Microsoft, capitulation and the end of Windows Everywhere

Tech culture is very fond of persistence, stubbornness, perseverance, and the idea that you should never give up. We're surrounded by stories of visionaries who were told they'd never succeed and went on to change the world. But sometimes, you should put selection bias aside and, yes, give up. 

This applies to big companies perhaps even more than for startups. Big companies have entire strategy teams devoted to working out what to do next and how to do it, and budgets to hire strategy consulting firms for millions of dollars to produce hundred-page decks with more strategies and ways to achieve them. Such people have little interest in saying 'give up - it won't work' (perhaps because that might mean you don't need a strategy team anymore). And there's no SmartArt for failure. 

Microsoft today, I think, is a case study in knowing when you should indeed give up, and what you should do after that.

As (hopefully) we all now understand, mobile is replacing the PC as the dominant computing platform. Smartphones sell in much larger numbers, have a much larger user base and are already close to taking a larger share of internet use than the PC in leading markets (such as the USA and UK). PCs aren't going away any time soon, any more than faxes or mainframes did, but they are the past, not the future. 

Since Microsoft's mobile operating systems have failed to achieve meaningful market share, Microsoft has hence gone from dominating sales of personal computing devices to powering less than a fifth. 

If you were running Microsoft, what would you do about this? 

The old option, from Steve Ballmer, was to apply the weight of Windows and Office on the PC to force Windows into mobile, using sheer effort to come from behind (as Microsoft had done many times before). Leverage Office and Windows against each other. Put together, late and in stages, an operating system that can answer iOS.  Pay developers to make apps for Windows Phone to seed the market, if you have to. Put Office onto your own platforms first, regardless of where the users are. When all else fails and Nokia tell you it's giving up, buy them out and make the handsets yourself. 

In the past, leveraging Windows and Office was the key to Microsoft's success, but that didn't work this time. Windows had actually ceased to be the dominant development platform in the late 1990s with the rise of the web (though that mattered less at the time since you still needed to go online, and for almost everyone that meant a Windows PC). Hence, though a big part of Microsoft's mobile strategy has been to push towards common code across Windows on the desktop and on mobile, so that it's easy to write apps for both at the same time, in practice that's largely irrelevant. The apps that people want on smartphones are not being written for desktop Windows anyway. Uber doesn't have a desktop Windows app, and neither does Instacart, Pinterest or Instagram. The apps and services that consumers care about are either smartphone-only or address the desktop using the web, with only partial exceptions for the enterprise. You can't tempt developers to support Windows Phone by saying 'it's easy to deploy your desktop app to mobile' if there is no desktop app. So Windows is not a point of leverage for Microsoft in mobile. Neither was Office. Few people really want to edit an Office document on a phone - a viewer is normally enough. And as Blackberry also discovered, enterprise support is not enough if the broader phone experience is sub-par. As Apple has added enterprise features, the appeal of Windows Phone has fallen away there too.  

There was a slim chance that Windows Phone might have been able to overcome all this and establish itself in 2011, when Nokia's partnership was announced. There was no chance at all by the time Microsoft bought Nokia in 2013. and so the write-down of that acquisition now comes as a matter of regret but not surprise. Windows Phone has failed to achieve enough scale to be attractive to developers: it is a third choice, or perhaps even fourth, after the greater appeal of just doing another iOS or Android project. There are fewer apps, and those that are there are later and have fewer features than those on iOS or Android. Consumers notice. Some people love their Windows Phones, but not nearly enough. 

Lumia unit sales (m)

Source: Nokia/Microsoft/a16z

So, Microsoft has missed mobile. Consumer PCs, slowly, will be a shrinking platform. Meanwhile weakness in mobile also bleeds back to the desktop and undermines Office. The shift away from the PC will be slower in the enterprise than in consumer internet, and so will the rise of alternative software models. But as I discussed here, the rise of SaaS services and new productivity models on one hand and more and more capable mobile devices on the other means that Office, and hence desktop Windows in the enterprise, is also probably a declining model. You may need a PC to run Office, but you can no longer assume you'll be running Office at all. 

This brings us to capitulation. The new CEO is acknowledging the end of 'Windows Everywhere' as the driving strategic engine for Microsoft, and also acknowledging the decline of Microsoft Office as the monolithic, universal experience for productivity. Windows Phone is no longer the centre of the strategy (if it survives at all as anything more than a Nexus-like niche). Microsoft is also suggesting that Xbox is not strategically core either, reflecting the reality that it will be the smartphone, not the TV or a box plugged into it, that will be the hub of the digital experience for most people. The smartphone is the sun and everything else orbits it. 

This is a little like Google's transition away from the plain-text web search as the centre of everything, and indeed Facebook's tentative shifts away from the Newsfeed. Microsoft has two huge, profitable businesses in Windows and Office: they will slowly go away, so how do you use them to create something new? Instead of every new project having in some way to support Office and Windows, how do you use Office and Windows to support the future? You must distinguish between things that prop up the legacy Office and Windows businesses (and Microsoft is doing plenty to do that), while using them to drive the new things.

But you also need to work out what that 'new' would look like. Google's mission is generalizable beyond 'web search' - really, (as I wrote here) it's a machine learning company whose mission is to understand everything and help you find it, and that doesn't have to mean a text search box. For Microsoft it's less clear. It delivered 'a PC on every home and on every desk', which once seemed like a crazy goal, but now mobile means 'a computer in every pocket' and Microsoft has little role to play in delivering that, so what does it do? Again, as I suggested here, enterprise platforms and productivity are going to change fundamentally, and that in turn will enable and feed off a shift away from PCs. Sharing document files (or copying them as web apps) is not the future - rather, the connective tissue of work needs to be rebuilt. By someone. I don't have a complete sense of what that looks like, but admitting defeat is the first step to working it out. 

Apple's victory laps: notes in the margin

As I wrote here at the end of last year, tracking precisely how well Apple and Google's mobile platforms are going has ceased to be very interesting. They both won, and both got most of what they wanted, more or less, and at this stage iPhone or Android phone sales announcements are really just victory laps. That's reflected in the fact that Google hasn't given a new Android sales statistic since last summer.

Meanwhile, it's too early to tell how the Apple Watch is going to do. Apple implies 2-3m unit sales and something over $1bn in revenue in less than a quarter, which clearly kills the 'flop' stories, but as this chart shows, that's not a strong indication of what will happen next. 

Hence, rather than analyzing the results in detail (what as a sell-side analyst we used to call 'maintenance research'), here are a few notes I scribbled in the margins of Twitter, together with an essential emoji summary. 

Google Now, Maps and Apple Music

Google Now looks like Google's most magical, artificially intelligent product. And in some senses it is, but it is also in some senses Google's most manual product. Each category that Now knows about and makes magical suggestions around is the result of a human making an editorial choice to create that category. Google Now gained cricket scores not because it 'learned' that there was cricket but because someone at Google wrote a recipe for cricket. 

Obviously, you can't do this for everything - Google Now cannot scale to the whole internet. It can do a lot, and certainly enough to be a great product, but it can't do everything, not until we have 'true' HAL 9000-style AI, which is a long way off. The underlying domain is not actually infinite, but like Borges's library it might as well be for any manually-edited project - Yahoo's manually-edited directory peaked at 3.2m sites but had become absurd long before that. So there are gaps - you're making something that looks like AI but isn't. Google Now covers the gaps by keeping quiet, where Siri covers the same kinds of gaps by making jokes (that was probably a better product management decision - if you baffle Siri you get a blank stare or a laugh but if you baffle Now you never know about it). 

 Contrast this with Google Maps, which is also partly a manual product. Hundreds or thousands of people drive Streetview cars down roads and hundreds or thousands of people check the images and edit the maps. There's lots of machine vision too, as well (and crowd-sourcing), but the back-stop is lots of paid employees. The algorithms alone aren't good enough yet. Editing the internet by hand was just too big, but editing maps isn't - it's just very expensive. Google is willing to spend that (and Apple is going to try too). 

So, mapping the internet by hand is impossible - you need algorithms. But mapping the world with algorithms alone is also impossible, for now - you need people to look at the data too.

I see Apple Music as sitting somewhere in this second category - manual curation at scale. Here the problem to be solved is that the commodity streaming service's 'search box plus 30m tracks' offers no way to discover anything you might like but haven't already heard of. You need personalised suggestions, from somewhere. But how do you suggest things from 30m tracks? It is in principle pretty easy to get some data about what people like, especially once they start listening (and so stop lying). But how do you know what aligns to that preference data?

Pandora applies an algorithm. Apple Music is trying the Google Maps approach: "doing this at least partly by hand is not an impossibly large problem, just a very big one". Just as Google manually adds metadata to the raw maps (street names, one-way streets, business names etc) Apple manually places tracks into playlists with its own metadata. Then it takes the interest data it has about a user and suggest 10 or 20 or 30 playlists, and, hopefully, there's a good chance that enough of them will be right. (It covers the gap that Siri covers with jokes and Now covers by keeping quiet by overshooting - it's probably OK if 15 are not quite right if 10 are on target.)

How many playlists you need? 5,000? 10,000? With a few hundred editors that's not actually an impossible challenge, just a big one. And a cynic might suggest that really, you only need a few hundred for the vast majority of music listeners, if you do them right. 

Office, messaging and verbs

This is a still from the classic film 1960 'The Apartment'. Jack Lemmon plays CC Baxter, a clerk in a large insurance company in New York, and so here you see his office - drones laid out at desks almost as far at the eye can see. Each desk has a telephone, rolodex, typewriter and a large electro-mechanical calculating machine. 

In some of the other shots you can see the manufacturer - 'Friden'. 

In effect, every person on that floor is a cell in a spreadsheet. The floor is a worksheet and the building is an Excel file, with thousands of cells each containing a single person. CC Baxter is on the 19th floor, section W, desk 861. The links between cells are made up of a typewriter, carbon copies ('CC') and an internal mail system, and it takes days to refresh whenever someone on the top floor presses F9. (Shirley MacLaine plays an elevator attendant, so this is actually a romance between a button and a spreadsheet cell.)

When people talk about productivity - about PowerPoint and Excel and how Google Docs and the cloud will or won't kill them, or messaging and the cloud, or how you need a PC for 'real work' -  I'm reminded of CC Baxter and his Friden calculating machine. What killed those machines was not better, cheaper competitors but a completely different way to address the same underlying business need. Instead of hundreds of people recalculating insurance rates, the company bought a mainframe. The business need was being met, but the mechanism changed completely and the old tools disappeared. 

That is, the way forward for productivity is probably not to take software applications and document models that were conceived and built in a non-networked age and put them into the cloud, or to make carbon copies of them as web apps. This is no different to using your PC to do the same things you used your typewriter for. And of course that is exactly how a lot of people used their PCs - to start with. Just as today we make web app copies of software models conceived for the floppy disk, so the first PCs were often used to type up memos that were then printed out and sent though internal mail. It took time for email to replace internal mail and even longer for people to stop emailing Word files as attachments. Equally, we went from typing expense forms (with carbon copies) to entering them into a Word doc version of the form, to a dedicated Windows app that looked just like the form, to a web page that looked just like the form - and then, suddenly, someone worked out that maybe you should just take a photo of the receipt. It takes time, but sooner or later we stop replicating the old methods with the new tools and find new methods to fit the new tools.  

Hence, channeling Marshall McLuhan, new tools start out being made to fit the existing workflows, but over time the workflows change to fit the tools.

Some times and only sometimes, it's possible to see the new model springing into existence, like Athena fully formed. This 1979 preview of VisiCalc, the first spreadsheet software, captures just that. The writer has to explain, slowly, what the concept we call a spreadsheet is (the term itself is never used), and what it might be good for, because for most people there was no paper analogue. 

(Almost as good as this review, incidentally, is the fact that the same page has a discussion of competing videotext services.) 

Yet, very early, it became clear that though spreadsheets are indeed 'magic paper', they're often, indeed perhaps mostly, used for things that aren't actually calculations. As Joel Sposky pointed out here, Excel is a great way to make tables. It's also both the world's most widely deployed IDE and the world's most widely deployed desk-top publishing program. In Japan people used it instead of Word to write letters (for the layout control). I spoke on Twitter a while ago to a consultant who told me that half his jobs were telling people using Excel to switch to a database and the other half were telling people using a database to switch to Excel. The tools fit the workflow, and then the workflow fits the tools. 

I'd suggest that Microsoft Office is actually somewhat unusual in the field of productivity apps in how broad its use ended up being. Most normal people don't use Photoshop just to crop their holiday pictures or AutoCAD to sketch out where to put a new sofa, but Office encompasses both people who are using it for what it's designed and optimised for, and people using it because it's there, since it's so widely and cheaply distributed (perhaps 1bn copies are in use today in one way or another) and so broad and flexible. (One could debate this, of course - using Photoshop to generate web layouts is arguably as 'abusive' as using Excel as a database.) So there are people using Excel to build complex financial models and people using it to manage timetables, people using Word to create complex structured documents and people using it to write memos, and people using Powerpoint to communicate complex data to wide audiences and people using it for internal metrics reporting. 

For decades, this has prompted the idea that if most people don't need most of the features, a competitor, with fewer features but cheaper or with different routes to market, can peel away more and more of the users, leaving behind only the very core power users. This never really happened, and it seems to me that this may be the wrong way to think about the issue.

The power users need those features because it is their job to make complex financial models, sophisticated presentations or structured documents and so on, and those features help them do that. But the reason that all the other users can sometimes do without all the features is not because their job is to make simpler, more basic documents - rather, their job is not actually to make presentations or models at all. They're in sales or marketing or logistics or dozens of other roles and they're using these tools because that's what they have, but that's not their job - it's just a tool. So these people should not be using a simpler cheaper alternative to Powerpoint or Excel (or indeed Photoshop), or one with a few different features. Rather, the way they change tools is if you give them fundamentally different ways to achieve the underlying task.

Hence today, in a thousand companies, a thousand execs will pull data from internal systems into Excel, make charts, put the charts into PowerPoint, write some bullets and email the PowerPoint to a dozen other people. What kills that task is not better or cheaper (or worse and free) spreadsheet or presentation software, but a completely different way to address the same underlying need - a different mechanism. That Powerpoint file could be replaced by a web app for making slides that lets two people work on it at once. But it should be replaced by a SaaS dashboard with realtime data, alerts for unexpected changes and a chat channel or Slack integration (Slack is an a16z investment). PowerPoint gets killed by things that aren't presentations at all. The business need is met, but the mechanism changes. You can see some of these use cases in the suggestions in the 'File/New' menu. Each of these is a smartphone app or a web service - the unbundling of productivity apps. And none of these have to be 'spreadsheets'. 

What also changes, of course, are our assumptions about what hardware, precisely, you need. Do you need a large or small screen, do you need a keyboard, a mouse or just touch, and do you need a complex multi-window OS (Windows, Mac OS) or a simpler model based on full-screen use (Windows 8 et al, iOS, Android)? If you have to make an Excel file, paste charts into PowerPoint and write bullets or a memo then yes, keyboards, mice and windowing make things much easier. But if you have to flag a few key changes on a dashboard and tag them for review by three colleagues, you might not. The business task being achieved might be the same. Again - you need a keyboard to do x, but is x actually your job, or it it just the tool you use today to do your job?

What this points, to, I think, is that productivity breaks down into a set of verbs. In CC Baxter's office you see each of those verbs made into a physical object. Over time, those verbs get combined, broken apart, linked, created and removed as the tools change, the organization is changed by the tools and of course the underlying business itself changes. You don't actually send email or make a spreadsheet - you analyze, delegate, report, confer, decide, track and so on. Or, perhaps, 'what's going on, what are we doing and what should we be doing?' Each set of tools fixes that into a different pattern, but one should not look at that pattern and assume that that's the way things must be done - that that's what 'real work' looks like. 

A thread through all of this is communication, which prompted the slide below (created in PowerPoint, of course). Communication changes from a typed memo hand-carried to your desk in a manila internal mail envelope, to a carefully-laid-out presentation laboriously crafted in PowerPoint (maybe emailed, maybe presented on screen, maybe printed), to threads in Slack, a chat app with third-party service and data integrations. The real, underlying task is to communicate around the problem "how are sales of widgets going, why, and what should we do about it?", and that might not have changed at all, though you might have gone from a week to a day to a minute to get the answer.  

Distilling that further, there is information and the creation and analysis of it, and then there is communication - the connective tissue of the organisation. Right now, both of these generally mean the creation and the passing around or talking through of document files. But there's nothing eternal about that model. 

Ironically, Lotus Notes, one of the earliest corporate messaging programs, was intended to be much more than email, calendaring and so on - there was a vision of a unified development environment, database and messaging system - 'groupware'. It didn't quite work out like that, and actually using Lotus Notes as I had to 15 years ago was rather like using an email client built with Microsoft Access - theoretically possible but not a very good idea. OLE in the 1990s was another concept that didn't quite work, embedding pieces of one program's document inside another. But today, Facebook's platform on the desktop is pretty much Ray Ozzie's vision built all over again but for consumers instead of enterprise and for cat pictures instead of sales forecasts - a combination of messaging with embedded applications and many different data types and views for different tasks. Hence, one could propose one future model as 'Facebook for the enterprise', but with the platform, not the social, being the point of the analogy. 

This is also what productivity concept videos, like this one from Microsoft, tend to look like. Somehow scifi interfaces never have multiple applications from different vendors with patchy cross-compatability. Rather, every different data type flows seamlessly between nodes and screens. Everything is structured data (no more typing numbers from a PDF into Excel) and everything is a message, or can be.  This video is really showing Facebook, but for the enterprise and with lots of really thin sans-serifs. 

The challenge here is the trade off between breadth and flexibility on one hand and focus and single-purpose efficiency on the other. It's easy to make everything flow together in a single UI if you have a narrow domain, but much harder if you're trying to encompass lots of different tasks and types of data. Sometimes the right 'unified UI' is a dedicated app and sometimes it's Windows, or a web browser, aggregating lots of different apps with different UIs. But mostly, it's the email app itself that's the universal connector, linking documents, data and ideas. That is, 'Send' is the universal verb that ties the others together. 

There have been lots of experiments in this area besides Notes or Facebook - one could include Google's abortive Wave all the way back in 2010, Microsoft's Courier concept and now Microsoft's 'Gigjam' project. But a good place to see this now is to compare two productivity startups, Slack (previously mentioned) and Quip

Slack is a person-to-person messaging platform that's more than messaging - through third-party integrations, it acts as a common aggregation layer for every other system in an enterprise. All the messages, updates and notifications from the dozens of other systems in a company can be inserted into the relevant team or project chat channels, and then be scanned through or searched later on. Slack is effectively a networked file manager, but instead of folders full of Photoshop, Word or Excel files you have links to Google Docs, SAP or Salesforce, all surrounded by the relevant context and team conversation. Then, if Slack is messaging that adds software, Quip is the other way around - productivity software (tables, analysis, text, tasks and so on) with messaging as an integral part of the data model, rather than siloed into a separate app ('email this document'). What both are trying to do is blur the boundaries between 'messaging' and 'applications' such that the data all sits within the communication space - rather like that Microsoft concept video above. 

Much the same is happening in smartphone interfaces more generally, as I discussed here: the difference between messaging and apps is blurring. All of this recalls the old tech joke ('Zawinski's Law') that all software expands until it can read email, but now this works in both directions - messages become software, but software becomes messages. 

Search, discovery and marketing

 

"However vast any person’s basic reading may be, there still remain an enormous number of fundamental works that he has not read"

Italo Calvino

Erasmus is said to have been the last person in Europe to have read everything. He lived just at the point that printing took off, and so it was in his lifetime that it ceased actually to be possible to have read everything. Some time later there was presumably also a last person to have heard of every book, or every book worth reading - in the 18th century, perhaps. Then, sometime in the past few decades, it also became impossible to have heard every good song. It is still possible, just, to see every good film, but the so-called 'golden age of television' means it's harder and harder to watch all the great shows. 

This is not a problem of search or availability, but a problem of too much availability, and of course the internet magnifies this a thousand-fold. Paradoxically, the internet makes it possible to get anything you've ever heard of but also makes it definitively impossible to have heard of everything. It allows anyone to be heard, but how do people hear of you?

The first attempt online was Yahoo's directory - an editorial home-page and a directory of every single website, organized by category. This sounds like a joke now - "there was a website that listed every single website that there was". Yahoo actually worked pretty well when there were 20,000 websites, just as a book shop with 20,000 titles works pretty well. But the Yahoo directory peaked at 3.2m sites and at that that point it definitely didn't work - you can't possibly scroll past that many entries (though it lingered on in a half-life until Marissa Mayer shut it down this year). And in the meantime, Google invented PageRank, coming at the problem from an entirely different direction. Suddenly, search worked, and that seemed like the answer. 

Today, app stores look a lot like the Yahoo of 20 years ago, and they don't work for the same reasons - you can browse 20,000 apps but not a million. Hierarchical directories don't scale. And so while it's easy to make a list of things that Apple and Google should fix on their app stores, that misses the point - it's like making a list of ways that the Yahoo home page should have been better. You might have been right but the answer was still Google. (I suspect that the same applies, just a little, to the current moves towards app search and deep linking, incidentally. PageRank uses the signal of links between pages - the ability to link of itself is only half the picture.) This is one reason mobile messaging apps are so hot - because they might become acquisition and discovery channels. 

However, I think our preoccupation with the problems of apps and app stores and with the ways that they broke Google masks a deeper issue - that Google didn't really solve the problem either. Or rather, it moved the problem. Google is very good at giving you what you're looking for, but no good at all at telling you what you want to find, let alone things you didn't know you wanted. Like Amazon, it's essentially a passive product (which is why Now is so interesting). It relies on waiting for you to find out what you want somewhere else, in some other way, and then it gives it to you. No-one complains that ‘I put my book on Amazon and no-one can discover it there’, but that’s really no different to saying ‘I put my app in the app store and no-one can discover it there’, or indeed 'I made a web page and no-one came'. 

So Google moved the problem from 'I want to find this and can't' to 'yes, but what do I want to find?' That is, 'What should read next?', 'what lamp would I like?' or 'where should we go on holiday next weekend?' are not valid search queries. And that's just for problems you know you have - the most interesting businesses are often things you'd never given any thought to. What search would you do that would tell you about Lyft, Instacart, Pinterest, AirBnB or Evernote if you had no idea that they existed?  (This prompts the question, indeed, of what didn't or couldn't work before 2007 because search was the dominant model.)

One of the clearest places to see this problem of ‘too much’ is Yelp. I’ve been fascinated by how many companies are effectively trying to unbundle Yelp, despite that fact that (unlike Craigslist) it’s a modern technology company that does most of the things one would expect it to. But where people unbundling Craigslist generally try to peel off a category and deliver a modern experience, the people going after restaurant listings are often doing so with constraint. That is, instead of giving you every single restaurant that’s within 2 miles and that lots of people liked, they give you 10 restaurants. YPlan gives you one, and just one, thing to do tonight. People are attacking crowdsourced universal scale with constraint, curation and personal preference. 

Looking at these companies, it strikes me that actually, saying that ‘Yahoo’s directory didn’t scale’ misses the point. What we’re really seeing is a trade-off between two problems. You can have a list, solving discovery and recommendation, but once the domain gets big then your list is either unusably long or partial and incomplete (and perhaps uneconomic to maintain). Or you can have a searchable index of everything but you’re on your own working what’s good and finding things you didn't know to search for. Time Out is an interesting attempt to sit in the middle of that scale - enough coverage to be quasi-universal, and to promise something good nearby wherever you are, but also enough curation that you don’t just get 5,000 listings all with five stars.  ProductHunt is an attempt to use community to surface quality at scale, as is Pinterest (both are a16z investments). In contrast, Canopy uses hand-curated selections on Amazon. The question for all of these: do you filter crowdsourcing down enough to get quality, or scale up editorial to get coverage, or you give up on coverage and do a purely curated product? 

One answer is that the machine will scale to solve the problem - you aggregate the opinions of the many (Yelp), or data about your purchases (Amazon) or perhaps you yourself (Google Now). Amazon’s failure to do this reliably makes me hesitate (one suggestion it gave me is above - I collect these), but more deeply, there are some questions, again, that just do not make good text search queries. I can ask Amazon for Owen Hatherley’s new book on Communist architecture and it’ll find it, and Google will give me reviews. But if I ask them “what should I read next?” then you quickly fall into the uncanny valley between data mining and the 'real', HAL 9000 AI that we don't actually have yet. That is, a machine can learn that I like architecture and history books, but that’s not the same as knowing that I will buy Owen Hatherley’s book but never Jacqueline Yallop’s book on Victorian utopian model villages, and we're not quite there yet. 

You can also see this challenge right now in both books and fashion. Amazon, after 20 years of ruthless execution, still only has under a third of the entire print books market. Most people buy most of their books in physical retail, because book shops are not just relatively inefficient end-points to a physical logistics network, but also filters and recommendation platforms. They’re high-latency but also high-bandwidth. Fashion, meanwhile, is going online very fast, but not through Amazon. Rather, dozens of companies are circling around the right models or recommendation, curation and discovery.

So, perhaps, a split might be:

  1. There is giving you what you already know you want (Amazon, Google),
  2. There is working out what you want (Amazon and Google's aspiration),
  3. And then there is suggesting what you might want (Heywood Hill). 

Perhaps this is just the next step in retail. Amazon let people in one-bar towns buy products that could only previously be had in big cities, but it doesn't let you shop the way people can shop in big cities - once you understand that physical logistics is a very small part of what shopping means (this can be hard to spot if you've only ever lived in the suburbs of the South Bay). Buying and shopping are not the same thing. That's what the new generation of internet retailers are trying to do - to scale curation instead of catalogues. 

This is also, very obviously, what Apple News and Apple Music are trying to do. Each of them approaches a data set in which the default answer is a million options and a search box. Each asks the question: "how do we take a commodity in a database (web pages, music tracks) and layer curation and recommendation in ways that are more usable and friendly than just giving people a search box and pushing them out of the door?" It's not as though this is a solved problem anywhere else. RSS (which actually powers Apple news) failed as a consumer technology  and following a magazine or musician on Facebook means you won't see more than one in ten of their posts (and I don't choose my friends for their taste in music, or book reviews). And you can't Google for 'what do I read next? I hesitate to declare that this can't work. 

Another strand here, which really does take us back to the beginning, is that I always thought the most useful part of Flipboard (now the last man standing of all the iPad news aggregators that followed the iPad launch, and superficially similar to Apple News) is not any of the formatting or design but the built-in directory of sites. If you want to see 15 sites about basketball, or typography, or hats or makeup, where would you find such a list? Yahoo did that once, and so did link rolls and web rings or, in another way, del.icio.us. (It's funny how many people keep trying to rebuild Yahoo - it didn't work out that well for Yahoo itself). Tumblr today provides one route into this, if you invest the time in surfing the topics, as does Pinterest. But you don't get that from Facebook or Google. 

Of course, once you give up on a universal search and a universal store - that is, Google and Amazon - as the only answer, then you've moved the problem again. There's an old Soviet joke that a man walks into a shop and asks “You don’t have any fish, do you?" And the shopkeeper says “No, we’re a butcher - we don’t have any meat. The shop next door is a fishmonger - they don’t have any fish”. So: where do you want to be hard to find? Do you want to be one of a million listings in Google or the app stores, or do you want to be one of ten or 100 listings in a carefully curated selection - but where that selection is one of a million listings in Google or the app store? 

All of this takes us to marketing. I sometimes tease my Xoogler colleagues by suggesting that if PageRank Really Worked, SEM wouldn't exist - if the links were always the right answer then no-one would click on search advertising. (Larry Page is fond of asking challenging questions, but that might be one step too far.)  Until then, though some companies can make it entirely through organic search or Facebook virality, most cannot. (Indeed, very often the mere fact that you've made these channels work for acquisition means they stop working, since your link advantage gets arbitraged away by imitators or Facebook decides you're taking just a little too much of the newsfeed.) For the rest of us, that means marketing. In effect, by removing all other constraints, the internet makes advertising more important than ever. 

Presentation: Mobile is eating the world

This is an updated version of a presentation I first gave last autumn: the macro view of how mobile is changing the technology industry, the internet and the broader economy, here with slides and an accompanying talk track as I presented it to our limited partners earlier this month. 

US tech funding

I spent some time in the last few weeks with my colleagues Morgan Bender and Scott Kupor going through the state and history of US tech funding. It's pretty easy to point out that the current situation bears little resemblance to 1999 or 2000. This time it's different. But then, it's always different - what's going on now? 

Update: we also did a podcast talking though these issues. 

What are you afraid of?

As a company moves from insurgent to incumbent, and gets big and complex and involved in lots of different things, it tends to end up with lots of different objectives, tactics and strategies. At that point, trying to understand it from outside, it can be useful to think not about what it's trying to do but what it's afraid of. This company want to do lots of things, but what's the existential threat? What does it want not to happen? What scares it, late at night?

For Google, the fear is around reach. Google is a data company, and a machine learning company, and everything it does is about reach - reach to get data in so that it can understand everything better, and then reach so that it can serve that understanding out to the users. And so Android exists partly to enable the expansion of the mobile internet, but also, and more fundamentally, to ensure that no-one (meaning first Microsoft and later Apple) would be able to block Google from reaching those users, both to give them each results and to see what they are doing. Google is afraid of going blind. 

For Apple, I'd suggest the fear is that the developers leave. This is what happened in the 90s and it was a key part of the company's near-death experience (and arguably Apple only survived because the web made the lack of Mac apps matter less as a reason to buy a computer). Once developers start leaving you're in a vicious circle that's very hard to reverse (this is where Windows Phone is now). Today the iOS ecosystem is smaller than Android in absolute users and downloads, but has 7-800m live device, which is three times the size of the PC install base in 1995, and twice as much app store revenue per user as Google Play. More importantly, perhaps, the users are highly concentrated in key locations - Chase isn't going to abandon its iPhone app because there are 500m Android users in China.  So right now the ecosystem looks sustainable, but that could change. Developers can leave. That's Apple's existential fear.  

This is a useful lens to apply to the announcements at WWDC and IO. Google, this year in particular, always seems happier and more comfortable talking about the great stuff it can do with its own unmatched cloud intelligence - Now on Tap, for example. Losing that intelligence is what Google's afraid of. Apple is happiest talking about the new platforms and technologies that it wants developers to use - Apple Pay, iBeacons, Extensions or App Search. And losing that developer adoption is what Apple's afraid of. 

The (lack of) app store metrics

Apple and Google both give headline statistics of how well their respective app stores are doing, generally at their summer developer conferences. These are rounded numbers at scheduled events and they're not always comparable, but they do give us a sense of what's going on.  

Last summer, at their developer events, both Apple and Google gave numbers for the money they had paid to developers in their respective app stores: $5bn in the previous 12 months for Google Play and $10bn for the iOS App Store. Given Android has double the user base of iOS, this meant that the average iOS user was worth around 4x the average Android user in app store revenue. 

This year Apple gave the same number - $10bn (more precisely, it gave a cumulative figure of $30bn this WWDC versus $20bn last WWDC). The lack of growth may be partly due to rounding but still implies that people are spending less on average, since the user base is still growing.  Google gave no number at Google IO but it gave one earlier in the year of $7bn. It looks as through Play is growing faster than iOS and might overtake it this year (unless Apple is rounding down very aggressively - certainly the uneven shape of the graph in 2013 is due to rounding).  

Screen Shot 2015-06-13 at 11.47.31 PM.png

Since Google Android has close to double the number of users, this implies that the average user is spending perhaps half as much as the average iOS user - a change from 1/4 a year ago but still a big gap. 

Meanwhile, this was the first year since 2013 that we could compare downloads.

Google Play had 50bn app downloads in the last 12 months and iOS had 25bn, with Play appearing to be growing faster. Since, again, Play has more users, this implies roughly the same downloads per user on Android and iOS. 

Incidentally, these numbers show annualized consumer spending on apps of around $25bn, and 75bn apps downloaded in the last 12 months.