Why Google’s struggles with the EC – and FTC – matter

Margrethe Vestager, the Danish-born EC competition commissioner. Photo by Radikal Venstre on Flickr.

“Google doesn’t have any friends,” I was told by someone who has watched the search engine’s tussle with the US Federal Trade Commission and latterly with the European Commission. “It makes enemies all over the place. Look how nobody is standing up for it in this fight. It’s on its own.”

The release, apparently accidentally, of the FTC staff’s report on whether to sue Google over antitrust in 2012 to the Wall Street Journal has highlighted just how true that is. We only got every other page of one of two reports. But that gives us a lot to chew on as the EC prepares a Statement of Objections against Google that will force some sort of settlement. (It’s obvious that the EC is going for an SOO: three previous attempts to settle without one foundered, and the new competition commissioner Margrethe Vestager clearly isn’t going to go down the same road into the teeth of political disapproval.)

The Wall Street Journal has published the FTC staffers’ internal report to the commissioners. And guess what? It shows them outlining many ways in which Google was behaving anticompetitively.

The FTC report says Google
• demoted rivals for vertical business (such as Shopping) in its search engine results pages (SERPS), and promoted its own businesses above those rivals, even when its own offered worse options
• scraped content such as Amazon rankings in order to populate its own rankings for competing services
• scraped content from sites such as Yelp, and when they complained, threatened to remove them from search listings
• crucially, acted in a way that (the report says) resulted “in real harm to consumers and to innovation in the online search and advertising markets. Google has strengthened its monopolies over search and search advertising through anticompetitive means, and has forestalled competitors and would-be competitors’ ability to challenge those monopolies, and this will have lasting negative effects on consumer welfare.”
• among the companies that complained to the FTC, confidentially, were Amazon, eBay, Yelp, Shopzilla and more. Amazon and eBay stand out, because they’re two of Google’s biggest advertisers – yet there they are, saying they don’t like its tactics.

Now the WSJ has published what it got from the FTC: every other page of the report prepared by the staff looking at what happened, with some amazing stories. It’s worth a read. Particularly worth looking at is “footnote 154”, which is on p132 of the physical report, p71 of the electronic one on the WSJ. This is where it shows how Google put its thumb on the scale when it came to competing with rival vertical sites.

What does Google want, though?

Before you do that, though, bear in mind the prism through which you have to understand Google’s actions.

Google’s key business model is to offer search across the internet, and sell ads against peoples’ searches for information (AdWords) or reading on sites where it controls the ads (AdSense).

For that business model to work at maximum efficiency, Google needs
• to be able to offer the “best” search results, as perceived by users (though it’s willing to sacrifice this – see later – and you could ask whether the majority of users will notice)
• to have the maximum possible access to information across the internet to populate search results. Note that this is why it’s in Google’s interests to make cost barriers to information to be pushed to zero, even if that isn’t in the interests of the people or organisations that initially gather and actually own and collate the information; it’s also in Google’s interests to ignore copyright for as long and as far as possible until forced to comply, because that means it can use datasets of dubious legality to improve search results
• to capture as much search advertising as it can
• to capture as much online display advertising as it can

None of those is “evil” in itself. But equally, none is fairies and kittens. It’s rapacious; the image in Dave Eggers’s The Circle (a parable about Google), of a transparent shark that swallows everything it can and turns it into silt, is apt.

YouTube (which it has owned since 2005) is an interesting supporting example here. It’s in Google’s interests for there to be as much material as possible on it, regardless of copyright, so that it can show display adverts (those irritating pre-rolls). It’s in its interests for videos to follow endlessly unless you stop them (an “innovation” it has recently introduced, and from which you have to opt out).

It’s also in its interests for YouTube to rank as highly as possible in search results even if it isn’t the optimum, original or most-linked source of a video, because that way Google captures the advertising around that content, rather than any content owner capturing value (from rental or sale or associated advertising).

It’s also in its interests to do only as much as it absolutely has to in order to remove copyrighted content – and even then, it will often suggest to the copyright owner instead that they just overlook the copyright infringement, and monetise it instead in an ad revenue split. Where of course Google gets to decide the split. (Example: film studios, and all the pirated content from their productions; record labels, and all the uploaded content there, which is monetised through ContentID. Pause for a moment and think about this: you and I wouldn’t have a hope of making money from content other people had uploaded without permission to our website. And particularly not to be able to decide the revenue split from any such monetisation. That Google can and does with YouTube shows its market power – and also the weakness of the law in this space. The record labels couldn’t get a preemptive injunction; so they were left with a fait accompli.)

Think vertical

In building associated businesses (aka “vertical search” – so-called because they’re specific to a field) – such as Google Shopping (where listing was at first free, but then became paid-for just like AdWords), or Google Flight Search (where Google could benefit from being top), or Google Product Search, the FTC report confirmed what everyone had said repeatedly: Google pushed its own product above rivals, even when its own were worse, and even at its own expense.

The FTC report is instructive here. It cites a number of examples where Google either forced other sites to give it content, or took that content (even when the other sites didn’t want it to), or sacrificed search quality in order to push its own vertical products.

Forcing sites to give it content? In building Google Local, Google copied content from Yelp and many other local websites. When they protested – Yelp cut off its data feed to Google – Google tried for a bit, and then came up with a masterplan: it set up Google Places and told local websites that they had to allow it to scrape their content and allow it there, or it would exclude them altogether from web search. Ta-da! There were all the reviews that Google needed to populate Google Local, provided by its putative rivals for free, despite all the effort and cost it had taken them to gather them.

Classic Google: access other peoples’ content for free; ignore the consequential benefits. For Google, it isn’t important whether those local websites survive or not, because it has their data. For a company like Yelp, which relies on people coming to its site and using it, and inputting data, and makes its money from local ads and brand ads, any move by Google to annex its content is a serious threat.

This also points to Google’s dominance. Sites like Shopzilla, the FTC noted, were scared to deny Google the free rein to its data because they worried that people wouldn’t find them.

Shopzilla worried about exclusion from Google's listings

Google offers you a ‘standard licence’, and you’d better accept it.

That’s arm-twisting of the first order.

Google was definitely worried about verticals taking away from its core business: in 2005 Bill Brougher, a Google product manager, said in an internal email that “the real threat” of Google not “executing on verticals” (ie having its own offerings) was

“(a) loss of traffic from Google.com because folks search elsewhere for some queries (b) related revenue loss for high spend verticals like travel (c) missing opportunity if someone else creates the platform to build verticals (d) if one of our big competitors builds a constellation of high quality verticals, we are hurt badly”.

You’ve got questions

Obviously, you’ll be going “but..”:
1) But aren’t “verticals” just another form of search? No – though they need search to be visible. A retailer of any sort is a “vertical”: a shop needs to know what it has to sell in order to offer it for sale. But populating the shop, tying up deals with wholesalers, figuring out pricing – those aren’t “search”. Amazon is a “vertical”; Moneysupermarket is a “vertical” (where it sells various deals, and wraps it with information in its forums). Hotel booking sites, shopping sites, they’re all “verticals”.

Their problem is that they need what they’re offering (“hotel tonight in Wolverhampton”) to be visible via general search, but they don’t want that to be something that can be scraped easily.

Amazon, for example, gave Google limited access to its raw feed; but Google wanted more, including star ratings and sales rankings. Amazon didn’t want to give that up for bulk use (though it was happy for it to be visible individually, when users called a page up). Google simply scraped the Amazon data, page by page – and used the rankings to populate its own shopping services. It did the same with Yelp – which eventually complained and sent a formal cease-and-desist notice.

In passing, this is a classic example of Google having it both ways: if your dataset is big enough, as with Amazon’s, then Google – and its supporters – can claim that scooping up of extra data such as shopping rankings and star ratings is “fair use”; if your dataset is small, then you’re probably small too, and will be threatened by the possibility of exclusion if you refuse to yield it up – witness Shopzilla, above.

(Side note: Microsoft wasn’t above doing something similar when it was dominant. Just read about the Stac compression case: Microsoft got a deep look at a third-party technology that effectively doubled your storage space in the bad old days of MS-DOS; then it took the idea and used rolled it into MS-DOS for free, rather than licensing it. Monopolists act in very similar ways.)

2) But rival search sites are “just a click away”. You don’t have to use Google. The FTC acknowledges this point, which is one that Eric Schmidt and Google have made often. There’s a true/not true element to this. The search engine business effectively collapsed after the dot-com boom in 2000: Alta Vista, which was then the biggest (in revenue and staffing terms) lost all its display ads. And Google did the job better. That’s undeniable. But for at least five crucial years, it had pretty much zero competition. Microsoft was in disarray, and Google was able to attract both search data and advertisers to corner the market.

What’s more, it was the default for search on Firefox and Safari, which helped propel its use. The combination of “better, unrivalled and default” made it a monopoly. Most people don’t even know there’s an alternative, and couldn’t find one if asked. Just listen how many times in everyday conversation – on the radio, in the street, in newspapers – you hear “google” used as a verb.

One thought on that “just a click away” – Google has poured huge amounts of money into making sure that people aren’t presented with any other search engine to begin with. The Mozilla organisation’s biggest source of funds for years has been Google, paying to be its default search (until last autumn, when Yahoo paid for the US default and Google, I understand, didn’t enter a bid – because Google Chrome is now bigger than Firefox). Google pays Apple billions every year to be the default search on Safari on the Mac, iPhone and iPad.

Clearly, Google doesn’t want to be in the position where it’s the one that’s a click away. That’s because it knows that the vast majority of people – usually 95% or so, for any setting – use the defaults.

The reality is that we are where we are: Google is the most-used search engine, it has the largest number and value of search advertisers, and crucially it is annexing other markets in verticals. This dominance/annexation nexus is exactly the point that Microsoft was at with Windows and Internet Explorer.

The difference, the FTC acknowledged, is in the “harm to consumers”. Antitrust, under the US Sherman act, rests on three legs: monopoly of a market; using that monopoly to annexe other markets; harm to consumers. In US v Microsoft, the “harm to consumers” was that by forcing inclusion of Internet Explorer,

“Microsoft foreclosed an opportunity for OEMs to make Windows PC systems less confusing and more user-friendly, as consumers desired” and “by pressuring Intel to drop the development of platform-level NSP software, and otherwise to cut back on its software development efforts, Microsoft deprived consumers of software innovation that they very well may have found valuable, had the innovation been allowed to reach the marketplace. None of these actions had pro-competitive justifications”; furthermore, in the final line of the judgement, Thomas Penfield Jackson says “The ultimate result is that some innovations that would truly benefit consumers never occur for the sole reason that they do not coincide with Microsoft’s self-interest.”

In the case of the FTC and Google, the harm to consumers is less clear-cut; in fact, that’s part of why the FTC held off. Yet it’s hard to look at the tactics that Google used – grabbing other companies’ content, demoting vertical rivals in search, promoting its own verticals even though they’re worse – and not see the same restriction of innovation going on. Might Shopzilla have turned into a rival to Amazon? Could Yelp have built its own map service? Or become something else? History is full of companies which have sort-of-accidentally “pivoted” into something remarkable: Microsoft with MS-DOS for IBM (a contract it got because the company IBM first contacted didn’t respond); Instagram into photos (it was going to be a rival to Foursquare).

What’s most remarkable about the demotion of rivals is that users actually preferred the rivals to be ranked higher according to Google’s own tests.

Footnote 154: the smoking gun

In footnote 154 (on page 132 of the report, but referring to page 29 of the body – which is sadly missing), the FTC describes what happened in 2006-7, when Google was essentially trying to push “vertical search” sites off the front page of results. Google would test big changes to its algorithms on “raters” – ordinary people who were asked to judge how much better a set of SERPs were, according to criteria given them by Google. I’m quoting at length from the footnote:

Initially, Google compiled a list of target comparison shopping sites and demoted them from the top 10 web results, but users preferred comparison shopping sites to the merchant sites that were often boosted by the demotion. (Internal email quote: “We had moderate losses [in raters’ rating – CA] when we promoted an etailer page which listed a single product because the raters thought this was worse than a bizrate or nextag page which listed several similar products. Etailer pagers which listed multiple products fared better but were still not considered better than the meta-shopping pages like bizrate or nextag”).

Google then tried an algorithm that would demote the CSEs [comparison shopping etailer], but not below sites of a certain relevance. Again, the experiment failed, because users liked the quality of the CSE sites. (Internal email quote: “The bizrate/nextag/epinions pages are decently good results, They are usually formatted, rarely broken, load quickly and usually on-topic. Raters tend to like them. I make this point because the replacement pages that we promoted are occasionally off-topic or dead links. Another positive aspect of the meta-shopping pages is that they usually give a variety of choices… The single retailer pagers tend to be single product pages, For a more general query, raters like the variety of choices the meta-shopping site seems to give.”)

Google tried another experiment which kept a CSE within the top five results if it was already there, but demoted others “aggressively”. This too resulted in slightly negative results.

Unable to get positive reviews from raters when Google demoted comparison shopping sites, Google changed the raters’ criteria [my emphasis – CA] to try to get positive results.

Previously, raters judged new algorithms by looking at search results before and after he change “side by side” (SxS), and rated which search results was more relevant in each position. After the first set of results, Google asked the users to instead focus on the diversity and utility of the whole set of results, rather than result by result, telling users explicitly that “if two results on the same side have very similar content then having those two results may not be more valuable than just having one,” When Google tried the new rating criteria with an algorithm which demoted CSEs such that sometimes no CSEs remained in the top 10, the test again came back “solidly negative”.

Google again changed changed its algorithm to demote CSEs only if more than two appeared in the top 10 results, and then, only demoting those beyond the top two. With this change, Google finally got a slightly positive rating it its “diversity test” from its raters. Google finally launched this algorithm change in June 2007.

Here’s the point to hold on to: users preferred having the comparison sites on the first page. But Google was trying to push them off because, as page 28 of the report explains,

“While Google embarked on a multi-year strategy of developing and showcasing its own vertical properties, Google simultaneously adopted a strategy of demoting, or refusing to display, links to certain vertical websites in highly commercial categories. According to Google, the company has targeted for demotion vertical websites that have ‘little or no original content’ or that contains ‘duplicative’ content.”

On that basis, wouldn’t Google have to demote its own verticals? There’s nothing original there. But Google also decided that comparison sites were “undesirable to users” – despite all the evidence that it kept getting from its raters – while at the same time deciding that its own verticals, which sometimes held worse results, were desirable to users.

Clearly, Google doesn’t necessarily pursue what users perceive to be the best results. It’s quite happy to abandon that in the pursuit of what’s perceived as best for Google.

Fair fight?

Now, that’s fair enough – up to a point. Google can mess around with its SERPs but only until it uses its search monopoly to annex other markets to the disbenefit of consumers. It’s easy to argue that in preventing rival verticals getting visibility, it reduced the options open to consumers. What’s much harder is proving harm. That’s where the FTC stalled.

But in Europe, that last part isn’t a block. Monopoly power together with annexation is enough to get you hauled before the European Commission’s DGCOMP (directorate-general of competition). The FTC and EC coordinated closely on their investigations, to the extent of swapping papers and evidence. So the EC DGCOMP has full copies of both the FTC reports. (If only they would leak..)

There’s been plenty of complaining that the EC’s pursuit of Google is just petty nationalism. People – well, Americans – point to the experiment where papers prevented Google News linking to them. Their traffic collapsed. They came back to Google News. Traffic recovered. Sure, this shows that Google is essential; cue Americans crowing about how stupid the newspapers were.

However, if you stop to think about the meaning of the word “monopoly”, that’s not necessarily a good thing for Google to have demonstrated – even unwittingly – in Europe. Now the publishers, who have what could generously be called a love-hate relationship with Google, can show yet another piece of evidence to DGCOMP about the company’s dominant position.

What happens next?

Vestager will issue a Statement of Objections (which, sadly, won’t be public) some time in the next few weeks; that will go to Google, which will redact the commercially confidential bits, then send it back to Vestager, who will show it to complainants (of whom there are quite a few), who will comment and then give it back to Vestager.

Then the hard work starts. Whether Google seeks to settle will depend on what Vestager is demanding. Will she try to forestall Google from foreclosing emerging spaces – the future verticals we don’t know about? Or just try to change how it treats existing verticals? (Ideally, she’d do both.) Many of the issues around scraping and portability of advertising which Almunia enumerated in May 2012 have been settled already (now that Google has wrapped them up; the scraped datasets aren’t coming out of its data roach motel).

Neither is going to make all the revenue lost to Google favouring its own services come back. And as with record labels and YouTube, it’s likely that Google will try to stretch this out for as long as possible; the more it does, the more money it gets, and the less leverage its rivals have.

Even so, I can’t help thinking that rather as with Microsoft and Internet Explorer, the chance to act decisively has long been missed. Instead, a different phenomenon is pushing Google’s dominance on the desktop aside: mobile. Mobile ads are cheaper, see fewer clicks, and search is used less compared to apps. I’d love to see a breakdown of Google’s income from mobile between app sales and search ad sales (and YouTube ad sales): I wonder if apps might be the bigger revenue generator. Yelp, meanwhile, seems to do OK in the new world of mobile. It’s possible – maybe even likely – that Google’s dominance of the desktop will be, like Microsoft, broken not by the actions of legislators but by the broader change in technologies.

Right and wrong lessons

But the wrong lesson to take from that would be “legislators shouldn’t do anything”. Because there’s always the potential for inaction to corner a market and foreclose on real innovation. Big companies which become dominant need to worry that legislators will come after them, because even that consideration makes them play more fairly.

And that’s why the Google tussle with the FTC and EC matters. It might not make any difference to those that feel wronged by Google on the desktop. But it could forestall whoever comes next, and it will focus the minds of the legislators and the would-be rivals. Google might not have any friends. There might come a time when it will wish it had some, though.

Samsung knows how many replacement batteries it sells. Which is why the S6 doesn’t have one

A replaceable battery for the S6. Photo by japanexpertena.se on Flickr.

I’m interested to see the result of this poll on Android Pit, which asks Samsung owners how important a replaceable battery and (separately) microSD card is to them.

Part of my interest is that I helped inspire it, through a conversation on Twitter. Modest, moi?

Ever since it was announced that the Galaxy S6 wouldn’t offer either a removable battery or a microSD slot, there was all sorts of kerfuffle on tech blogs, and the comments therein: people said that they bought Samsung stuff specifically for those elements, and that those were key things which set them apart from the (reviled, in their eyes) iPhone range, which has never offered a removable battery or slot-in storage.

However, I’m pretty sure that Samsung’s move is not only idealistic – not having to make the back removable avoids all sorts of design compromises – but also driven by clear data.

Consider this: Samsung knows exactly how many Galaxy S phones it has sold. It also knows exactly how many replacement batteries it has sold – it certifies them or similarly gets data from any company that’s selling it with its name. So that’s the battery story sorted. (See the update below if you’re just now saying “ah, but third-party batteries…”)

I suspect that it also gets analytics on the use of microSD cards – anonymised, of course – since I know that it gets that sort of data about stylus use on the Note series (apparently about 10% of users ever use the stylus).

In which case, its decision to dump removable batteries and microSD cards is simply one where it knows a few people will be upset, but the vast majority won’t. And that once again will show that people who make a lot of noise on the various tech blogs (whether above or below the line) may feel keenly about these issues, but aren’t necessarily representative of wider use. (The comments on that Android Pit story are typical: people annoyed about the change, but no clarity on what they’ll do instead.)

I’ve long felt anyway that the arguments about replacement batteries don’t hold water. External battery chargers are comparatively cheap, and don’t require you to take the back off your phone and the battery out (which immediately means your phone has to be restarted). And a good point made by Janak Parekh via Twitter: Samsung has focussed on fast recharging for the S6. (I’ve been impressed for some time by how quickly iPhones recharge. I haven’t seen it documented in comparative benchmarks, though. Update: but of course Anandtech, the site for which nothing is not worth a benchmark and a graph, documents it. Here’s the chart comparing how quickly various top-end smartphones recharge, in its HTC M9 review: it shows that the Galaxy S5 and Note 4 have the fastest recharge times, about 10% faster than the iPhone 6.)

Similarly on the microSD point: Google offers a lot of free cloud storage, and so does Dropbox, and I don’t buy the idea that you really need to have a bazillion gigabytes of music on your phone all the time. 1GB of music is 1,000 minutes of listening time, or over 16 hours. And that’s not very much to have devoted to music. When you get back in range of Wi-Fi, you can download a whole new lot.

The direction there is only towards more availability of cloud storage, not less – so, away from microSDs. (What’s more, microSD cards are amazingly fiddly and easy to lose.)

So I don’t think that the absence of these elements is going to significantly affect sales of the S6.

However, that doesn’t mean that I think the S6 is necessarily going to do gangbusters. It’s much too early to say that. Samsung’s position at the top end is being chewed in China and elsewhere by the iPhone 6 and 6 Plus, and from below by Xiaomi and local vendors. It’s going to be a tough row to hoe.

Update: plenty of responses on Twitter (as well as here). Let’s deal with a couple.

1) “Samsung doesn’t know about sales of third-party batteries

You know, I think a company the size of Samsung might have the resources to carry out a bit of market research to find out how many third-party batteries fitting its specification are being sold.

And I’m not denying that there are people on Twitter who have bought replacement batteries and are happy with them. That’s not my point. My point is that I suspect Samsung has taken a careful look at how pissed off people are going to be if it doesn’t allow replaceable batteries, and concluded – based on sales data – that actually it’s not going to be that much of the Galaxy-buying population. I don’t see why people are struggling with the idea that a big company that’s not called Google might be data-driven in some of what it does.

Another bonus for Samsung, pointed out by @misterleoni:

Certainly, being sure that you can keep “OMG SAMSUNG BURNT DOWN MY HOUSE” stories out of the papers has its attractions. Also, for the makers of third-party batteries, there’s the chance instead to move up the value chain by developing a brand around their external battery (as in the picture at the top of this piece): much better margins doing that than being some faceless unknown inside a phone.

2) “Dammit, I need an SD card for my 200GB of music!”

Again, if you’ve got 200GB of music then you’re such an outlier you can’t see the shore any more. Some data from 2011, looking at the “average” iTunes library, found that the average person (who uploaded their list, so self-selecting, so likely larger) had 7,160 tracks in their library. At 4 minutes per song, that works out to 28GB. (At 1 min = 1MB.) I think that’s an exaggeration too – people in this Yahoo Answers thread offer much smaller numbers. I’d suggest it’s likely closer to 5GB or 10GB, like the early iPods.

So again, I’m not saying there aren’t people who want or need microSD cards to store their music or photos. I’m saying that I think Samsung has looked at the market, and is taking a decision based on that.

Of course, there’s another possibility, which is that almost everyone wants removable batteries and microSD cards, and that Samsung has decided to spite them all by forcing them to buy non-replaceable ones to force an upgrade (or costly repair), and force them to buy pricey inbuilt storage. (Well, the latter certainly works for Apple.) But that’s so customer-hostile for a company that’s in straitened times, in smartphone terms, that it fails Occam’s Razor. So I’ll stick to the simpler answer: it’s the result of analysing data, and getting simple answers that allow some design compromises (backs that come off, slots for slow microSD cards) to be jettisoned.

When will Apple build a weightless laptop? Let 26 years’ data tell us

All the talk around Apple’s new MacBook – the lack of anything other than a USB-C port and a headphone port, the new “Force Touch” trackpad (which really is amazing, and I’ll write about another time) and how light it is got me thinking.

Particularly, how light it is. That thing really doesn’t weigh at all: you pick it up, and it barely seems to be there. It’s a helluva contrast with the Apple laptops of yore; I recall people toting around PowerBook G3s. Hell, I remember when the first portables came in from Compaq: they were gigantic things which could have been used as pricey weight training add-ons.

The discussion about the ports and lightness of course centred around two sides of a really quite polarised debate. One side said: “Apple is being stupid with this ‘thin and light’ stuff. We need ports.” The other said: “People want their portable devices to be ever more thin and light, and Apple is just giving them what they want.”

Well, how far can this go? Doesn’t there come a point when there’s nothing more to be shaved off?

I thought it would be fun to dig back and see how Apple has driven the “thin and light” idea, going right back to the original Apple Portable from 1989. Look, here it is. Isn’t it adorable?

Macintosh Portable

The Macintosh Portable from 1989. Bigger than a Macbook. Source: Wikipedia.

Introduced September 1989, weighed 7.2kg, measuring 10.3cm (closed) by 38.7cm by 37.7cm for a total volume of 15,027 cubic centimetres, and density of 479 kg/m^3 – a little more dense than barley (which is 400kg/m^3).

And then you compare this year’s model, 26 years later:

The new MacBook.

It’s thin. Also, light. Look, it’s almost floating off the table.


(All the data about the Apple products, by the way, comes from the invaluable resource of Mactracker, which tells you everything about pretty much every Apple product going back to the beginning of time, which for these purposes is 1975. And before you ask why I’m not comparing it with other PC makers’ laptops, it’s because there isn’t a simple or sensible comparison to be made, nor any easily available data as there is with Mactrack, which has a neat desktop and mobile app. So there.)

So when you start adding in the data about various laptop launches, trying to focus on the ones which are 12in or 13in (so that the screen size, and hence weight, is comparable), you find a definite trend.

He, He, He

It’s this: if we were relying on straight-line trends, we’d have weightless MacBooks by 2017. Yes. Perhaps they’d be airtight and filled with helium? Why not?

Apple laptop weights since 1989

Fitting Apple laptop weights to a straight line: turns out they’ll weigh nothing by 2017. Data: Mactracker

Before you snigger and yawn, just bear in mind that hard drive makers are now making them airtight – and filling them with helium. Not laughing so hard now, are you?

However if you take the more reasonable logarithmic trend, then it looks like we’re some way off weightlessness – although we might expect a half-kilo laptop in five years or so.

Apple laptops' weight: fitted to log curve

Not as much fun: seems weightless laptops are still some distance off.

(The log curve is a better fit than the polynomial one, which I’ll use below for other graphs.)

Note too that the PowerBook 180, released in 1992, seems to have been a real outlier: much lower weight than you’d expect.

Turning the volume down

But what about volume? If Apple’s making these things thinner, while keeping the screen roughly the same size, then the volume’s going down. What’s happening there?

Apple laptops volume since 1989

Fitting a power curve works best here, but 1989 distorts it a bit

Apple laptops' volume since 1992

Fitted to a power curve (which works best): the most recent models are below trend, ie lighter than expected

I’ve done two graphs, so you can see what it looks like if you exclude the original Portable. Notice how the latest machines lie below the trend line (which seems to work best as a power, rather than logarithmic, fit). There has to be a limit on how low volume can go, of course – you need a screen, you need some sort of keyboard – but Apple has really pushed the latter down in the new MacBook: the keys have less travel and the trackpad doesn’t move. That reduces the empty space needed to fit everything in, but I’d expect we won’t see much change in the next few years.

Gravity. I mean, density

Finally, if the weight is going down, and the volume’s going down, what’s happening to the density (which, let’s remind ourselves, is weight divided by volume)?

Turns out it’s increasing – quite quickly. From that barley-equalling density, it has moved quite quickly past the density of water (with the 2003 PowerBook G4 12in) and then solid magnesium (with the 2010 MacBook Air).

Apple laptops' density over time

A straight line fit seems best here. Barley, water, magnesium… all past. But pure silicon (metal) and aluminium remain way off.

There’s however quite some way to go before it goes past solid aluminium (2700 kg/cubic metre). There’s a lovely list of densities of common materials – butter, 865kg/cu m, who’d have guessed? – which might allow one to play “guess the material the Macbook model matches”.

But what does it mean?

All those graphs. But what’s the conclusion? Quite obvious really – given enough time, Apple will produce a weightless laptop of infinite density, which will suck its buyers into its gravitational well, from which they won’t be able to emerge. It’s one way to hold on to your users.

Then again, given the gravitational pull that its designs exert on a significant chunk of the internet, you could argue that it’s doing that pretty well already.

Google’s new London retail store: only if ‘new’ means ‘four years old’ (and not a Google store)

Photo: Evening Standard in September 2011

Jonathan Prynn, writing in September 2011, described how “Google’s first store pops up in London”:

The world’s first “Google store” opened not in California but in the less glamorous setting of PC World in Tottenham Court Road at 9am.

The 285sqft pop-up “shop within a shop”, which only sells Google’s Chromebook laptop and a few accessories such as headphones, will run for three months up to Christmas.

But if the low-key experiment is successful Google could follow its great rival Apple in opening permanent stores around the world.

Unlike the hugely hyped launch of the first Apple Store in Regent Street, very few customers were even aware of the Google shop – officially known as “the Chromezone” – and there were certainly no queues round the block.

As Nate Hoffelder pointed out on Friday, this story then got rewritten as “Google’s first retail store” in March 2015 when Google did exactly the same thing, though with a bigger budget, in exactly the same store.

I didn’t even remember the 2011 setup myself, and there’s no story about it in the Guardian at the time; it whizzed past my radar.

But someone ought to have remembered, surely? Especially because it seems that the “Google Zone” idea was then rolled out to 50 PC World stores around the UK. Did nobody notice those?

As Hoffelder notes,

“so far as I can tell the only real difference between the 2011 store and the 2015 store is the devices carried. The new store carries smartwatches, Chromebooks, Android tablets, the Chromecast, and other Google devices (a lot of which didn’t exist four years ago).”

Photo: Business Insider

No Google Glass, I’m guessing. And as Hoffelder also points out, there were rumours of standalone Google retail stores in 2013 and 2014. Those haven’t panned out either.

Now, to be fair, the new look is dramatically improved on the old one. Google didn’t have much to sell then; now it can offer all those hardware products. There’s even a Google Maps installation to a screen wall that lets you “fly” across the world.

Google seems to be claiming that this is its first branded store.

That’s what the PC World press release suggests:

In a world first, Google today unveiled its first “shop in shop”

and that

This is the first Google shop experience Google has opened anywhere in the world.


“Google shop experience”? The store-in-a-store isn’t new. And Google’s own spokesman seems unaware of its previous existence in this Business Insider story.

Update: via Stuart Miles of Pocket Lint, this “branded store” isn’t actually a Google store at all – as in, Google isn’t taking the money. Here’s the evidence:

From which maybe one concludes that both Google and journalists believe the company’s spin, and that four years is long enough for something to have been completely forgotten in technology. (Thanks to commenter Paul for pointing this out.)

Although I bet if it had been either Apple or Microsoft which had tried to claim their store-in-a-store was a “first” they would have had everyone on them like a ton of bricks.

Apple’s loose ends: HomeKit, Apple TV, Photos, retina MacBook Air, Apple Pay?

There are some things so far missing from discussion about what Apple’s going to announce today, which seem worth noting here, not because they’ll necessarily happen, but because it might be worth reflecting on them if they don’t.


Remember HomeKit? Announced at WWDC 2014, it was going to be the way for apps to control smart items around your dumb house (or perhaps the other way round). Things have been awfully quiet around this, despite many manufacturers being keyed up for it as far back as January at CES. (Thanks @lessien for the reminder on this.)

Apple TV

Long overdue an update, it will anyway have to get some sort of software update so that it can be controlled by the Apple Watch. Remember how Tim Cook said that he controlled his Apple TV with his watch? You won’t be able to do that without a software update, so one is surely coming there. Quite probably this will wrap together with the HomeKit update.


There have been betas of the new Photos app (to replace the ageing iPhoto) kicking up and down the web for over a month now, so it must be time to have something concrete. This seems like a good time to do it – perhaps it would also beam them to, oh, a watch?

“Retina” MacBook Air

There was all the fuss around this a couple of months back, and now it’s gone quiet. This feels like it would be a distraction from the main event, and doesn’t fit into the flow above – HomeKit links to Apple TV links to Apple Watch, and Photos is expected – but it’s one to perhaps consider.

Apple Pay

Apple has a team in Europe who have been working with banks on implementing Apple Pay; if they’ve been doing the right work then they could have it on track to get started as soon as April in the UK, just in time for the Watch to go on sale and be used by all those people who have iPhone 5S, 6 or 6 Plus models. It would be quite a coup, and no doubt would see all sorts of records being broken. Again, this would fit into the flow, but I’m not sure how likely it is. There haven’t been any murmurings, but then again, banks are quite good at keeping secrets (such as how they’ve reduced the interest rate on your savings again while offering great rates to people joining the bank.)

I don’t expect anything around iTunes Radio or Beats. Although something on that is getting overdue, it just doesn’t gel with the whole Watch theme.

Committing acts of journalism: the New Yorker profile of Jony Ive

Pretty sure they’re not rewriting that week’s New Yorker. The Toronto Star newsroom in December 1930. Photo by Toronto History on Flickr.

Ian Parker’s magnificent profile of Jony Ive (and to a lesser extent Apple) in the New Yorker has received lots of attention – mainly from journalists at a multitude of news outlets who each spent a jolly hour or so filleting it for 23 Things You Maybe Didn’t Know But Might Be Persuaded To Read Because They’re In The Form Of A List Rather Than A 17,000-Word Article.

What’s been largely overlooked is the sheer amount of work done by Parker in putting this together. Clearly, Apple’s PR people played an important part: they set up at least three interviews with Ive (some time in late July/early August; the day after the iPhone release; some time later when they go to see the new campus being built). There’s also an interview with Tim Cook, whose time isn’t exactly limitless.

But once you’ve got past those, you realise that Parker has spoken to loads more people. Let’s list the people internal to Apple who Parker spoke to, even briefly, to quote:
• Marc Newsom
• Craig Federighi
• Jeff Williams
• Bart Andre
• Hartmut Esslinger
• Eugene Whang
• Julian Honig
• Jody Akana
• Dan Riccio
• Evans Hankey
• Alan Dye

That’s 11 people beside Ive and Cook, and that’s only inside Apple.

Now here’s a list of the people Parker spoke to outside the company:

• Laurene Powell Jobs
• Robert Brunner
• Jeremy Kuempel
• JJ Abrams
• Richard Sapper
• Richard Seymour
• Clive Grinyer
• Paola Antonelli
• Doug Satzger (who Parker contacted, but who wouldn’t comment – but Parker at least tried)
• Bob Mansfield
• Michael Ive (Ive’s father) [added after a reader familiar with the article pointed out this omission]

That’s another 10 11 people, of varying difficulty on access. (He also has a brief conversation with Heather Ive, Ive’s wife, but it hardly counts as an interview. It’s not clear from the text whether he communicated with Paul Smith to verify something about the contents of notes, but you can bet that – this being the New Yorker – the point did get fact-checked.)

One score and three

In all, that’s 23 24 people with whom he had longer or shorter interviews. And there’s also a hell of a lot of reading and information mixed in there – worn very lightly (such as the point about how much profit Apple makes during a 25-second pause, or how quickly phones were coming off Chinese assembly lines as Tim Cook announced them). Deciding what parts of 23 24 interviews to use (once you’ve transcribed them, of course) and what to throw away, plus what part of the observations around them to use (Jony Ive’s manner), and then simply writing it and getting it straight, and through fact-checking, subbing, and editing, is a huge task. At that length, it’s a short novella.

Parker’s reward? I’m sure that a ton of people have read the article and learnt from it. Lots of people have done rapid rewrites of salient bits of it, but the only piece of journalism has been Parker’s, in the first place. Sure, Apple gave him access, and lots flowed from that.

Sure, Apple chose to set it up – and this is the first substantial piece about Apple I’ve ever seen in the New Yorker, which I’ve been reading for decades – but the company wouldn’t have been able to dictate the content or conclusions; notice Parker’s sardonic take on Cook’s comments about how the forthcoming Watch, which Cook was wearing, would be all about notifications: “I noticed that, at this moment in the history of personal technology, Cook still uses notifications in the form of a young woman appearing silently from nowhere to hold a sheet of paper in his line of sight.”

And the access that Parker managed to get to other people? Clearly, he was helped by the name of the New Yorker (and quite possibly some nudging from Apple to people like Powell Jobs and Bob Mansfield).

Views differ. So go and find some

But the contrast between Parker’s in-depth cover-the-bases profile, and the many pieces that get thrown off every day on news outlets, where the writer doesn’t bother to get an opposing, or a neutral, or indeed any outside view about a piece of information – bugs me often. When I worked at New Scientist, it was an ironclad rule (and I think, reading it, still is) that in any news piece you were writing, you had to get an outside opinion about whatever marvel had been unveiled, whether it was the birth of the universe or the discovery of a new species of beetle under a log in the Amazon. As a writer, this put one to the test, and the stress was only magnified at a daily paper when one had to be able to find people willing to do similar things to tight deadlines.

Yet those outside voices play an important part in news: they counterbalance, and stop the pell-mell rush towards what can otherwise be the recycling of press releases. It dismays me when, as often happens, a Big Company announces something which is both (a) years away from fruition (b) being spun furiously for its “innovative” potential, and then sees yards of approving writeups in which no news editor has thrown the copy back at the writer and said “No. Go back and find someone – preferably two people – with an opinion about it.”

Yes, I know, web deadlines, bla bla bla. I cleave to the view that if someone else has already written the story, your job as a journalist/writer is to move the story on – do something more, get a new perspective, find out something new, unearth the fact nobody else has brought to light.

This applies even more when everyone’s rushing to put something, dear god anything, online so Google News will anoint it as this minute’s top story. If you move the story on, then you become the one they look to. Quality will tell because monkeys (well, software) will eventually eat those jobs doing rewrite anyway. Look, a company that automatically writes sports and finance stories just sold for $80m. It’s coming for your listicles next.

Parker, obviously, demonstrates that quality becomes a virtuous circle. Do it well, and you get better access. But you have to do it well in the first place.

How many Android Wear devices out of 720,000 were activated in 2014? Here’s the number

Canalys announced on Wednesday that 720,000 Android Wear devices shipped in 2014.

But how many got onto peoples’ wrists? You can get an idea by looking at the number of downloads of the Android Wear app on Google Play.

However it’s intentionally vague – Google just gives you limits between which the install number lies. The levels are between 5k-10k installs, 10k-50k installs, 50k-100k installs, 100k-500k installs, 500k-1m installs, and so on.

Digging in

If you do some digging on the past captures of the Android Wear page on Google Play, you can quite quickly figure out when the number of installs passed various milestones. (Updates don’t count as extra downloads, so each “download” is a unique activation.)

A little bit of twiddling shows that the Android Wear app
• passed 10,000 downloads on July 3
• passed 50,000 downloads on July 15 (good progress!)
• passed 100,000 downloads on August 3
• passed 500,000 downloads between December 2 and 9.

The page also includes the number of reviews, which is also a useful thing to track.

So I graphed it out, and once you add in the download breakpoints, and especially the reviews, there’s a fairly clear straight-line relationship over time.

Here is is with just the download breakpoints:

Android Wear: download breakpoints

Using the Internet Archive shows you when downloads of Android Wear passed particular points.

Then we can draw a straight-line trendline

Android Wear: trendline

The download breakpoints offer a straight line interpolation

The number of reviews (graphed on the right-hand axis) seems to reinforce that there’s roughly a straight-line relationship.

Android Wear: reviews

Adding in the number of reviews from the app page suggests a straight-line growth in downloads

If you look in more detail, there seems to have been a rush of reviews early on, a bit of a dip from August to the end of October, and a fairly clear linear relationship after that. But that’s not surprising, given that Google handed out a few thousand Android Wear devices at Google I/O in June.

So that trendline looks even better once you’ve got the reviews graphed:

Android Wear: downloads and reviews

Downloads and reviews of Android Wear seem to agree pretty well. So how many by the end of 2014?

That’s pretty solid. And here’s what this graph says: by December 30, there had been about 560,000 downloads of Android Wear.

By 11 February, based on the number of reviews, it’s around 700,000.

Android Wear: all the numbers

Put it together, and we have about 560,000 Android Wear activations by the end of 2014, and 700,000 to mid-February.

Is that a good or bad figure? I’ve no idea, to be honest. (If you assume 1.2bn Google Android phones in use, with Android Wear working on everything from 4.3 upwards, which takes in 47.6% of Android phones, then you have a potential addressable market of over 570m phones. (If you take it as 1bn Google Android phones, then it’s an addressable market of 476m.)

The much more important question – which I don’t know how you measure – is how many are still being used. Abandonment of wearables has been quite high, at around a third of all users.

Update: at the suggestion of @jasonostrander on Twitter, one might get an idea of the split of use via watchface downloads. There’s a ton of them, and people seem to have downloaded at least two each.

Android OEM profitability, and the most surprising number from Q4’s smartphone market

I’ll admit – I’m surprised myself by how low the profitability even of “high-end” Android OEMs is. I hadn’t investigated it in detail until I wrote this post.

The end of 2014’s fourth calendar quarter, and hence year, brought forth a blizzard of data about the mobile and the smartphone markets. The mobile market (so including featurephones) passed the 500m mark for the quarter, according to Counterpoint. The smartphone chunk is growing as a proportion of that faster than ever: in Q4 it made up about 75% of sales.

As a proportion, smartphone sales are rising healthily:

Smartphones are a growing proportion of all mobile phone sales

Smartphones are a growing proportion of all mobile phone sales

That compares with a 58% smartphone mix in 4Q 2013. Even so, I don’t expect the 90% mark to be hit before 1Q 2018 (yes, 2018), on the assumption their sales rise as a diffusion curve.

If the proportion continues at about a 3% increase in smartphone share per quarter (as happened roughly in 2014), smartphones will be 90% of sales in 1Q 2016 – just a year away – and 95% in 3Q 2016. By then, effectively all the market is smartphones.

Apple’s enormous sales – 74.5m units shifted – attracted lots of the attention. It ended the quarter with less inventory in the channel than at the start, suggesting that sell-through (ie the number bought by people) was actually higher.

But the really surprising figure – the one that had me firing off emails asking for more detail – came not from Apple, nor any individual company, but from ABI Research. Its headline: “Android Smartphone Shipments Fall for the First Time”.

Android shipments fell in Q4 2014, ABI says

This is surprising: Android shipments have never fallen from quarter to quarter before

Wow. I mean, truly wow. As it says, that’s never happened before. Android shipments have always increased from quarter to quarter, both for “Google Android” and AOSP, since the platform’s first phone. (Unlike pretty much every other research company, ABI also breaks its Android figures down into “Google Android” – ie Google Mobile Services certified, carrying all Google’s services – and “AOSP” – principally, China.)

Yet here ABI is, saying G-Android shipments fell by 11.9m, and AOSP by 0.47m, a total of 12.4m. That’s quite a lot more than a margin of error.

Now, you could object right away. There are lots and lots of research companies to choose from, and all are pushing their own datasets, and all diss each others’ datasets in more or less subtle ways. There’s IDC, Gartner, Counterpoint, ABI, Strategy Analytics, Canalys, CCS Insight, Kantar, and more. Couldn’t ABI be wrong?

It could. But then you also have the numbers that come from the companies themselves. Samsung helpfully said that in the fourth quarter it shipped a total of 95m mobile phones (ie featurephones and smartphones) – and that smartphones were in the “high 70 percent” range.

So… how high in that 70s exactly? Because if it wasn’t in the 78.5% or above range, it was less than Apple. But Samsung wouldn’t want to admit that. Sony also reported numbers, as did LG, and Lenovo. LG’s numbers fell from 3Q to 4Q, Sony’s grew, Samsung’s fell, Lenovo’s and Motorola’s both fell independently. Together, they lost 6.5m sales from 3Q to 4Q, at a time when you’d expect sales to rise.

Let Jeff Orr of ABI Research explain it in a bit more detail.

“China was definitely the lead influencer in the OS share change during 4Q’14. ABI will be a lot more precise once the final vendor tallies are announced and tallied.

“Thinking about Apple’s financial call, Tim Cook highlighted exceptional iPhone growth in the US (up 44% QoQ) and BRIC [Brazil, Russia, India, China] region (up 97% QoQ), adding that China was the company’s second largest iPhone customer for the quarter and that Singapore and Brazil also saw significant increases (though these last two were not given any numbers).

“Apple also noted the amount of Android-to-iPhone switching it observed in the quarter. Given Apple’s premium price points in the smartphone world, one has to believe that switching impacted a subset of all Android handset makers that would be comparable in ASP to iPhone (or at least within a similar pricing tier). Again, it’s too early to name names here just yet, but you can probably guess the likely Android vendor names to put in that bucket.”

The competitors
Basically, it’s the big companies that compete directly with Apple – so Samsung, Sony, LG, HTC – and to a lesser extent Chinese companies such as Lenovo and perhaps (though it’s not certain) Xiaomi, whose shipments fell from the third to the fourth quarter.

Apple stole customers who might otherwise have bought Android phones away from those makers, and that helped cause a fall in sales.

How many iPhone users upgraded?
Note Cook’s comments on the earnings call about how many of the existing iPhone base upgraded to the iPhone 6 and 6 Plus: he put it in the “low to mid teens percent of the existing base” (these chief executives and their vague percentages, eh?). So how many is that?

Assuming a 400m iPhone base before the fourth quarter, and 13% upgrading to the 6/6 Plus (the minimum possible): 52m iPhones.

Assuming a 440m iPhone base, and 15% upgrading: 66m.

So if you subtract those from the sales figures, you get between 8.5m and 22.5m who were new to the iPhone – either upgrading from a featurephone, or coming over from Android.

Split the difference and you get 15.5m. That’s surprisingly close to the 12.4m figure earlier of “missing” Android sales from ABI Research; and you’d also expect that there would be sales growth from the third to the fourth quarter, which is the largest by volume of the year. Perhaps up to.. 15.5m more phones? (Update: Canaccord puts the user base at 404m, and the upgrade percentage at 15%, which gives 60.6m upgrades, and so 14.4m upgrading a featurephone or coming from Android.) (Second update: Mav, in the comments, points out that the total iPhone 6/6 Plus sales includes those sold in the third calendar quarter, at the end of September when they went on sale. That means an extra 10m and more sales of the iPhone 6/6 Plus. So it’s a very substantial landgrab from featurephones and Android.)

So it looks like Apple actually skimmed off some of Android’s growth in the fourth quarter. What’s worse for the Android OEMs is that Apple tends to grab two classes of customers: the loyal ones (who just keep buying Apple stuff – see how the majority of buyers were loyal ones here), and the high-paying ones, especially in China and the west, where its brand is able to command a premium price. Then it converts those high-paying customers into loyal ones.

That’s notable from Samsung’s apparent mobile ASP (average selling price – what it gets from carriers and wholesalers), which dipped badly in Q3 and came back only slightly in Q4, helped by the Galaxy Note 4. Even that couldn’t help against Apple’s iPhone 6 Plus – note how Apple’s ASP rose while Samsung’s fell in Q4.

iPhone v Samsung average selling price

Figures for Samsung ignore revenues from featurephones, which are comparatively tiny compared to its smartphone revenues

Samsung’s smartphone shipments are reckoned to have fallen from 78.5m in the third quarter – to (if we’re generous) 74.5m, equal to Apple, in the fourth. That’s down from 82m a year before (a 9% fall).

The value trap, Android version
Let’s be clear: I think Android is a boon to the world; quite possibly it’s the best invention of this century so far. (I’ve said as much many times, but some people find this hard to understand.) It’s a great thing that people who previously couldn’t get internet access at all can now get a cheap handheld device capable of running apps that can provide all sorts of information, and use far less power and are far more portable than a PC. Smartphones put power in the palm of your hand; and Android is the OS of choice for that.

However, for handset manufacturers, Android isn’t such a boon. Take a look at the operating profit and margins for the top-end Android OEMs. (I use operating profit rather than gross profit because it takes into account the costs of actually competing in the market through marketing, R+D and so on – not just what you take over the counter).

I’ve estimated HTC’s shipments at 6m, based on Sony and LG’s revenues and shipment numbers (it comes out between 5m and 7m, depending which you use).

Top-end Android handset revenue and profits, compared to Apple

OEM Handset revenue US$ (approx) Operating profit US$m Operating margin % handsets shipped Implied ASP per phone Implied profit per phone
HTC $1.6bn $6m 0.38% 6m *est $266 $1
Sony $3.6bn $79m 2.2% 11.9m $305 $6.64
LG $3.45bn $74m 1.8% 15.65m $220 $4.72
Lenovo (inc Motorola) $3.39bn -$89m -2.6% 24.7m $137 -$3.13
Samsung $22.8bn $1,790m 10.0% 74.5m $306 $24.02
for Android
$34.84bn $1.86bn 5.3% 132.75m $262 $14.01
Apple $51.2bn $14.3bn
(at 28% margin)
28% (est) 74.5m $687

Note that these are rough-and-ready figures. Here are a few of the caveats.
• For Sony, Lenovo and Samsung I haven’t accounted for tablet sales, which in Samsung’s case were 11m at an unknown ASP; for Lenovo tablet numbers weren’t stated (though it puts them at 4.8% worldwide, which would be 3.6m out of 76.1m). I haven’t even tried to estimate them for HTC, which in its financial report on Q4 doesn’t even mention the Nexus 9 it made for Google which went on sale in Q4.

• The Samsung figures overstate smartphone revenues, because they ignore the 20m or so featurephones which will have had an ASP of around $15 and unknown profit. Samsung’s mobile profits have risen in line with its smartphone shipments, so we can reason that its featurephones have less profit than smartphones; so ignoring their revenue and assigning their profits to smartphones is generous, but not madly so. Also, Samsung’s profit figure is for its “IM” division, which includes PCs, of which it sells a few million per quarter but probably doesn’t get much more than $10-20 of operating profit per sale.

• ASPs are thus approximate, and so are profit estimates: tablets might be money losers, depressing the apparent profitability of the handset business. But what all these companies except Samsung have in common is that their handset businesses have lost, or are losing, money. So these probably aren’t that far off; tablets might be less profitable, but there are fewer of them to spoil the numbers.

• These are only top-end Android phone makers. There is a ton of others: ABI reckons there were a total of 303m G-Android and AOSP phones shipped in the quarter, so more than half of those aren’t accounted for in this total. Many of them are cheaper (Huawei, ZTE, Coolpad, Micromax come to mind) and ship in volume. Being cheaper, they’re unlikely to make a lot more profit, and since they don’t appear in the top five, they probably don’t make much difference to these figures except downwards in ASP and per-handset profit.

• Apple gives the number of phones shipped and total revenues (so you can calculate ASP) but doesn’t break out the operating profit of its divisions; analysts make estimates. I got the 28% operating profit figure from this 2013 analysis by Canaccord. Update: for 4Q 2014, Canaccord puts iPhone operating profit at 38%, which is colossal, and would make that table look even more lopsided. (Canaccord reckons Apple has 79% of mobile profits, and Samsung 25%; the others have between 0% and 1%, if they’re profitable.)

So what does this tell us?
That 2013 Cannacord estimate isn’t the most recent but things are unlikely to have shifted far – and even if they’ve moved by a few percentage points, it still doesn’t change the overall picture. Apple makes a ton of profit per phone, and top-end Android OEMs generally don’t. (Motorola has been a basket case for years, and dragged down Lenovo’s figures; see Jan Dawson’s analysis of how that’s affected the mobile bottom line.)

You can argue – and lots of people do – that Apple is therefore “charging too much for the iPhone, and it should cut the price so that more people would buy it”. This is superficially attractive logic to people who (a) aren’t running a business and (b) can’t think long-term.

For Apple, the cash it rakes in is used to reinvest in factory and supply contracts in ventures for forthcoming products. If it didn’t have that surplus cash, it couldn’t buy fingerprint reader companies, lock up supplies of camera sensors, guarantee enough factory production to make 74.5m phones, fund machinery to diamond mill the sides of phones, and pay forward for whatever it’s going to do over the next two, five, ten years.

(But what about that cash mountain? Well, lots of that is profits earned abroad that Apple doesn’t want to repatriate to the US, because that would attract a high rate of interest tax, so it leaves it sitting in Ireland and reinvests it in those things as above.)

By contrast, the Android OEMs whose phone divisions are living hand-to-mouth on those incredibly slim margins can’t afford to reinvest. They’re essentially at the mercy of the rest of the smartphone and component ecosystem.

For example, they can implement a fingerprint scanner (HTC and Samsung have) but it’s incomplete; HTC didn’t use it across all its models and it wasn’t part of a payment system – as Apple Pay is, carefully planned over a two-year arc. Similarly, 64-bit Android hasn’t happened to any appreciable extent, and while you can argue about whether 64-bit makes a difference (these ARM engineers reckon it does, and explain why), Apple is still a mile ahead of the rest in implementing it. Lots of Android has to move at the pace of the slowest part of the hardware ecosystem, much of which is 32-bit.

Note that three of the four top-line OEMs are part of large conglomerates which collectively make everything from camera sensors to games consoles to TVs to washing machines to memory chips. That means the smartphone divisions are effectively a bit of icing on the main, hopefully profitable, other parts of the business, and also that they can bear quite sizeable losses for a while (LG and Sony have). For Samsung, mobile enjoyed a spell in the sun; now the chip business has become dominant again. HTC’s survival is anomalous, but somehow heartening.

High-end Android – trying to compete with Apple, especially in the fourth and first quarters (because the latter is a gift-giving time in China particularly) – is becoming a rich man’s game, with low returns. Nor are these ASPs and profits for the fourth quarter unusual; tracking them over time you see similar figures. LG’s average smartphone operating margin for the past 8 quarters is 1.3%; for HTC it’s -1.4%; for Sony it’s a few percent.

In fact, they’re all caught in the “value trap” that I wrote about a while back for the PC market: because they don’t control the software, there’s little chance to differentiate. These companies are vulnerable to customers who choose simply on price. That means only those who can manufacture at scale or compete locally can benefit.

So what’s do we conclude?
• Android will continue to be gigantic. For local OEMs in countries like India and China, it offers huge opportunities for scale

• high-end Android handset makers will keep struggling, against Apple and notably against Samsung – which is meanwhile struggling with the aforementioned local OEMs (which eat into its low-ASP base) and Apple, which is stealing its top-end customers

• there’s little opportunity even for high-end Android OEMs to invest and innovate, because it’s not profitable enough. Only Samsung is an exception, because it’s part of a gigantic conglomerate. All are weak in software, and there’s no sign of that changing.

The giant in the niche

Apple's share of the overall mobile market

A niche player? Apple’s share of the overall mobile market (including featurephones) is at 10% for the past year, and 15% in the past quarter.

Apple’s a niche player – if that’s what you want to call the largest smartphone OEM – but it’s the most valuable niche, and also the one that lets it decide what people view as “valuable” in a phone. (For example, waterproofing hasn’t helped the Sony Xperia range or the Galaxy S5 sell, despite being a distinction compared to the iPhone range. Having a fingerprint sensor apparently has helped the iPhone 5S onwards.)

If Apple has the most valuable and the most loyal customers (and especially if it gets the most loyal valuable customers) then that means it can continue to expand its ecosystem, continue to charge a premium, continue to make big profits, continue to buy up companies that rivals had their eyes on, aim to undercut others in the services it offers.

Skimming off the top end gives Apple huge leverage. Ben Bajarin reckons Apple has about 60% to 70% of the “premium” market. On the basis of that 400m-440m user base (and don’t forget another 100m iPads and some iPod Touches too, though probably with a lot of overlap), that suggests there’s a total premium market worldwide of about 750m smartphone users worldwide. If Apple keeps pulling in 10m or so of them every quarter, it’s going to be a monster.

Update: Canaccord’s latest February 2015 calculation (which appeared the day I published this) reckons it will reach 650m by the end of 2018 – though it thinks that will be only a third of a global premium audience of 1.8bn. Android’s got plenty of room to grow. But so, it seems, has Apple.

In the land of the refuseniks: what Kantar’s latest smartphone data reveals

Kantar ComTech Worldpanel’s latest numbers for smartphone sales share are out (or dig the groovy but very dark animation), and pretty much as expected they show that iPhones have had a terrific time in sales terms for the three months to the end of November – so that covers September, October, and November, basically all the time that the new range has been on sale.

My only frustration with Kantar is that it doesn’t index the numbers from year to year; there’s no way to know if sales in 2014 are lower, higher or the same than the previous year or year before. You might think that because more people own smartphones that volume must be increasing, but it’s not necessarily the case; GfK data suggested that mobile phone sales value fell in Christmas 2013 compared to 2012, though increased in 2014 (helped, one suspects, by the Apple phones).

I’ve asked Kantar previously to include an indexing figure (eg 2012 = 100, 2013 = 105, ie 5% greater in sales volume) with these numbers, but they haven’t. (In a future post I’ll try estimating this from general levels of mobile sales from quarter to quarter, and country populations and smartphone shares.)

In the absence of that, we just have the graphs – which I’ll put below. But there’s a much more interesting story which hasn’t been picked out of the press release, though they put it in there. It’s about the refuseniks: the people who have a featurephone, but are determined not to move to a smartphone.

US smartphone sales share to end November 2014

Data from Kantar. iPhone share is high, but not as high as the iPhone 5 launch. (No comparative data on volumes, however.)

UK smartphone sales share to end November 2014

UK smartphone sales share to end November 2014, from Kantar; iOS share is highest ever recorded

EU5 and Australia smartphone sales share

Share of smartphone sales for three months to end November 2014, via Kantar. (EU5 = Germany, UK, France, Italy, Spain = 184m smartphone users; Australia = 16.4m smartphone users)

(Note among all these that Windows Phone is still not making anything resembling progress; nor is “other”, though Firefox phones are few and very far between in these countries. The smartphone platform space is played out.)

Here’s the most interesting part, left to the very last paragraph of the release:

Smartphone penetration reached 58% in the US and 65% across Europe’s big five economies. “While die-hard featurephone owners state they are not planning to buy a smartphone in the next 12 months, they might not have a choice as vendors continue to transition their portfolio away from featurephones to smartphones”, concluded [Kantar head of research Carolina] Milanesi. Forty-seven percent of featurephones owners looking to change their current device in the next six months in the US and 35% across Europe’s top five [countries] are not planning to upgrade to a smartphone.

Once again: 47% and 35% of featurephone owners in the US and EU5 (Germany, UK, Italy, Spain, France) who are looking to change device won’t go to a smartphone.

Now, you could flip those numbers over: out of all (remaining) featurephone owners, 53% in the US and 65% in the EU5 may go to a smartphone in the next six months. It’s only “may” because they haven’t said they will, only that it isn’t definite they won’t.

The longrunning ComScore data in the US, meanwhile, which tracks installed base, shows that at the end of October there were about 65m featurephone users remaining there. For the UK, I calculate that smartphone penetration of mobile phone users is now 80.2% – based on the data and calculations I did for a piece last April using Kantar’s data.

In the US, the number of featurephone users converting to smartphones had been fairly constant, at a few million per month, but as the graph below shows, it nosedived during 2014 (the data only goes up to the end of October) – indicating that fewer are shifting up to smartphones.

Number of US featurephone users moving to smartphones, by month

Data calculated from ComScore, showing how many featurephone users shifted to smartphones in the US. The trendline seems to have fallen off in 2014.

And indeed, why would you shift to a smartphone in the US if you don’t particularly need one, given its crazy system where you pay a huge per-month fee for the phone, and then for carrier service, and then for data, and then perhaps for “extras” (loosely defined)?

Thus we may now be at the point where the only ones without smartphones are the refuseniks – the people who don’t want a smartphone. In other words, we’re hitting the “laggards” – the 16% who don’t care. Given that the best fit for the data suggests smartphone penetration will top out at 92%-95%, and with 80% already using them (in the UK), we’re clearly in laggard territory.

Diffusion of innovations: segmentation

Stages of adoption of innovations. Source: Wikipedia

What does this mean? Mostly, that selling more phones becomes a battle where the existing smartphone platforms try to win people over from their rivals, while there’s also a gradual accretion from the last featurephone holdouts – who, as Milanesi says, might find they have barely any choice when it comes to replacing their dead one.

For BlackBerry, this is almost played out; it’s down to something like 40m users worldwide, of whom perhaps 8m are consumers and thus remain to be poached.

For Windows Phone – well, it’s unclear quite what’s happening there. The numbers sold remain consistently small, and low-end, and there just doesn’t seem to be a lot of movement. There’s some anecdotal suggestions that many of the low-end ones are used as sort of semi-smartphones, with pay-as-you-go contracts and little use made of their internet capabilities.

Android and iOS

The more interesting flow, aka churn, is between iOS and Android, and Android and iOS: in percentage terms for the three months, 11% of iPhone buyers were previously Android users; and 13% of Android buyers were previously iPhone buyers.

I know – the simplistic view would be that iOS is losing users overall. Except the numbers don’t work out that way; there are more Android users than iPhone users in the US (92m v 73.7m), so that iOS is actually gaining users.

Here’s how:
• 13% of 73.7 = 9.6 iPhone users shifting to Android;
• 11% of 92 = 10.1 Android users shifting to iOS.
Assume that the number of people changing phone is proportional to the total installed base at any time (which is likely), and the iPhone user base grows – just.

Even so, that churn must be a concern to Apple. Maybe the new screen sizes of the iPhone 6 and 6 Plus will reduce that, or perhaps it’s inherent in the dynamics of smartphone platforms.

Meanwhile, if you know a refusenik, do ask them what they’ll do when their phone breaks. Smartphone without data? Buy a featurephone on eBay?

Whatever; it’s unlikely the refuseniks are going to make a big impact now. In the developed world, the smartphone platform wars are so played out.

Analysis: another BlackBerry quarter: there must be a horse in there somewhere

Those BlackBerry results for the quarter to the end of November. Terrible? Yes. A fall in revenues quarter-on-quarter, and a whopping operating and net loss. The title of this post is, yes, a reference to the old joke – the guy with a spade who is working his way through a mountain of horse crap. Someone comes up and asks him what he’s doing. “With this much crap, there’s gotta be a horse in here somewhere,” the man replies. In the same way, I’m fairly sure there’s a profitable business somewhere inside BlackBerry. The trouble is finding it. John Chen has done an impressive job since he came on board just over a year ago. But he hasn’t found the profit either (we’ll get to the little financial twiddling that let Chen claim a profit later). I wrote about how much trouble BlackBerry is in after its last quarter. How are things three months on? There’s obviously a profitable business hidden in there: it provides high-security keyboard-based smartphones to governments and businesses which put a premium on connectivity and security. The trouble is extricating that from the loss-making consumer side, and the high costs of making smartphones at far less than scale, and how to fight off the competition to be in charge of mobile device management (MDM) at big companies and in governments. So, let’s start digging. (Click on images for a large-sized version.) Bad headlines The headline numbers are bad. Revenue fell 33.5% year-on-year to $793m, and 14% sequentially.


BlackBerry revenues by geographic region, by quarter

Shown separately: EMEA is the largest, but all are struggling. Source: BlackBerry

BlackBerry geographic revenue by quarter

EMEA is still the biggest region by revenue, but it’s all got much smaller. Source: BlackBerry

The graph shows how things aren’t working for the company. EMEA (Europe, Middle East and Africa) has long been the largest segment, principally because of subscribers in South Africa and the rest of Africa; the UK used to be a stronghold too, but that’s faded. Asia-Pacific hasn’t looked too rosy either for the past few quarters; but those are consumer areas, and that side of BlackBerry’s business just isn’t happening.


BlackBerry shipped 2m units but “sell-through” (to end users) was 1.9m, which suggests there’s no surplus in the channel now; it’s cleaned out. Does that mean there are no more BB7 devices like the Bold to buy? Again, unclear. But the ASP (average selling price) cratered:

Derived from total hardware revenue divided by handsets shipped. Source: BlackBerry

Derived from total hardware revenue divided by handsets shipped. Source: BlackBerry

Note that the ASP has halved from the December 2008 quarter, six years ago. Chen expects that to pick up now. Yet BlackBerry is struggling to fulfil even quite small demands, it seems. Chen is very vague about how many Passports were shipped (and paid for) in Q3 (transcript from Seeking Alpha)

Going forward, because of the new products, we expect the ASP to start picking back up again. We’re able to fulfil about 200,000 Passport orders that was preordered at the time we announced it, while reducing the manufacturing lead times to roughly now between four to six weeks. However, because we have sold our stock out a number of times in the quarter waiting for the fulfilment, we were only able to fulfil order backlog of Q3 by December 12. So that was clearly already into Q4. And I also want to remind everybody that our revenue of these devices are all recognised on a sell-through basis. So not every one of those units, in fact most of those units revenue are not recognised in Q3. And we obviously will recognise as they lead [ph] up in throughout the next few quarters.

Data point: 61% of Passport sales were in the US. See how much that helped revenues? Oh, sure, it didn’t. Revenues fell in North America. Also, Chen said that the hardware side could be profitable with sales of 10m units per (fiscal) year. It’s some way from that: 5.7m shipped after three quarters of the fiscal year (7m in the past four quarters); or 6.8m “sell-in” (to customers) in three quarters – and the magic 10.2m of sell-in over the past four quarters.


BlackBerry has made lots of profit from services. But it needs to make revenues from services for that to work. The Service Activation Fee (SAF) is what it gets from people who activate BB7 phones. And it’s declining: SAF revenues were down 13% sequentially, and will fall by about 15% in the next quarter according to the finance director. Looking at BlackBerry’s service revenues, those fell from $424m to $364.8m – which is a 13% fall. So that suggests that most of its service revenues are from SAF. That makes sense, looking at how the service revenues track the numbers of subscribers. The correlation between published subscriber figures and service revenues is really strong:

Correlation between BlackBerry subscribers and service revenue

Service revenues and subscriber numbers are closely linked – BlackBerry used to get more money per subscriber than it does now.

The linking lines show the progression of the link between revenue and subscribers; notice how it goes below the correlation line to begin with (more money for fewer subscribers) and then goes above the line, and starts heading backwards (less money for the same number of subscribers). So whereas in the June-August 2010 quarter there were 50m subscribers generating $785m in revenue, for the March-June 2014 quarter, when there were 50m subscribers, they only generated $519m in revenue. And that’s going to continue to decline; on the 15% decline forecast by Chen, that means a $310m service revenue announced in March, covering the current quarter (to the end of February).


For some time the best questions in the BlackBerry analyst call have come from Ehud Gelblum at Citi, and he didn’t disappoint. “Is there a subscriber number update that we can hang our hat on?” he asked Chen. Answer: “No, not right now.” We can do some calculating about subscriber numbers, though. There’s a figure of “service revenue per subscriber account” that you can work out from the published number of subscribers and the service revenues. To get to the drop in service revenues, there are three choices: • subscriber numbers stayed the same, but per-subscriber payments fell • subscriber numbers fell, but per-subscriber payments stayed the same • both subscriber numbers and per-subscriber payments fell. If it was the first, wouldn’t you expect Chen to have spoken up and spun it a bit? He didn’t. If it were the second, wouldn’t you expect him to spin it a bit? He didn’t. There’s a clue, though. Here’s Chen talking about the problem of going from BB7 to BB10:

I’m not go to be providing you the margin of the Classic but it is a positive margin and revenue of course is in the $400 plus. But then I lose $3 to $5 a month [in service fees] when that conversion happens. Sometime over the lifetime of this will cross over.

At $3 per month in service revenues, $364.78m suggests about 40m subscribers. If we model about 5m fewer subscribers per quarter over the past four or five quarters, we also get a consistent fall in per-account subscription.

BlackBerry subscriber numbers

The fall in BlackBerry subscriber numbers (red = estimated, based on public data) is pretty fast.

BlackBerry service revenue per subscriber account

Service revenue per subscriber account is constantly falling as consumers, carriers and businesses move away from the Service Activation Fee (SAF).

And according to Richard Yersh, about 80% of those subscribers are business users – so that’s 32m business users and 8m consumers on BB7. On the upside, with 6.9m business licences (pretty much all on free tryout) for its EZPass BES12 server, that does mean lots of potential clients. If – big if – it can get them to sign up for its services. The other problem about falling subscriber numbers is that it gives you fewer people to sell new handsets to. On a basis of a two-year refresh cycle, the further you fall below 40m, the lower the refresh. Over the past seven quarters, the sell-through (to end users) has been 24.6m – average 3.5m per quarter – which would be 28m per two years. Trouble is, the sell-through is falling; the latest is the smallest recorded. Why should things improve?


Still, it made a profit, right? CFO James Yersh:

In the quarter we also turned in a non-GAAP net profit of $6 million or $0.01 per share. These results were largely attributable to disciplined management of margins and expenses.

“Non-GAAP” means “not using official measurements”. (“GAAP” is “Generally Accepted Accounting Principles”.) I don’t mind companies leaving out exceptionals – one-off payments for laying off staff or closing factories, say. The reasons here are trickier: GAAP operating expenses included $150m of revaluation of its $1bn of convertible debt, the value of which increased. It’s a non-cash charge, and doesn’t affect anything about how the business is actually run. Just as a flashback, the reason why BlackBerry is carrying a debenture debt is because it decided to go for a bailout after a buyout effort failed. The debenture raised $1bn at 6%, giving debenture holders the right to convert the debt into $10 shares in November 2020. The debentures have to be marked-to-market – that is, valued – because they’re a continuing obligation: they’re a risk to the company, because convertible debentures can be a risk to a company through short-selling. But even if we allow the non-GAAP profit of $6m, we have to ask what BlackBerry has ahead of it. Let’s recap: • it’s struggling to fulfil hardware orders, but has cleared out a lot of BB7 inventory, so now has to try to sell lots of handsets, which will be made for it by Foxconn • service revenues are falling • subscriber numbers are falling • it now has to convert business users who had the free EZPass deal to actually pay money, rather than going with a rival for less Actually, a forecast that looks about right to me comes from one of the (many) BlackBerry analysts on Seeking Alpha, who sets out what it will take for the company to achieve profitability in its next financial year: • hardware sales of 13m over the fiscal year (my comment: possible, but it shows no signs of coming close this fiscal year) • service revenues as a result of those hardware sales of $1.27bn for the year (my comment: might just be achievable, at $300m per quarter, which is where it’s heading). There’s a newer article by the same author, who seems pretty sensible. Recommended. BlackBerry used to straddle the smartphone world; now it has become a curio, one with a niche business that it has to turn into profit as the consumer side of its business subsides. But there must be a horse in there somewhere, right?