.

A fun story, but it was the underlying technology, which Pete Warden, then CTO at Jetpac, explained to me that made me realise AI was back on the agenda again. He had used a neural network (which mimics, in machine form, the way neurons in the brain work: certain stimuli are reinforced, others are de-emphasised, in a learning process) to do the analysis.

It wasn’t surprising when eight months later Google bought Jetpac. The fit with Google’s broader AI drive was so obvious.

Where’s that tech now? Almost certainly powering the recognition system in Google Photos. Isn’t the Photos recognition system clever? But equally, isn’t it so obviously a progression from the face recognition we’ve seen in apps for years? The temperature is rising.

In fact, the AI temperature is now so high that this week we may witness a key event: a machine winning against a human at one of the subtlest board games ever. AlphaGo, an AI program developed at Google’s Deepmind in London, learned how to play the Chinese game Go at a professional level – and then beat Europe’s best player 5-0. On Wednesday – Tuesday night in the US – AlphaGo takes on Lee Sedol, the game’s top player. (If you haven’t played Go (most people in the West haven’t) let me put it like this: it makes chess look crowded and trivial: the board has four times more points than a chessboard, and the number of possible moves is far, far larger.)

AlphaGo isn’t like Deep Blue; it isn’t programmed just to play Go. Instead, it has a “learning” system which was tuned to play Go by working through millions of games and learning what outcomes were best. It could probably learn to win at chess. The core program learned to play video games.

This is the warming bath: how did we get to the point where computers could learn to beat the best player in the world at a game where intuition and “feel” are seen as essential?

The ice bucket

By contrast, some technologies are ice baths – so dramatically different from what has gone before they upend our expectations. Virtual reality (VR) fits this well. Immersive VR is an utterly different experience from what has gone before and the potential for creating new ways of interacting are what have so many people excited about it.

To people who haven’t tried it, VR tends to be “that thing where you wear the stupid helmet”. But that’s because they haven’t experienced the ice bucket. In the past, trains were a similarly disjunctive experience, able to travel at absurd speeds. There were even fears that the velocity would make passengers’ bodies fly apart.

Are there other “ice bucket” technology examples? The original iPhone was a shock to pretty much everyone, even though the technologies it contained (notably the multi-touch screen) were already known. From January 2007, Google’s Android team sidelined their work on a BlackBerry-like device and focussed instead on a multi-touch product.

Your preference doesn’t matter

Ice buckets change the game abruptly; warm baths surround us and raise the temperature so we can’t imagine life before them. There’s no way to pick which is “better” – and we don’t get to pick anyway, because they happen quite independently of our wishes or expectations. But in truth, there are more warm baths than ice buckets. The gradual improvement of smartphone screens, battery life, chip speeds, mobile reception, mobile speeds, design improvements – they’re all slow improvements which you don’t notice until you don’t have them. For dramatic change, though, the ice bucket beats the lot.

Moment of truth

There’s an instant as you first experience a splash of water when you don’t know whether it’s hot or cold. The match between Lee Sedol and AlphaGo could be like that: an odd mixture of hot and cold, a “where were you when…?” moment. Garry Kasparov’s loss to Deep Blue in 1997 was an iconic moment, remembered by many. It has taken nearly 20 years for a computer program to be able to challenge the top human in Go, which tells you about the gap in complexity between the two games. Fewer people understand Go than chess; but everyone understands winning and losing. Computing’s advance is bringing us a moment when the ice bucket comes from a warm bath.

The first match between AlphaGo and Lee Sedol starts at 1pm Seoul time on Wednesday (4am GMT Wednesday, 11pm EST Tuesday, 8pm PST Tuesday). The match will last up to four hours. It can be viewed on Youtube; there will be commentary (which might not mean much to non-Go players) at Gogame.