Stratechery Plus

Subscribe

Learn More
Member Forum

Menu

Stratechery Plus

Subscribe

Learn More
Member Forum

Latest Podcast

Stratechery Plus Update

  • Apple Music and Apple’s Focus

    In a spot-on introduction to a spot-on piece, Dr. Drang nails why, at least in the context of yesterday’s WWDC keynote, Apple Music was such a disaster:

    Endings are important, which is, unfortunately, why today’s WWDC keynote will be remembered as a flop.

    Two weeks ago I wrote about how Google I/O (members-only) was two separate keynotes: the beginning was about Google doing stuff primarily, at least as far as I could tell, because they were big; the second part was about Google furthering their original mission and doing what they do best — “Organize the World’s Information” — and it was incredibly compelling and exciting.

    WWDC went in the other direction. The first 90 minutes were excellent: very tight, with excellent clarity and momentum, well-rehearsed speakers delivering mostly iterative announcements with the occasional surprise. The final 60, on the other hand — the “one more thing” — were the exact opposite: unclear and dragging, with unprepared speakers delivering…well, I’m honestly not sure what most of them were saying. If there was a surprise the lack of a coherent message has to be top of the list.

    After all, as Apple CEO Tim Cook so frequently reminds us, “Apple loves music”. Most of us, including myself, have for years given Apple the benefit of the doubt when it comes to such statements, even as the company’s flagship music product, iTunes, has fallen into an increasingly unusable state, and even as the App Store has long since surpassed the iTunes Music Store as a source of not only revenue but more importantly differentiation. When Apple’s annual music event was switched to a grab-all product announcement in the face of plummeting iPod sales, itself reduced to an app, no one questioned Cook’s insistence that music is “in our DNA”.

    There is no question that music was in Steve Jobs’ DNA, and you could certainly argue that that specific aspect of Jobs’ DNA was a key component in transforming Apple from a barely solvent niche computer maker to one capable of creating an iPhone, an App Store, and everything else that makes Apple the most valuable company in the world. None of the latter happened without the iPod, without the iTunes Music Store, without that DNA.

    Moreover, to be perfectly clear, the ending of the annual music event, the diminution of the iPod, the eclipse of the iTunes Music Store — each of these was the natural order of things. An important part of what makes Apple so impressive as a company is its willingness to destroy its own businesses, an approach that requires always looking forward to the way the world will be, not backwards to the world as it once was (not to mention an organizational design and incentive structures that remove the very real human tendency to preserve and protect). Recall what Tim Cook said about Steve Jobs at the Apple founder’s memorial:

    “Among his last advice he had for me, and for all of you, was to never ask what he would do. ‘Just do what’s right,’” Cook said. Jobs wanted Apple to avoid the trap that Walt Disney Co. fell into after the death of its iconic founder, Cook said, where “everyone spent all their time thinking and talking about what Walt would do.”

    Here is my question about Apple Music: it certainly is what most of us think Jobs would do. But is it right?

    We know streaming is the future (and, for all intents and purposes, especially when you include YouTube — as you must — the present). Sure, iTunes music is DRM-free, and pirated content is as easy to access as ever, but on mobile devices where space, bandwidth, and time are limited, actual music files you have to move from device to device are as obsolete as physical CDs were once the iPod came along.

    The business and strategic implications of this shift are profound: the labels would have never made the original deal with Apple had the Cupertino company not been so clearly the lesser of two evils, which means Apple would never have come to dominate their revenue (and thus to gain monopsony power). But now that the piracy threat is inconsequential the labels are unencumbered in their ability to exact monopoly rents from their music collections, ensuring any streaming service operates at their pleasure, on their terms. Including Apple.

    Moreover, the fact that the iPod became just an app is symbolic of a far more profound shift: Apple has become a platform company in a way they never were even in the Mac’s heyday, and certainly unlike they were when the iPod was king. Perhaps the right thing to do is to enable — and extract rents from — services with not just the DNA but also the incentives and focus to deliver a compelling music experience.

    I can hear the pushback now: A streaming service is table stakes, it’s an essential part of the ecosystem, Apple needs to diversify into services (Android app!). I get it. It makes no strategic sense to not have a music service. On paper anyway. The great thing about Apple, though, what has made the company so unlike every other, is another lesson Cook learned from Jobs. From the Goldman Sachs Technology and Internet Conference earlier this year:

    We are the most focused company that I know of, or have read of, or have any knowledge of. We say no to good ideas every day. We say no to great ideas in order to keep the amount of things we focus on very small in number, so that we can put enormous energy behind the ones we do choose, so that we can deliver the best products in the world…

    That is not from just saying “Yes” to the right product which gets a lot of focus. It’s saying no to many products that are good ideas, but just not nearly as good as the other ones. And so I think that this is so ingrained in our company that this hubris that you talk about which happens to companies that are successful but then decide that their sole role in life is to get bigger, and they start adding this and that and this and that. I can tell you the management team of Apple would never let that happen. That’s not what we’re about.

    “That” sure sounds like Apple Music: there is this (streaming music) and that (curated lists) and this (BeatsOne radio) and that (Ping Connect) and no cogent thread to tie them together beyond the assumption that Apple must do a music service because that is what they do. That’s what big companies do.

    Moreover, it’s not like this was the only example of Apple failing to question its assumptions and subsequently losing focus. Yesterday’s demonstration of native Watch apps was very good, but in a very “Why does WatchKit even exist?” sort of way. When the Watch came out most reviews concluded that the Watch made a lot of sense in the long run, but few could get over just how poor the app experience was, with many concluding customers ought to wait.

    Imagine an alternate reality where the Watch had the exact same Watch face functionality (including complications), the exact same notifications and communications capabilities, the exact same performant Apple apps, the exact same unexpectedly strong battery life, but no apps beyond a promise they were “coming soon.” Surely reviewers would gripe, but with a “It’s already great, and it’s going to get better” sort of vibe. Yet Apple couldn’t bring themselves to say “no”.

    Indeed, what made the first half of yesterday’s keynote so compelling wasn’t just the speakers’ delivery; rather, the clarity of the delivery flowed from the fact Apple was doing what they do best: iterating on products that, once upon a time, were minimum viable products from a feature perspective delivered with a maximum focus on the user experience. John Gruber perfectly captured this quality of Apple in a 2010 piece called This is How Apple Rolls:

    Apple has released many new products over the last decade. Only a handful have been the start of a new platform. The rest were iterations. The designers and engineers at Apple aren’t magicians; they’re artisans. They achieve spectacular results one year at a time. Rather than expanding the scope of a new product, hoping to impress, they pare it back, leaving a solid foundation upon which to build. In 2001, you couldn’t look at Mac OS X or the original iPod and foresee what they’d become in 2010. But you can look at Snow Leopard and the iPod nanos of today and see what they once were. Apple got the fundamentals right.

    Maybe Apple Music will be great. Maybe the fundamentals are spot on. But the messaging certainly was not, and as I noted after the Watch unveiling, muddled messaging often stems from a muddled product, and muddled products come from a lack of focus. Maybe it’s time for Cook to spend less time talking about how “the management team of Apple would never let that happen” and make absolutely sure that a loss of focus is not, in fact, happening.

    Tomorrow’s Daily Update for Stratechery members will include a point-by-point analysis of the rest of yesterday’s keynote

  • The Funnel Framework

    For folks who weren’t there, it’s difficult to describe just how dominant Microsoft was at the end of the 20th century. It wasn’t just that the company and its products were everywhere, both literally and figuratively, but also the longevity of said domination: Microsoft and Windows were the most important company and product in the industry for 23 years.1

    Platforms Versus Ecosystems

    That longevity was built on the fact that Windows was not only a platform but also the cornerstone of a massive ecosystem. These two terms are often used interchangeably, but I think it’s important to draw a clear distinction:

    • A platform is something that can be built upon. In the case of Windows, the operating system had (has) an API that allowed 3rd-party programs to run on it. The primary benefit that this provided to Microsoft was a powerful two-sided network: developers built on Windows, which attracted users (primarily businesses) to the platform, which in turn drew still more developers. Over time this network effect resulted in a powerful lock-in: both developers and users were invested in the various programs that ran their businesses, which meant Microsoft could effectively charge rent on every computer sold in the world.

    • An ecosystem is a web of mutually beneficial relationships that enhances the value of all of the participants. This is a more under-appreciated aspect of Microsoft’s dominance: there were massive sectors of the industry built up specifically to support Windows, including value-added resellers, large consultancies, and internal IT departments. In fact, IDC has claimed that for every $1 Microsoft made in sales, partner companies made $8.70. Indeed, ecosystem lock-in is arguably even more powerful than platform lock-in: not only is there a sunk-cost aspect, but also a whole lot more money and people pushing to keep things exactly the way they are.

    Microsoft’s dominance had a big impact on everyone, and today most people in technology still talk in the language of platforms and ecosystems. However, in today’s era I’m not so sure either holds the sway it once did, particularly in the consumer market. Consider today’s dominant operating systems: Android and iOS. According to platform logic, Android should have long ago achieved total dominance thanks to the network effect that is supposed to accrue to whichever platform has the most users. Android’s advantage is arguably even stronger when it comes to its ecosystem: nearly every non-Apple OEM in the world and every carrier has good reason to push Android, and while it is by a comfortable margin the largest operating system and the closest thing we have to a modern Windows, it is nowhere near as dominant, powerful, or profitable.

    The fundamental ideas of platforms and ecosystems also don’t translate well to today’s other giants: Internet companies like Google, Facebook, and Tencent. The latter two clearly benefit from network effects, but it is a one-sided network: users connecting to each other. As for Google, by platform and ecosystem logic, they should be the most vulnerable of all: competing search engines are simply a URL away. Clearly there is another factor at work.

    Scarcity Versus Abundance

    It’s common to think of the modern computing era as having been launched by Windows and the IBM PC back in 1981 (or the Apple II in 1977): for the first time computers were for individuals, not just departments (minicomputers) or back-end offices (mainframes). The individual nature of computers has only intensified now that we carry them everywhere.

    The structure of the PC business, though, while an order of magnitude larger than previous eras, wasn’t that much different from minicomputers or mainframes: there was a defined market to sell to (primarily enterprise), and limited distribution channels. In other words, the opportunity for software companies was scarce, which is why the biggest providers — Lotus, Ashton-Tate, Borland, etc. — all relied on large sales teams. Smaller outfits relied on getting shelf-space at hobbyist stores or a friendly contact with a value-added reseller.

    This reality greatly magnified Microsoft’s platform and ecosystem advantages by tying them together: the ecosystem, heavily invested in Windows’ continued success, was the channel for the applications built on Microsoft’s platform; application providers, forced to invest heavily in sales, could hardly spare the money to build for another platform (i.e. the Macintosh) that didn’t have an ecosystem to sell their products anyways.

    What ultimately broke Microsoft’s stranglehold was the Internet: specifically, the fact that it made distribution free and freely available. While this didn’t necessarily matter to Microsoft’s core enterprise customers, there was an entirely new market developing: consumers. This, though, was a completely different kind of market: you couldn’t reach consumers through a sales team — there were simply too many of them, and they were only valuable in the aggregate. To put it another way, the shift to Internet distribution and consumer markets meant a shift from scarcity to abundance. Instead of constricted channels and known customers there was a new world of free distribution and an effectively infinite-sized market.

    This didn’t affect Microsoft immediately: all those Internet applications ran on a browser that ran on Windows. But while the PC-era Microsoft was king, the post-Internet Microsoft was the proverbial emperor with no clothes: its old platform and ecosystem lock-in were gone, even if the company didn’t know it. That reality would only manifest itself after both consumers and enterprise customers abandoned Windows Mobile as soon as something better came along.

    The Rise of “User-First”

    Referring to iOS (as well as Android) as “something better” may sound blasé, but the idea of a new offering winning on the substance of its user experience was a profound change: instead of convincing a centralized buyer via a limited distribution channel, winning products had to appeal directly to widely dispersed independent users. And, as I’ve discussed at length, in the case of Apple and the iPhone in particular, the implications of winning via the user experience were profound: no matter how much cheaper “good-enough” modular Android alternatives have become, the integrated iPhone retains a (growing) hold on the high end.

    What is interesting, though, is the impact on platforms and ecosystems of a user-first approach. iOS has maintained a platform lead, just as Windows did, but unlike Windows said lead is not based on owning the ecosystem (and thus distribution);2 rather, iOS owns the best customers, i.e. the customers who are most willing to pay. This is hardly a revelation, but I think there is a larger lesson to be drawn: success no longer depends on platforms or ecosystems; rather, platforms and ecosystems themselves depend on access to desirable customers. By extension, the companies who own that access — who own the funnel, to use a marketing term — are the ones who gain outsized influence and, in the long run, outsized profits.

    Consider Google; the basis of the company’s success is very simple: their search engine was, and is, superior. They didn’t have a sales team pushing their product to CIOs, or a revenue-sharing plan with resellers; they built an Internet application that was available to everyone and that everyone (eventually) used because it was the best. And, now that they own access to the most consumers looking to buy something online — now that they own the funnel — they have outsized influence and outsized profits.

    Facebook fits the same mold: the company built a superior product that connected an extraordinary number of people, all generating content that made the product even more attractive and able to capture even more attention. And, now that they own an outsized amount of consumer attention — now that they own the funnel — they have outsized influence and outsized profits.

    The Funnel Framework

    All three companies succeed with very different product focuses, but all share the ability to capture a specific type of customer and funnel them to someone who is willing to pay:

    Product Focus Captures Funnels To
    Apple Superior UX Willingness-to-pay Apple (and developers)
    Google Superior Search Motivated Users Direct Marketers
    Facebook Personally Important Content User Attention Direct and Brand Marketers

    Note the critical link is first delivering a superior user experience and then leveraging that into owning a funnel to a specific type of customer. To put it another way, in contrast to the PC era, distribution and access is no longer the means of control but the end result.

    If you look around you can see the funnel framework everywhere:

    • Taxis used to dominate because they owned distribution; Uber delivered a superior user experience, captured the most desirable customers, and leveraged that into dominance of the transportation market
    • Hotels used to dominate because they owned supply; Airbnb delivered superior value to end users and created a virtuous cycle that resulted in a two-sided network
    • Messaging apps deliver huge value to end users and monopolize attention; LINE and WeChat are leveraging that by pushing users to pay-to-play games in particular, while Snapchat is delivering advertising
    • Newspapers used to dominate because they had geographic dominance; BuzzFeed and other successful Internet publishers attract readers by first figuring out what they want and then giving it to them, and then sell either that skill (in the case of BuzzFeed and Vice) or the attention they gather (everyone else)
    • Broadcast channels used to dominate because they owned the airwaves; Netflix and HBO earn consumers’ attention by delivering superior programming and capture the value directly via subscriptions

    It’s hard to not be excited about the long-term implications of the funnel framework; I know that a lot of pre-Internet companies are struggling, and I sympathize with employees caught out by failing business models, but it’s worth noting that most of these doomed enterprises were based on something other than providing a superior product. On the Internet, on the other hand, being the best is a first order concern; as a consumer — and as a wholly independent entrepreneur — that makes me pretty happy.

    Discuss this Article in the Stratechery Forum (members-only)

    1. By my count, measuring from the launch of the IBM with DOS in 1981 to Google’s IPO in 2004; it’s 26 years if your end date is the launch of the iPhone; 19 if you start with Windows proper in 1985 ↩

    2. This is also why iOS’s lead is much, much smaller than Windows’ ever was ↩

  • 21 Inc. and the Future of Bitcoin

    It seems like no one has been talking about Bitcoin, at least for quite a while. That is what happens when the price of a seemingly magical currency plummets from a high of $1,242 in December, 2013 to a mere $238 as I write this, a drop of 81%. That means the total Bitcoin market cap is about $3.4 billion — far more than perhaps many expected even a few years ago, but a tiny number in the grand scheme of things (for the sake of comparison, Apple sold $3.9 billion worth of iPhones every week last holiday quarter).

    That’s why many were shocked back in March when 21 Inc., a Bitcoin-focused startup co-founded by Andreessen Horowitz board partner Balaji Srinivasan, announced that it had raised a shockingly large $116 million in funding for…well, no one knew exactly. Srinivasan’s co-founder Matthew Pauker said at the time, according to the Wall Street Journal:

    “Bitcoin is going to change the way that people and businesses and even machines interact with each other,” he says. “But for Bitcoin to realize that vision we need mass adoption. It can’t just be for Silicon Valley.”

    Indeed. Perhaps the single biggest obstacle facing Bitcoin is that most folks don’t really understand exactly what it is, or why it’s so interesting.

    Bitcoin in Three Bullet Points

    Here’s my best attempt to explain Bitcoin in three bullet points:

    • Using cryptographic keys, I can convey to you X amount of Bitcoin (fractions or multiples). This is cool because I’m transferring something digitally — no real-world analog (and associated clearinghouse, like a bank) required.
    • Our transaction is submitted to a global peer-to-peer network, which verifies the transaction and enters it into a public record (the blockchain). Because the blockchain is a ledger of every Bitcoin ever made, any one Bitcoin, or fraction thereof, can only ever be owned by one person. This is cool because it prevents me from double-spending — transferring something multiple times. In other words, Bitcoin makes digital goods rivalrous — uncopyable.
    • That global peer-to-peer network isn’t verifying transactions out of the goodness of its heart; rather, you may know of this verification process by a different name: mining. About every ten minutes one of the verifying computers is awarded with 25 Bitcoins for its effort (a current value of $5,950). This is cool because it means Bitcoin is a self-supporting system: transactions are verified with absolute certainty for free.

    Note that only the third bullet point really seems to have anything to do with money: what is most exciting about Bitcoin, at least from my perspective, are the first two points — that Bitcoin specifically, and the applications enabled by the Blockchain broadly, are digital (with all its attendant advantages, including worldwide instantaneous transferability, divisibility, tracking, etc.) yet scarce — in stark contrast to everything on the Internet that can be copied at no cost, much to the detriment of content producers in particular. This means Bitcoin could theoretically be used for all sorts of transactions, including messages, contracts, identity verification, etc., and, of course, monetary ones, where today’s digital applications just don’t work.

    Mining and Money

    It is monetary transactions that dominate discussion of Bitcoin, not only because that application is the easiest to understand, but also because money is the driving force behind mining. While mining first took place on personal PCs, the fact it is an embarrassingly parallel computational problem meant it was well-suited to graphics cards. Soon, though, as the price of Bitcoin increased, miners began investing in Application-Specific Integrated Circuits (ASIC), chips custom-built for Bitcoin and placing them in data-centers wherever electricity was cheap.

    While ASICs represent the fixed cost of Bitcoin mining, the expense of electricity is the marginal cost; indeed, Bitcoin is, for all intents and purposes, the digital representation of electricity. This presents a downside of Bitcoin — it is “free” in part because many of the externalities of its production are borne by society broadly in the form of pollution — but also highlights that even when it comes to mining money is only tangential. Were Bitcoin to go to $1 the vast majority of miners in the world would stop production tomorrow, yet Bitcoin the protocol would continue, automatically changing the difficulty of its algorithm to keep up the same steady pace of transaction verification on whatever processing was available to the network.

    Let me be explicit on this point: Bitcoin inherently has nothing to do with money, all of its representations as a digital currency to the contrary. Rather, it is a protocol that enables rivalrous (non-copyable) digital goods.

    21 and Mining Everywhere

    Last week 21 Inc. finally announced exactly what they were working on. Well, not “exactly,” but at least they said something. Here’s Srinivasan on Medium:1

    21 is now officially open for business — and business development. After much hard work, we’ve created an embeddable mining chip which we call the BitShare that comes in a variety of form factors. The 21 BitShare can be embedded into an internet-connected device as a standalone chip or integrated into an existing chipset as a block of IP to generate a continuous stream of digital currency for use in a wide variety of applications. You can request a dev kit by signing up on our website to get started.

    That’s right, with the BitShare chip, your Internet-of-Things toaster can mine Bitcoin, along with your cell-phone, and your refrigerator, and, well, your USB charger, which 21 Inc. is starting with.2 Srinivasan lists a whole hosts of potential benefits from this “free” revenue stream, including micropayments from a “free” revenue source,3 subsidized devices, even subsidized chips with 21 Inc. IP.

    Frankly, I think Srinivasan massively overstates his case with these examples — the economics for consumers simply don’t add up, for now:4

    • A single BitShare chip, given its minuscule processing capabilities, would need around 93 years to “win” a Bitcoin award
    • Thus, 21 Inc. will operate all of the BitShare chips it sells as a pool: Bitcoin income will be shared across all BitShare-installed devices, but only some — 21 Inc. will keep 75% of the revenue
    • Meanwhile, remember what Bitcoin is — digital electricity. And the consumers who own all of those BitShare-enabled devices will be paying 100% of the cost of that electricity

    Indeed, it makes far, far more sense for consumers to simply buy Bitcoin that has been harvested by the sort of industrial mining operations I talked about above on the open market.

    But that brings us full circle: what consumers, and why? Read that Pauker quote again:

    “Bitcoin is going to change the way that people and businesses and even machines interact with each other,” he says. “But for Bitcoin to realize that vision we need mass adoption. It can’t just be for Silicon Valley.”

    Indeed, all the benefits of Bitcoin are right now mostly theoretical: making it real requires a massive journey from here to there, to a world where everyone uses Bitcoin just as they use the Internet, and I think 21 Inc. is best understood as a massive bridge-building project from today’s speculative curiosity to tomorrow’s essential trust network.

    A (Digital) Wallet in Every Device

    When 21 Inc. first began in 2013, it was as a traditional mining operation: they brought on chip designers and built at least two generations of mining chips, earning 5,700 Bitcoins in 2013, and 69,000 Bitcoins in 2014. While it’s possible the company is pivoting to BitShare chips, I don’t think so; rather, 21 Inc.’s traditional mining pool gives it flexibility in making Srinivasan’s promises a reality. The company will actually have sufficient Bitcoin to guarantee an OEM that they will get a certain return for having placed a BitShare chip in the hardware they build, long before 21 Inc. achieves the sort of scale necessary to even fantasize about meaningful returns. 21 Inc. could also ensure consumers are compensated for the electricity they use, although I’m a bit skeptical about this — I think the play is primarily focused on getting manufacturers on board and hoping that future scale eventually ensures consumers get their fair share.

    That may be dubious ethically, but the payoff could be huge: the dynamics of Bitcoin and its application for a variety of trusted transfers shifts tremendously when everyone has multiple hardware-based digital wallets — whether they know it or not! For example, if your phone has a hardware-based digital wallet, applications for trusted messaging or contract signing or a whole host of applications no one has yet thought of become approachable not only for those capable of setting up a Bitcoin wallet but also those able to simply download an app. Identification depends not on usernames or passwords but on unique digital signatures. The possibilities are endless, and, I think, good — and they only rely on a single Satoshi (0.00000001 Bitcoin, the smallest amount that can be transferred).

    This is what I think Srinivasan means when he says “we are less concerned with bitcoin as a financial instrument and more interested in bitcoin as a protocol” and that “embedded mining will ultimately establish bitcoin as a fundamental system resource on par with CPU, bandwidth, hard drive space, and RAM.” It turns out lots of people are willing to spend electricity for essential functionality: nearly all of us run routers 24/7, whether we are using them or not, because instant access to the Internet is worth the electrical cost. A more extreme example are DVRs: our set-top boxes are the biggest power drains in our houses, costing $3 billion annually in electricity in the U.S. alone — nearly the entire worth of Bitcoin! — all in the service of making watching TV just a little bit more convenient. Why wouldn’t we eventually be ok with using just a bit of electricity not for profit but for mining the Satoshis we need to use an entirely new class of apps?

    Bitcoin the Network

    This, then, is the bridge I believe 21 Inc. is building, and why they specifically and Bitcoin broadly deserve more attention: Bitcoin at a conceptual level is not about money, and neither is 21 Inc., at least not directly.5 It’s about dispersed trust and digital scarcity, and the massive number of new applications that can be built on such concepts both similar yet wildly different from what is enabled by today’s Internet. And, while many of those applications are unimaginable — just as Facebook and Google were unimaginable at the birth of the Internet — what is very much certain is that step one is getting everyone, somehow, connected. As Coindesk reported, 21 Inc. is seeking to learn from the best:

    “The AOL CD of bitcoin,” the document called the strategy. “Give every user a free trial of bitcoin at near-zero marginal cost. A proven model to onboard millions.”

    Best keep an eye on your mailbox — and electricity meter.

    Discuss this Article on the Stratechery Forum (members-only)

    1. Has Medium replaced Tumblr for the default startup blog? ↩

    2. That is a slide from the 21 Inc. pitch deck; it’s quite old (note the reference to 21e6, 21 Inc.’s original name), and possibly not accurate ↩

    3. Journalism’s unicorn ↩

    4. Unless otherwise cited, 21 Inc. facts and figures come from this report from Coindesk ↩

    5. Indeed, an under-appreciated part of 21 Inc.’s approach is that all of these embedded chips will verify transactions without the incentive of monetary reward, allowing Bitcoin the protocol to survive even when there is no more Bitcoin to mine ↩

  • Apple Watch and Continuous Computing

    From day one of this blog I have insisted that I don’t do product reviews: I don’t buy every phone or, in this case, watch on the market, and I happily defer to publications that specialize in exactly that.1 That’s not to claim ignorance: I read voraciously, including reviews, talk to as many “normal” people as I can in as many places as I can, and think I have a sense for where various categories are at. And given that, I can’t quite shake the feeling that the Apple Watch is being serially underestimated. Nor, I think, is the long term threat to Apple’s position being fully appreciated.

    The Wrist is Interesting

    Back in 2013, at AllThingsD, Tim Cook said, “I see [wearables] as a very key branch of the tree,” and that “The wrist is interesting. The wrist is natural.” I fully agreed, and that, more than anything, was the basis for my optimism about the Watch over a year before its release. Over the long arc of technology, the fundamental characteristic of every new wave of devices that eclipses what came before is that they are smaller and (thus) more convenient:2

    • Minicomputers were smaller and more convenient than mainframes, meaning various departments in a corporation could have their own computers, instead of time-sharing
    • Desktop PCs were smaller and more convenient than minicomputers, meaning individuals could have their own computers
    • Laptops were smaller and more convenient than desktops, meaning individuals could have their own computers in more places
    • Phones were smaller and more convenient than laptops, meaning individuals could have their own computers with them nearly all of the time

    “Nearly all the time” is pretty amazing, and I get the arguments that the smartphone is the pinnacle of computer evolution: it’s portable, but readable, and multi-touch works and works well. But any interaction with your phone is still a cloistered one — you, your phone, and nothing else — which to my mind leaves at least one more evolutionary jump: continuous computing, no matter your context. Indeed, when it comes to ensuring a computer is always present and always accessible, the wrist is both natural and interesting.

    Still, success isn’t guaranteed: the ability of a watch, or in this case the Apple Watch, to enable a new area of continuous computing depends on three factors:

    • The physical design of the Watch
    • The interaction model for the Watch
    • The ability of the Watch to interact with its environment

    Apple and its competitors’ ability to deliver on each of these factors will determine whether the category ends up being a nice side business to phones, or the next step in the trend towards ever smaller and ever more convenient personal computers. And, for what it’s worth, after a few weeks with the Apple Watch, I’m increasingly bullish that it is the latter.

    Physical Design

    It may seem odd to declare that physical design is of equal import to the other factors that will determine a computer’s success, but a Watch is no ordinary computer: it’s one you wear. At Apple’s second Watch event Tim Cook said:

    Apple Watch is the most personal device we have ever created. It’s not just with you, it’s on you. And since what you wear is an expression of who you are, we’ve designed Apple Watch to appeal to a whole variety of people with different tastes and different preferences. But the one thing that is consistent is that we crafted each one of them with the care that you would expect from Apple and used incredibly beautiful materials.

    I added in How Apple Will Make the Wearable Market:

    There has been a bit of consternation about Apple’s focus on “fashion” and all that entails, but there is a very practical aspect to this focus: people need to be willing to actually put the wearable on their body. While “form may follow function” for tools, the priorities are the exact opposite when it comes to what we wear: function is irrelevant without a form we find appealing. In this case, design actually is how it looks.

    I’m not going to convince you that the Apple Watch is attractive or not, but to my eyes at least, it is significantly ahead of anything else on the market. And, frankly, I don’t think that surprises anyone. Apple’s industrial design is generally superior, but that superiority is maximized when a product or manufacturing technique is new: given that smartphones like the Xiaomi Mi Note, HTC One, or Samsung S6 are only just now approaching the iPhone, it’s reasonable to expect a substantial head start for the Watch.

    I get why most reviews only spent a sentence or at best a small section on the Watch’s appearance: it seems so shallow, and beauty is in the eye of the beholder. One’s skin, though, is as “shallow” as technology has ever gotten, at least compared to the depths of an office, a bag, or even a pocket. Each step towards the light increases the importance of aesthetics, and Apple’s advantage here can’t be overstated.

    On the other hand, “beauty being in the eye of the beholder” accentuates the downside of Apple’s integrated approach: what people wear is an expression of who they are, and while even the most idiosyncratic individual may be willing to have a phone that looks the same as everyone else’s, that may be a bridge too far when it comes to something on their body. This is why Apple launched with such a wide array of straps and case materials,3 and why the company has already opened the door to 3rd-party bands.

    For now, though, I think Apple Watch checks this box in a way other wearables to date have not: more people than not will be willing to wear it.

    The Interaction Model

    There is a hierarchy when it comes to actually interacting with the Watch, something Horace Dediu laid out in The Battle for the Wrist:

    The Apple Watch offers a hierarchy of surfaces onto which software can compete for attention:

    1. The Complication Layer
    2. The Notification Layer
    3. The Glances Layer
    4. The App Screen

    These surfaces are arranged in a hierarchy where the highest is the most accessible and the lowest is the least accessible.

    The ranking is not just about accessibility: it’s also the order in which the Apple Watch executes, from best to worst.

    • Complications4 are invaluable, and the delta between pulling out your phone to check your calendar, or the weather, versus looking at your wrist is massive5
    • The Taptic Engine is a revelation when it comes to notifications; in fact, I think Apple should have the sound turned off by default to accentuate the utility of an outwardly imperceptible tap of your wrist. The way a notification is displayed when you lift your wrist in response works well, and it’s easy enough to scroll through what you have missed6
    • Glances are where you put things you need to know that don’t merit a complication7 (or, in the case of 3rd-party apps, aren’t allowed — for now); they also serve as a more easily accessible app launcher
    • The App Screen is a place you only visit deliberatively, when you need a specific function. My favorites to date include Authy, for two-factor authentication, the New York Times app, with bite-sized stories, and Twitterific, for browsing mentions and direct messages. There’s no question, though, that hybrid apps are slow to load and often frustrating to use; one wonders if Apple wouldn’t have been better served keeping the doors shut until native apps are possible

    What is missing in Dediu’s hierarchy, though, is the most important feature of the Watch: Siri specifically, and the cloud broadly. Siri in particular impacts every other part of the hierarchy:

    • You can have, at most, 4.5 complications.8 To get any other information, you need to either use a Glance, an app, or, most efficiently, Siri:

      Siri-watch

      The virtual assistant is fantastic when it works, like in the first two examples. The third, though, pushed me to my phone — a terrible experience that far too many apps are mimicking — and to Bing at that, which was entirely unuseful (in stark contrast to Google):

      search-watch

    • Notifications are all well-and-good, but many, particularly messages, require a response, and Siri is the only option beyond a few canned responses, and a ghastly set of emoticons and mime hands. That’s not a bad thing! I already find speaking into my wrist to be a far more natural activity than speaking into my phone ever was, and truthfully, Siri on the Watch somehow seems vastly improved over Siri on the iPhone.

      More broadly, it’s clear that what the mouse was to the Mac and multi-touch was to the iPhone, Siri is to the Watch. The concern for Apple is that, unlike the others, the success or failure of Siri doesn’t come down to hardware or low-level software optimizations, which Apple excels at, and which ensures that Apple products have the best user interfaces. Rather, it depends on the cloud, and as much as Apple has improved, an examination of their core competencies and incentives argues that the company will never be as good as Google.9 That was acceptable on the phone, but is a much more problematic issue when the cloud is so central to the most important means of interacting with the Watch.

      There’s a second issue with notifications: in contrast to many reviewers, I’ve had no problem with the number of notifications being pushed to my wrist. In part this is because I long ago limited the number of notifications that I received, and I pruned the list still further when it came to the Watch. It’s fair to ask, though, how many customers will go to the effort? Indeed, here Google in particular may have a significant advantage with their efforts around Google Now: the very premise of the cloud service is to intelligently notify you about what you need to know when you need to know it, a proposition that is particularly compelling when it comes to something so intimate as literally tapping your wrist.

    • Both Siri and Google Now can launch apps, and again, both are essential to input in particular. I’m very frustrated at apps that don’t bother to include Siri input (LINE, I’m looking at you), but Siri itself is still frustrating, particularly because whenever it does screw up the transcription, there is no way to edit what you said. Probably the best solution is to simply continue to get better at transcription, but again, Google is ahead here and it impacts the experience far more deeply than it does on a phone

    Ultimately, the interaction model is the mirror image of physical design: Apple is playing catch-up, but it’s not out of the question that Siri becomes good enough, if it’s not already.

    Interacting With Your Environment

    This was the primary focus of the aforelinked article, How Apple Will Make the Wearables Market:

    It’s increasingly plausible to envision a future where…our physical environment are fundamentally transformed by software: locks that only unlock for me, payment systems that keep my money under my control, and in general an adaptation to my presence whether that be at home, at the concert hall, or at work.

    To fully interact with this sort of software-enabled environment, I will of course need some way to identify myself; for all the benefits of the human body, projecting a unique digital signature is not one of them. The smartphone clearly works, but it’s not perfect: the more you need it for interacting with your environment, the more noticeable is the small annoyance of retrieving it from your pocket or handbag.

    A wearable is different, particularly if it’s on your wrist: simply raising your arm is trivial. This makes it much more likely you will actually interact in a meaningful way with software-enabled objects around you, which makes even having said objects much more likely. To put it another way, I don’t think it’s an accident that the two hot new technologies are wearables and the Internet of Things; they are related such that each is made better by the other.

    As I noted in that article, this vision has a bit of a chicken-and-egg conundrum: software-enabled objects won’t be built for a Watch that isn’t widely available, which is why I suggested that Apple has a big advantage — the company has millions of loyal customers who will buy an Apple Watch simply because it’s made by Apple. This is further accentuated by point number one: Apple’s superior physical design makes it significantly more likely that a critical mass of Apple Watches will be sold.

    Still, Apple’s success with initiatives like Apple Pay and HomeKit is not assured: I’m bullish about the former, but it’s not a slam dunk, and the latter depends on Apple delivering on an API that the company itself doesn’t depend on, a situation that hasn’t always ended well.

    Looking Ahead

    It’s important to note that the factors I listed matter in order: you first have to build a wearable people are willing to wear, then deliver a usable interaction model, and finally catalyze a world of smart objects that interact with your wearable.

    Indeed, for now I think it likely that one of Apple’s oldest and most cherished skills — its ability to make beautiful, desirable objects — will make the Watch exactly what Tim Cook promised: another tentpole product that rivals the Mac, the iPod, the iPad, and even the iPhone. Framed as nothing more than A Watch that Does Stuff — and that you actually don’t mind wearing — Apple will rightly sell enough to kick-start a world that gets just a little bit smarter and little bit better when it knows who and where we are.

    Moreover, the Watch may even help Apple to rival Google when it comes to Siri and the cloud: the best way to improve a service like Siri is to have millions of customers using it constantly, and I for one have used Siri more in the last two weeks than I have the last two years. Multiply that by millions of Watch users and you have the ingredients for a rapidly improving service. Perhaps more importantly, the fact that Siri is critical to the Watch’s success in a way it isn’t to the iPhone’s may finally properly align Apple’s incentives around improving its cloud services.

    Ultimately, the Apple Watch has exceeded my quite high expectations. The complications and notifications fit into all the slivers of my life the iPhone has not, and the criticism I’ve levied at Siri has been primarily fueled by the appreciation of just how powerful it is to have a virtual assistant on my wrist instead of my pocket. As for apps, speed is the most easily solved issue in technology, thanks to Moore’s Law. I’m confident apps will be fully performant sooner rather than later.

    That said, I suspect there will be a bifurcation when it comes to the Watch’s relative importance vis-à-vis the smartphone between developed and developing countries: in the long run I do think that convenience trumps all, but there’s no denying a smartphone is already pretty darn convenient. To put it another way, if you can afford it there is a sufficient delta between Watch and iPhone functionality to make the former worth owning despite its dependence on the latter. I also think that when the Watch inevitably gains cellular functionality10 I will carry my iPhone far less than I do today.11 Indeed, just as the iPhone makes far more sense as a digital hub than the Mac, the Watch will one day be the best hub yet.

    Until, of course, physical devices disappear completely:

    her-movie-poster

    That is the ultimate Apple bear case.12

    Discuss this article in the Stratechery Forum (members-only)

    1. I do give my opinion on Twitter and Exponent  ↩

    2. They are also, in the long run, cheaper, in part because of scale. Indeed, the Apple Watch packs in an amazing amount of functionality for a mere $350 ↩

    3. I suspect had you told Tim Cook in 1998 that he would eventually oversee a product launch that included 38 models and 56 SKUs, he would have assumed he was having a nightmare — and certainly the number of SKU’s contributed to the Watch’s uneven launch ↩

    4. Complications are the “extra” features on a watch face beyond the time ↩

    5. Windows Phone tried to differentiate on this point, but the delta between a live tile and opening an app was far smaller than the delta between pulling out a phone and lifting up your wrist ↩

    6. And there’s an easy way to clear all notifications finally ↩

    7. For example, this is where I have my battery meter: the truth is I haven’t worried once about the battery running low ↩

    8. Modular and Simple (ironically) offer 4 complications plus the date ↩

    9. Nathan Taylor nailed this point on his excellent Praxtime blog two years ago ↩

    10. I agree it is several years off, but it’s clearly not impossible: several standalone smart watches exist today, not that I would want to wear any of them ↩

    11. Developers: heed this: don’t always assume the phone will be there! My biggest complaint about apps, outside of the terrible loading times, is that few outside of RSS readers (natch) let me read full articles. I read tens of thousands of words on this — I’ll do the same on the Watch ↩

    12. Yes, I know much of Her is not realistic; that doesn’t mean it’s not pointed in a broadly realistic direction ↩

  • Verizon-AOL, Facebook Instant Articles, and the Future of Digital Advertising

    At first glance, Tuesday’s two big stories — Verizon’s acquisition of AOL, and Facebook’s official announcement of Instant Articles — have little in common. A deeper look, though, suggests that both are a consequence of an ongoing transformation of digital advertising that will affect nearly every consumer-focused services company.

    Facebook’s Advertising Strength

    There’s no need for me to dwell on the Facebook news; I covered it at length in March in an article called The Facebook Reckoning. In that piece I noted a significant problem with Internet advertising: ad inventory is ever-increasing, which means the rates for an undifferentiated ad spot are ever-decreasing; the best way to combat that trend is through better ads, better placement, better targeting, and better measurement.

    Facebook is progressing on all these fronts: the company’s ad units are best-in-class, the targeting is frighteningly specific, and with its overhaul of Atlas the company is working to draw a line from Facebook ads to offline purchases. The vast majority of publishers, though, are in the opposite boat: mobile ads mostly stink, there’s no room for them on the screen (which leads to audience-antagonizing interstitials), and perhaps most problematically, it is much more difficult to target and track mobile users than it ever was on the desktop.

    Cookies and the Rise of Programmatic Advertising

    On the desktop, tracking is relatively easy: various ad networks will set a cookie in your browser that collects information about your travels around the Internet, and then use that information to serve advertisements based on your browsing history. That information was sufficiently useful that said ad networks were able to charge advertisers rates well above those charged by publishers who knew nothing about their audience, meaning the ad networks, with a few notable exceptions, could offer publishers more money for their ad inventory than the publishers could make on their own. This was a fundamental shift: instead of focusing on specific sites as a proxy to reach specific types of people — a model that was very much in-line with non-digital advertising like television — advertisers could instead focus on specific types of people directly.

    This process has only grown more advanced: today users are identified not only by their browsing history, but also offline data like loyalty cards and other purchase histories, creating incredibly precise profiles. Advertisers can reach people based on those profiles through automated exchanges that use auction pricing and serve up ads in real-time, allowing for the collection of nearly immediate feedback on the campaign, making rapid iteration possible for the advertiser.

    The Mobile Advertising Challenge

    Programmatic advertising is not perfect: while the advertiser can specify the sort of content they don’t want their advertisement to be associated with, it’s questionable how well this works in practice; there’s also a mismatch between impressions (when an ad is loaded) versus the number of times an ad is actually seen (Google says 56% of ad impressions are never seen). Mobile, though, has presented the biggest challenge: apps don’t share cookies.

    Advertising networks have searched for some sort of common denominator that would allow them to track users in different apps; originally most settled on using a phone’s Unique Device Identifier (UDID), which the Wall Street Journal exposed as a privacy disaster in 2010:

    “The great thing about mobile is you can’t clear a UDID like you can a cookie,” says Meghan O’Holleran of Traffic Marketplace, an Internet ad network that is expanding into mobile apps. “That’s how we track everything.”

    Here’s a useful rule of thumb: the more giddy a quote from an ad network employee, the bigger the privacy violation for end users. To that end, both Apple and Google soon banned the use of unique identifiers, instead offering ad IDs that could be reset by users — or even turned off completely.1 Advertisers have adapted, but the reality is that tracking users and measuring campaigns on mobile remains a lot more difficult than on the desktop, and that’s even before you consider multiple devices.

    Tracking Users Effectively

    This reality is at the core of Facebook’s value proposition to advertisers: target users, not devices. Sheryl Sandberg said on Facebook’s Q4 2014 earnings call in January:

    If you look at how digital ads are being measured, they are being measured based on a cookie based world that assumes that people have one device, largely a PC. And that is not real, consumers have a phone, they have a tablet, they have PC’s as well and [we have] the ability to understand that one person to serve an ad and measure all the way through [to purchase] correctly.

    Note the transition Sandberg is highlighting: the locus of tracking and measurement is shifting from devices to users. It echoes the shift in advertising from publishers to ad networks, which is to say from content to users. Targeting and serving ads to individual users is the goal, and, as Ms. O’Holleran excitedly declared in 2010, the holy grail is some sort of “super-cookie” that a user can’t turn off.

    Facebook has this: it’s your name, and most of us have gladly handed over not only that but also our birthday, our history, our address, all of our friends, what we like, and more. Of course we’re logged in on our PCs, so that all of those Facebook ‘Like’ buttons can track our movement around the web, and Facebook is aggressively pushing developer products — including relatively highly paying ad units — so that our apps can keep an eye on us as well.

    Google is dominant as well; I’ve long maintained that Google+ was more of a success than people appreciate: it unified IDs across Google’s properties and ensured most of us are always logged in. Google may not have the explicit user-provided data that Facebook does, but the company makes up for that with their extensive ad networks across both the web and inside of apps that not only display ads but also track said users.

    Still, while both Facebook and especially Google have 3rd-party ad offerings, both — especially Facebook — are more focused on monetizing on their own sites. A whole plethora of ad-tech companies have risen up to fill in the rest of the web’s inventory, but no dominant player has emerged: until, potentially, now.

    Why the Verizon-AOL Deal Makes Sense

    AOL was a part of that plethora: while most people remember the eponymous dial-up service and “You Got Mail,” or perhaps know that the company is also the 4th biggest digital media property in the United States with nearly 200 million unique monthly users, Verizon was almost certainly interested in a much newer part of AOL’s business: its ad network. Over the last several years AOL has invested heavily, both through M&A and R&D, into its programmatic advertising offering both for its own sites and, more importantly, 3rd-party ones; it has also developed a particular expertise in video. This has made the company competitive in its ability to allow advertisers to target users instead of content, but the lack of an “AOL ID” puts a cap on upside, at least relative to Google or Facebook.

    Verizon, meanwhile, knows a lot about its users:

    • By virtue of being a paid service, Verizon knows users’ names, addresses, and even social security numbers (gotta run those credit checks!)
    • Because they are a phone carrier, Verizon knows your location, something that is useful not just for serving ads but also for ascertaining whether or not they were effective (seeing a McDonald’s ad and visiting the Golden Arches soon after is a powerful signal)
    • Because they are the ISP for your mobile phone (and for many customers, their home as well), Verizon doesn’t need a cookie or device identifier: they can set a “super-cookie” on their servers to track everything you do on the Internet, and that’s exactly what they’ve done2

    This is why the deal makes so much sense: AOL provides the technology to target individuals instead of content, and Verizon the ability to track those individuals — at least the over 100 million customers they already have — at arguably a deeper level than anyone else in digital advertising (for non-Verizon customers, AOL’s ad platform is still useful, albeit not as targeted; rates would be commensurately lower). The talk of this mashup joining Facebook and Google to form a “Big 3” of digital advertising is not unrealistic.

    That said, this isn’t a grand slam either. While there is a lot to like about this deal, it’s doubtful Verizon would be doing it were they not facing long-term growth questions, and that’s not always the best motivation to enter an entirely new kind of business. Moreover, there is a very real culture question: an infrastructure company and an online media company couldn’t have more different approaches to business, based on a history of solving fundamentally different problems. This clash will only compound the difficulty in realizing the vision I just painted — a vision that Facebook and Google have still only partially realized despite having the best software engineers in the world. And, of course, there is the potential for consumer backlash: if my skin crawled while describing a dystopian future of perfectly tailored ads based on perfect information about my activities, I trust your reading it inspired a similar sensation.

    The Re-ordering of Advertising

    Describing what Verizon and AOL are shooting for is another way of depicting just how much trouble most publishers are in. Beyond the fact most publishers know precious little about their users, few are must-reads that attract users to their streams (as opposed to their links). Worse, even those with compelling streams don’t have scale even remotely comparable to what programmatic ad networks offer: for now the most prestigious sites are looking to monetize on direct sales to brands, but if programmatic offerings ever figure out how to guarantee premium placement that is actually viewed by the exact right customer those revenue streams could dry up.

    Subscriptions remain an attractive proposition, but as the Toronto Star recently discovered, not all publishers are the New York Times or Wall Street Journal, which means many publishers will have to take what they can get. And, frankly, it’s hard to see them getting a better deal than Facebook’s Instant Articles offering:

    • Publishers use their own publishing tools, meaning no interruption to their workflow; Facebook makes it look good (although publishers can augment the post) to the publishers’ specification — a New York Times article won’t look like a BuzzFeed article
    • Facebook will share analytics data from Google Analytics and Omniture, and ComScore will count Instant Article visits as visits to the publisher
    • Facebook is giving publishers the opportunity to sell their own ads and keep 100% of the revenue (although, given Facebook’s superior targeting capabilities, I wouldn’t be surprised if 70% of a Facebook sale ends up being worth more)

    Meanwhile, most publisher websites will be increasingly reliant on programmatic offerings that care only about the site’s ability to attract a specific type of customer who a particular advertiser has bid on, a sure route to a meager existence predicated on driving page views.

    It’s not just publishers who might struggle, though: two weeks ago I chronicled Twitter’s cloudy future given their relative lack of scale and under-developed ad offering, and while I stand by that, it’s kind of amazing that a few hundred million active users might not be enough; that certainly has to be a sobering thought not only for unicorns like Pinterest and Snapchat (although I’m bullish on both), but for any startup looking to monetize through advertising. It wouldn’t surprise me to see more and more companies taking yet another page from Asia and increasing their focus on e-commerce and online-to-offline transactions, but it might take a few failures to fully presage such a shift. Hopefully there will be sufficient publishers to document the change in approach.

    Discuss this Article on the Stratechery Forum (members-only)

    1. On the iPhone, Settings -> Privacy -> Advertising -> Limit Ad Tracking ↩

    2. You can now opt-out  ↩

  • The AWS IPO

    One of the technology industry’s biggest and most important IPOs occurred late last month, with a valuation of $25.6 billion dollars. That’s more than Google, which IPO’d at a valuation of $24.6 billion, and certainly a lot more than Amazon, which finished its first day on the public markets with a valuation of $438 million.1 Don’t feel too bad for the latter, though: the “IPO” I’m talking about was Amazon Web Services, and it just so happens to still be owned by the same e-commerce company that went public nearly 20 years ago.

    I’m obviously being facetious; there was no actual IPO for AWS, just an additional line item on Amazon’s financial reports finally breaking out the cloud computing service Amazon pioneered nine years ago. That line item, though, was almost certainly the primary factor in driving an overnight increase in Amazon’s market capitalization from $182 billion on April 23 to $207 billion on April 24.2 It’s not only that AWS is a strong offering in a growing market with impressive economics, it also may, in the end, be the key to realizing the potential of Amazon.com itself.

    Understanding Amazon, Part 1

    Much of the analysis of Amazon tends to fall in two diametrically opposed camps that are strangely united in their lack of rigor and in-depth appreciation of the economics driving Amazon’s business. On one hand are the ardent skeptics, who see Amazon’s lack of paper profits as prima facie evidence that the company is dramatically over-valued by the stock market; on the other are the true believers who point to Amazon’s ever-increasing revenue numbers as equally obvious evidence that the company is undervalued and primed for ever bigger and better things.

    A more nuanced approach considers the fact that Amazon is not a monolithic operation, but rather a collection of businesses sharing resources, including a channel (Amazon.com), logistics, and a common technological foundation. These businesses range from bookshops to video game stores to home furniture to clothes to shoes to consumer electronics to auto accessories…the list is quite extensive at this point! True, consumers experience all of these different businesses as a unified Amazon.com, but inside the company some of these businesses are mature and (theoretically) throwing off cash, while others are reliant on investment as they work to get off the ground.

    The Amazon business model as drawn by Jeff Bezos on a napkin
    The Amazon business model as drawn by Jeff Bezos on a napkin

    This view is a more sophisticated take on the aforementioned bull argument: it’s not that Amazon is not making money, it’s that the company is reinvesting every dollar it makes into growing new businesses; were the company to stop investing, it would start throwing off cash. This is exactly how Jeff Bezos has represented the company — except for the stop investing and throw off cash part, of course.

    Why Amazon Started with Books

    In The Everything Store, Brad Stone explains that while Jeff Bezos had always planned to build a site that sold, well, everything, he started with books for a very particular reason:

    Bezos concluded that a true everything store would be impractical — at least at the beginning. He made a list of twenty possible product categories, including computer software, office supplies, apparel, and music. The category that eventually jumped out at him as the best option was books. They were pure commodities; a copy of a book in one store was identical to the same book carried in another, so buyers always knew what they were getting. There were two primary distributors of books at that time, Ingram and Baker and Taylor, so a new retailer wouldn’t have to approach each of the thousands of book publishers individually. And, most important, there were three million books in print worldwide, far more than a Barnes & Noble or a Borders superstore could ever stock.

    If he couldn’t build a true everything store right away, he could capture its essence — unlimited selection — in at least one important product category. “With that huge diversity of products you could build a store online that simply could not exist in any other way,” Bezos said. “You could build a true superstore with exhaustive selection, and customers value selection.”

    There was one more critical factor, though, that Stone only mentioned in passing a few chapters later: despite the fact that books were a commodity, they were exceptionally high-priced. The list price of a new book contained a 50 percent markup for the retailer, which meant Bezos — who a few years later would coin the famous phrase, “Your margin is my opportunity” — could sell books at a significant discount while still making money on each transaction (and over time, Amazon would not only bypass the wholesalers but eventually become large enough to dictate terms to publishers).

    Moreover, the nature of Amazon’s business — customers paid for books with credit cards immediately, while Amazon paid book wholesaler invoices months after delivery — resulted in a negative cash conversion cycle that freed up much more cash for investment than would otherwise be warranted by Amazon’s margins, an effect that was greatly magnified by Amazon’s growth rate.

    This made books, over the long-run, a truly profitable business for Amazon, and the company repeated the trick for many of the exact same reasons in CDs, DVDs, and video games: commodity products with massive selections traditionally sold at a significant markup and paid for 90 days later allowed Amazon to have a superior selection and lower prices and make money to boot.3

    Understanding Amazon, Part 2

    The problem with the bull case for Amazon is the assumption that all of Amazon’s different businesses are essentially the same: selling books is like selling DVDs is like selling clothes is like selling TVs, etc. However, while that may be true from an infrastructure perspective (same channel, logistics network, and technology stack), it’s absolutely not the case from an economic one.

    Consider books versus TVs: there are an effectively infinite number of books, but a relatively small number of TVs, which means it’s more likely a competitor will have the same TV, leading to lower prices and reduced margins. TVs are also relatively infrequent purchases, and customers are likely to research the best price, leading again to lower prices and reduced margins. Moreover, while an individual book is a commodity, that same book is highly differentiated from another book; not so with TVs, where the most expensive TV still accomplishes the same thing as a cheaper one, which leaves little room for luxurious 50% retail markups. This again leads to lower prices and reduced margins. And, on the flip side, while a book is a book is a book, so you can buy it anywhere with confidence, many consumers consider a TV worth checking out in person, and it’s expensive to ship to boot.

    That said, TVs are, relative to books anyways, expensive, as are computers, furniture, car accessories, and all the other businesses that Amazon has been developing over the last decade. In other words, all of this stuff that Amazon is selling is perfect for increasing revenue, but not so great at producing profit (that said, the benefits of a negative cash conversion cycle very much apply to these items as well; they are not being added in vain).

    That leaves the old stand-bys — books, CDs, DVDs, and video games — to produce the actual profit that funds all of the new businesses Amazon wants to build, and they have done just that for 20 years now. The problem, though, is obvious: each of these categories is being replaced by digital distribution, and Amazon has only been successful in reaping the benefits of that shift in the case of books and the Kindle (although they’re competing in all four areas with Amazon Music, Amazon Prime Video, and the Amazon App Store). Moreover, the impact of digital distribution is showing up in the financial results; Amazon has long broken out the sales of ‘Media’ and ‘Electronics and General Merchandise’ in their financial reports, and in the last quarter ‘Media’ declined three percent even as ‘Electronics and General Merchandise’ increased 20%. This raises the long-run question for Amazon: if ‘Media’ is in secular decline, how will they fuel eternal investment into their business, much less the fabled returns that at least theoretically underpin their sky-high stock price?

    Enter Amazon Web Services

    The incredible potential of Amazon Web Services is as clear as its initial prospects in 2006 were, well, cloudy. AWS only came about after Amazon had experimented with more full-service offerings like powering the websites of Target or Toys-R-Us,4 and there were plenty of skeptics as to whether companies would entrust critical operations to a 3rd party. It soon became apparent, though, that both economics and simplicity were overwhelmingly in the public cloud’s favor, and Amazon was years ahead of everyone.

    Today, public clouds are the future for the vast majority of businesses; the economics of scale achieved by Amazon (and its closest competitors, Google and Microsoft) are so incredible that multi-billion dollar companies like Netflix view it as more efficient to pay Amazon than to build their own data centers. The calculus is even more stark when it comes to any sort of startup: it’s so much easier and cheaper to get started with AWS that the idea of buying your own server infrastructure — an expense that consumed the majority of venture capital in the dot-com bubble era — is preposterous. This is great from Amazon’s perspective: the company effectively has a stake in nearly every significant startup, and for free; if the company succeeds, Amazon will be paid, handsomely, and if they fail, well, Amazon covered their own costs of providing cloud services along the way.

    The big question about AWS, though, has been whether Amazon can keep their lead. Data centers are very expensive, and Amazon has a lot less cash and, more importantly, a lot less profit than Google or Microsoft. What happens if either competitor launches a price war: can Amazon afford to keep up?

    To be sure, there were reasons to suspect they could: for one, Amazon already has significantly more scale, which means their costs on a per-customer basis are lower than Microsoft or Google. And perhaps more importantly is the corporate culture that results from a “your-margins-are-my-opportunity” mindset: Amazon can stomach a few percentage points of margin on a core business far more comfortably than Microsoft or Google, both fat off of software and advertising margins respectively. Indeed, when Google slashed prices in the spring of 2014, Amazon immediately responded and proceeded to push prices down further still, just as they had ever since AWS’s inception (the price cuts in response to Google were the 42nd for the company). Still, the question remained: was this sustainable? Could Amazon afford to compete?

    This is why Amazon’s latest earnings were such a big deal: for the first time the company broke out AWS into its own line item, revealing not just its revenue (which could be teased out previously) but also its profitability. And, to many people’s surprise, and despite all the price cuts, AWS is very profitable: $265 million in profit on $1.57 billion in sales last quarter alone, for an impressive (for Amazon!) 17% net margin.

    A New Foundation for Amazon

    The profitability of AWS is a big deal in-and-of itself, particularly given the sentiment that cloud computing will ultimately be a commodity won by the companies with the deepest pockets. It turns out that all the reasons to believe in AWS were spot on: Amazon is clearly reaping the benefits of scale from being the largest player, and their determination to have both the most complete and cheapest offering echoes their prior strategies in e-commerce.

    Moreover, Amazon has cleverly found a way to approximate the negative cash conversion cycle trick: instead of paying billions to build data centers up-front, before they are ever used to earn revenue, Amazon is building its data centers with capital leases that effectively let the company pay for the data centers as they use them. True, this is a riskier strategy, and one that casts a pessimistic hue on Amazon’s admonition that investors look at free cash flow,5 but it’s also a strategy that presumes growth, an excellent assumption to make when it comes to AWS — provided the company’s investment can keep up. Contrary to my initial skepticism (members-only), I think the capital leases are a win-win.

    Perhaps the biggest implication of AWS, though, is its impact on Amazon.com. Last summer I lost my patience with the company, wondering when if ever Amazon would fully focus on seizing what looked to be a massive e-commerce opportunity, instead of dallying with devices and video. More importantly, would they do so before the ‘Media’ money train ran out of steam?

    Today that is a moot point: the sky is the limit for AWS, and if the service is profitable at its current scale, what expectations should we have for five years from now, or ten? More importantly, that profitability can over time replace the role of ‘Media’ in the Amazon engine: cash to build new e-commerce businesses, or to explore what is next (a la AWS), or both of the above. Or, in the fantasy of Amazon’s investors, to actually provide a return to shareholders.6

    Discuss this post in the Stratechery Forum (subscribers-only)

    1. With an ‘m’! ↩

    2. Amazon’s market cap has since decreased to $197 billion; like I said, I’m being facetious — this isn’t a real proxy for AWS’s value — but the spike was meaningful ↩

    3. This was the foundation of my bullish article on Amazon in 2013 called Amazon’s Dominant Strategy  ↩

    4. but ultimately, why help your competitors? ↩

    5. Amazon is now reporting “Free Cash Flow Less Finance Principal Lease Repayments and Capital Acquired Under Capital Leases” which is what their cash flow would be if they purchased data centers up-front; it’s significantly lower than the number the company trumpets publicly ↩

    6. So yes! As predicted, I am changing my mind about Amazon. Again. ↩

  • Twitter Needs New Leadership

    Yesterday Twitter released very disappointing quarterly results, with misses on both user and revenue growth. From the Wall Street Journal:

    Twitter Inc. delivered its weakest quarterly revenue growth as a publicly traded company, casting a shadow on its fledgling advertising business, which until now has been a consistent bright spot. Investors, who sent the company’s stock down nearly 20%, were hit with a double whammy Tuesday as Twitter’s first-quarter results were leaked on the social-media service itself an hour before their release was expected.

    The symbolism of the leak was appropriate: Twitter the service is so clearly indispensable, at least for some, but Twitter the company can’t seem to get it right.1 And so, after five years in charge, I believe it’s time for Twitter’s leadership, in particular CEO Dick Costolo, to make way for new leadership that has improved credibility with Wall Street, with developers, and within Twitter itself.

    Note: At the bottom of the article I have included a number of links to previous articles and Daily Updates where I developed many of the individual pieces of this argument; in other words, this is not a knee-jerk reaction to one bad earnings report, but rather a culmination of years of concern

    Twitter’s Fundamental Problem

    Twitter’s fundamental problem is that their active user growth is simply too small given their current size. Twitter yesterday reported the service had 302 million Monthly Active Users (MAUs), an increase of only 18% year-over-year and 5% quarter-over-quarter (and the company said the current quarter would be worse!). This is a fraction of Facebook, half of Facebook Messenger, fewer than Instagram and not that much bigger than SnapChat; presuming the latter service passes Twitter later this year, Twitter will be only the 5th most popular U.S.-based social networking service looking to monetize through advertising. This distinction — which excludes WhatsApp, at least for now — is a critical one, because the issue with advertisers is most don’t have the time or ability to work with multiple services; it’s likely most digital advertising spending (which I believe is set to expand greatly) will be consolidated onto the biggest networks (along with Google’s properties), with Facebook taking the lion’s share. Were that to happen, it’s easy to see Twitter as the odd network out.2

    Worrying Signs from Advertisers

    Twitter’s recent results suggest this shakeout may already be happening. CFO Anthony Noto was asked repeatedly on the earning’s conference call about Twitter’s success in retaining advertisers, and Noto repeatedly refused to answer the question, instead stating that “year-over-year growth [in total advertisers] was very similar to what it was in the fourth quarter.” The problem, though, is that later in the call Noto noted that Twitter’s advertising load factor was flat: if Twitter was adding advertisers, but not ads, that strongly suggests that the company is in fact losing advertisers just as quickly as it is adding them.

    Noto’s other justifications for Twitter’s revenue miss weren’t very compelling, either:

    • The first was that “some advertisers limited spending at higher levels of scale because the bids required to win incremental auctions were higher than they were willing to pay, which limited additional spending.” In other words, larger advertisers had in mind a price they were willing to pay, and preferred to reach fewer users rather than pay a higher price, suggesting that advertisers saw limited value in Twitter’s ads

    • The second was that Twitter “improved the quality of leads for direct response advertisers using our Website Cards by raising the bar on what constitutes an engagement or click”; Costolo later explained that some types of ads were moving from a cost-per-engagement model to a cost-per-action model, which was worth more to advertisers but more difficult to achieve, and that while the number of clicks decreased more than the price Twitter was able to charge increased, in the long-run this trend would reverse itself. The problem is that Twitter started this shift last summer when they rolled out app-install ads, which calls into question why the company failed to forecast the decline? Clearly the decrease was worse than they expected.

    These two justifications are particularly interesting in light of one more factoid that emerged in the earnings call: the aforementioned app install ads were one of the key ad types that underperformed. What makes app install ads especially interesting is that the companies buying these ads are able to be very sophisticated in their measurements of just how much these ads are worth to them. Thus, for example, said companies would almost certainly have a price limit on how much they were willing to pay for an ad placement (justification one), and we know their particular ad type is measured on a cost-per-action basis (justification two). In other words, it seems likely that not only does Twitter suffer from a potentially fatally small user base, but that their direct response ads may in fact be less effective than hoped, as evidenced by the pullback by app install advertisers.

    Indeed, Twitter’s direct sales team, which is focused on brand advertising from large advertisers, has long dominated Twitter’s revenue numbers, and that was again the case this past quarter (Noto noted that “direct sales was once again the largest contributor to year-over-year dollar growth”); this is a significant contrast to both Google and Facebook’s reliance on self-serve ad products. This isn’t necessarily a bad thing — I think branded advertising is set to explode past direct marketing — except that it is specifically brand advertisers who care the most about having an efficient way to reach a lot of people. In other words, Twitter’s limited user base is a particularly pressing problem in the ad category it excels at.

    Twitter’s Abandoned Users

    The trouble for Twitter is that awareness of the service has long outstripped its usability. And yet, despite the fact that Twitter has struggled with new user growth for years, almost nothing was done to improve the product or on-boarding experience until just the last few months, when the company finally rolled out a new logged-out page meant to entice people with Twitter’s content, as well as an instant timeline that helped people get started. Unfortunately, both efforts seem to be too little too late: Twitter admitted on the earnings call that neither improvement had increased retention.

    This isn’t a surprise: Business Insider reported last year that Twitter likely had 697 million abandoned accounts (and that number, presuming it was correct, has certainly grown). The problem is that those 697 million users, having already decided that Twitter isn’t a useful service for them, are much less likely to even experience things like the new logged-out page or instant timeline, even though Twitter @-handles and hashtags continue to be plastered all over TV and the web.

    The Problem with Twitter’s New Strategy

    This suggests, then, that Twitter needs a new strategy: the timeline has taken the company as far as it can, so how else can the company grow?

    To that end Twitter has taken to increasingly focusing on logged-out users, both on twitter.com, and also those who see Twitter’s content around the web. Ad Age reported from last fall’s analyst day:

    The company’s central focus was on the “largest daily audience” notion, driving the point that Twitter content flows and appears across the web and beyond. The company told investors to consider Twitter’s reach in four concentric circles: logged-in users, logged-out ones, eyeballs from Twitter posts’ syndication around the web; and, finally, the expanse of Fabric, its new mobile software platform. With this reach, Mr. Noto said the platform is en route to becoming “better than anybody else in digital advertising.”

    The problem is in the last line: “better” when it comes to digital advertising is only partially about reach: targeting, effectiveness, and measurement matter just as much. To that point, such an approach abandons Twitter’s biggest differentiation when it comes to its ads: the knowledge it has about logged-in users’ interests. Jay Yarow of Business Insider pressed Costolo on this point in February:

    Q: How do you think about [logged-out users]? When you look at a company like Yahoo, the display business is down. It seems like all the ad growth online is mobile, or social, or native advertising. What you’re essentially creating is a display ad unit for logged out users. When you have people who come in through search you don’t get the rich targeting like you do in your app. Are these logged-out users going to be less valuable than people in the app?

    Costolo: We still are going to be using the same native ad units, the Promoted Tweets. The content that’s also an ad is going to be the same unit. We like our ad unit a lot, we know that others in the market have replicated it, so were going to stick with that.

    As regards targeting, look we have this whole internal scrap about people who are logged in, it’s also the case that we’ll know a lot about people who are coming as logged-out users and why they’ve come there. A lot of those users come direct via searches, directly from a search engine, so we’ll know about them. We’ll know of course around events and specific topics, and in the moment what’s happening in the world, we’ll know the kinds of things that are happening and be able to direct our targeted ads into those experiences. So there’s a lot of the existing targeting capabilities that we’ll still be able to bring to these experiences.

    You’ll note that Costolo didn’t answer Yarow’s question, because the answer is obvious: logged-out users are significantly less valuable, and all those signals Costolo is counting on are no better than Yahoo’s. It is difficult to interpret this new focus on significantly less valuable users as anything other than giving up on acquiring new users, and I believe it is absolutely the wrong strategy.

    A Different Strategy

    Instead, Twitter should redouble its efforts to acquire new users even as it redefines what Twitter the company is all about. I wrote about this in What Twitter Might Have Been:

    What makes Twitter the company valuable is not Twitter the app or 140 characters or @names or anything else having to do with the product: rather, it’s the interest graph that is nearly priceless. More specifically, it is Twitter identities and the understanding that can be gleaned from how those identities are used and how they interact that matters. If one starts with that sort of understanding — that Twitter the company is about the graph, not the app — one would make very different decisions.

    Specifically, Twitter should dramatically increase the number of applications — and thus the number of potential reasons — a potential user might create and maintain an active user account. For example, Twitter could follow the Facebook strategy and build out a family of apps — one for messaging, another for news, others for specific events — and enhance the ways one could interact with Twitter content, whether that be through comments, private communities, etc. It’s ok that this is aping Facebook; what differentiates social networks is not their feature set but rather their organizing principle. Facebook is about people you know, and Twitter about those that share your interests. Everything else — including all the quixotic features that Twitter holds dear — are implementation details.3

    Alternatively, Twitter could empower third-party developers to build these sorts of applications that feed back information into the Twitter interest graph. An application like Nuzzel, for example, which uses your Twitter graph to create a news app, has much more of a one-way relationship with the social network: Nuzzel is getting all the benefit, and not sending much information back to Twitter. Twitter would be better off retooling their API and developer agreements to ensure they are learning from every application they interact with, and in return sharing their graph along with advertising in the form of their MoPub or Namo Media-derived offerings. The advantage of this approach is that the imagination and ingenuity of a massive developer ecosystem will always be far faster and more innovative than anything any one company can do on its own — just ask Apple.

    As an aside, something that has hurt Twitter on the public markets has been the expectation/hope that the social network would follow Facebook’s path with regards to user numbers and monetization. Clearly the company as presently constructed isn’t anything close to that; however, the open approach that I’m advocating could in fact become something exponentially larger. Last week I wrote about Facebook’s AOL-like dominance and concluded, “What might be the broadband to Facebook’s dial-up?” The answer, I think, is this open Twitter: an identity system for the rest of the web that connects people and apps according to interests, not just superficial relationships, and monetizes accordingly.

    Obstacles

    Unfortunately, it’s difficult to see Twitter under its current leadership getting even close to this ideal:

    • A Facebook-type strategy would require a level of execution and speed that is pretty much the exact opposite of what Twitter has demonstrated throughout its existence. Facebook seems to be updating or testing new designs for its various products on a nearly weekly basis, while it feels like Twitter’s product hasn’t really evolved in years (in stark contrast, I might add, to the composition of its executive suite).

      Given that Twitter seems afraid to make even obvious changes to its core product for fear of messing up what they don’t seem to understand, I’m not particularly confident the current leadership team has either the vision or the internal credibility to lead such an effort4

    • The open strategy I endorse has an additional x-factor outside of Twitter’s control: the cooperation of 3rd-party developers. And, given that Costolo was the CEO when Twitter burned its bridges with developers in 2012 (and, two weeks ago, with its data providers), the company under its current leadership has no credibility with the folks who might make such an effort a success

    • Most importantly, now that Twitter is a (struggling) public company, it is exceptionally difficult to see Wall Street having the patience for the amount of time such a strategy realignment would take. After all, it took the company nearly nine months to come up with its current strategy of being Yahoo-lite; how much longer to map out much less execute the sort of ambitious plan proposed here? Unfortunately, any credibility the company may have had with Wall Street is, after yesterday’s huge miss, surely gone

    New Leadership is Needed

    To summarize:

    • Twitter has a user growth problem that, as of this quarter, is showing significant signs of manifesting into an advertising growth problem
    • Twitter’s current strategy of expanding to serve an audience they don’t know with undifferentiated ads is barely sustainable and a considerable waste of Twitter’s potential as the owner of the interest graph
    • There are different strategies available, but the current leadership team doesn’t have the credibility internally, with developers, or with Wall Street to consider or execute them

    That is why I believe it is time for a change. Twitter needs a better strategy, and more importantly, it needs fewer obstacles in its path to executing that strategy. There’s no question Costolo in particular has done excellent work at Twitter, taking over a dysfunctional company with a brilliant product and building it up into an actual business, and I wouldn’t begrudge him a particularly large golden parachute. But I firmly believe the time to act is short.

    In fact, it may already be too late. It seems ridiculous to, as I did earlier, put Twitter in the same conversation as Facebook, given that the former owns 70% of social referrals and Twitter a pitiful 3%. Moreover, Facebook’s developer offerings are considerably more advanced, and its monetization opportunities significantly greater. Instead, as I hinted at in my previous article, there is a lot about a Google-Twitter merger that makes sense: the former pays the bills, and the latter provides a road into the brand advertising future. That, though, would likely require an even greater run-down in the stock.

    I don’t want that for Twitter. I want one last shot for my favorite and most essential social network to truly take over the world, not just the headlines; I just want someone else taking that shot.

    Background

    I am not the first person to discuss this issue. Last November, the Wall Street Journal quoted Walter Price, an Allianz Global senior portfolio manager as saying “People are losing confidence in [Costolo]”, and in December Robert Peck, an analyst at Sun Trust told CNBC “We think there’s a good chance he’s not there in a year”, which may have caused a mini-rally in the stock.

    More importantly, I have been building up the substance of my argument for over two years here on Stratechery; the following are a selection of articles and Daily Updates that contain many of my concerns (please note the Daily Updates are members-only)

    • September 16, 2013 — There are Two Twitters; Only One is Worth Investing In — This article written when Twitter filed for its IPO, and predicted that Twitter would monetize relatively well but that it would struggle with user growth
    • April 30, 2014 — Daily Update: Twitter’s Earnings — This Daily Update discussed, in the context of Twitter’s earnings, why I was increasingly concerned that the company’s new user problem was intractable given the fact that many people had already tried the service
    • May 1, 2014 — Twitter’s Marketing Problem — This article posited that Twitter’s new user problem was exacerbated by the fact the company had never had to figure out product-market fit; the core product was so good it appealed to a very large number of users without any effort on Twitter’s part, but now the company did not know how to grow marginal users
    • July 30, 2014 — Daily Update: The Twitter Enigma — This Daily Update expressed my concern that Costolo did not seem to have a clear vision for the company
    • October 23, 2014 — Daily Update: Twitter Fabric — This Daily Update discussed the fact Twitter seemed to be giving up on Twitter the service and turning its attention to developer-based services and monetization options, and the difficulties they would have with such a strategy
    • October 28, 2014 — Daily Update: The Twitter Mismatch — This Daily Update discussed the disappointing number of advertisers on Twitter, both in terms of direct-sale brand advertisers as well as self-service direct marketing focused advertisers, and posited that Twitter was in danger of losing out on advertiser attention completely because it was simply too small
    • October 31, 2014 — Daily Update: Twitter Changes, Again — This Daily Update discussed yet-another Twitter executive reshuffle, this time around product
    • November 13, 2014 — Daily Update: Twitter’s Analyst Day — This Daily Update discussed Twitter’s analyst day announcements and my concerns that the company was increasingly focused on pandering to Wall Street at the expense of making need product improvements
    • February 6, 2015 — Daily Update: Twitter’s Results, Algorithm, and Value — This Daily Update discussed Twitter’s new plan to focus on non-logged in users and the financial ramifications of doing so
    • March 13, 2015 — Daily Update: Nuzzel and the Unbundling of Twitter — This Daily Update explored the possibility that the 140-character timeline was a minimum viable product that is being unbundled in favor of speciality services, and that Twitter should embrace that
    • April 15, 2015 — Twitter and What Might Have Been — This article, just two weeks ago, explained how Twitter was cutting off access to its interest graph due to its increasingly desperate need to monetize, and explored how much brighter the company’s future may have been had the company been more open with developers in particular

    1. To be fair, the specific mistake in this instance was Nasdaq’s reponsibility. But, as we’ll discuss, there seem to be excuses for everything that is, at the end of the day, Twitter’s responsibility ↩

    2. Moreover, Twitter’s MAU numbers are not necessarily what they seem, particularly from an advertising perspective. The company disclosed last year that 14% of its MAUs never visit Twitter’s website or apps; rather, these users have connected 3rd-party applications and websites that don’t display Twitter ads; for example, you can log into Twitter from Instagram in order to post a link to a photo. Moreover, that number doubled over the previous year, a stark contrast to Facebook’s <5% figure. Worse, a sizable portion of those users may be lapsed completely; according to Twitter approximately 8.5% of MAUs are due to applications that automatically ping Twitter without any user involvement.

      One example of such an application was iOS 7 Safari: in Twitter’s 2014 4Q results the company blamed its slowing user growth numbers on iOS 8, which ended Safari’s practice of pinging Twitter for its Shared Links whether or not the logged-in user ever read that section. The proper interpretation, though, was not that Twitter’s user growth numbers were accidentally and temporarily slowed, but rather that the company had been over-counting MAUs for at least a year. Add in the fact that, according to Twitter, up to 5% of MAUs are likely spam accounts, and it’s fair to wonder just how many people are actually being served ads from Twitter’s advertisers ↩

    3. Arguably Twitter is on this path with Vine and Periscope, but progress is painfully slow; the company needs to be moving 10x faster ↩

    4. And, I might add, I remain disappointed and disillusioned by Twitter’s hire of a head or product who, at the time, only had 72 tweets to his name; to Costolo’s credit he fixed this mistake quickly ↩

  • Facebook and the Feed

    In a week where much of the Internet was all atwitter about Mobilegeddon, Google’s pre-announced algorithm change that will favor mobile-friendly sites in mobile search results, a potentially far more impactful announcement was much more of a surprise: Facebook is tweaking the News Feed algorithm.

    This is a big deal for publishers in particular: according to Shareaholic, social referrals passed search referrals last summer and are now up to 31% of site traffic as of December, and Facebook is responsible for an incredible 79% of those social referrals.1 What is perhaps more interesting though, is what these changes mean for Facebook itself.

    Understanding Facebook’s Dominance

    It is increasingly clear that it is Facebook — not iOS or Android — that is the most important mobile platform. Facebook’s family of apps account for 24% of time spent on mobile, and the main Facebook app is responsible for 75% of that. Mobile apps thrive on “found time” — moments in line, or on the bus, or even on the couch where people simply want to look at something interesting — and Facebook consistently delivers for an increasing number of users all over the world. More impressively, Facebook isn’t just increasing its user base: its existing users continue to deepen their engagement with the app over time.

    In this respect Facebook really is the new AOL: in the 1990s the service provided a far easier and more accessible way to get online, and by 1997 nearly 50% of all Americans got online via the service. And, just as publishers and anyone else eager for people’s attention flocked to get their content on AOL, the same is the case for Facebook: as I discussed last month publishers like the New York Times and BuzzFeed are reportedly on the verge of placing their content directly on Facebook. As long as Facebook is the easiest way to access content on mobile, publishers have little choice but to go where their readers are.

    The problem with this comparison — at least from Facebook’s perspective — is that 1997 was AOL’s peak. North America’s first broadband service had launched the year previously, and as more and more customers got online through their phone or cable providers, AOL’s moat — dial-up access — evaporated away. And in the end, AOL’s content, compelling though it may have been, simply couldn’t compete with the breadth of the entire Internet.2

    Facebook’s Moat

    This is where, to my mind, the AOL comparison falls apart. AOL provided an essential utility that was far easier-to-use than the alternatives, but that utility was obsoleted by broadband. Facebook, on the other hand, is built on the social graph: its users’ relationships. And given that the very nature of humanity is to connect and communicate with other humans, the need that Facebook has traditionally met will be with us forever. The only danger is that another service somehow takes Facebook’s place as the Rolodex of the world.

    I simply don’t see that happening. At this point, my most extreme Facebook bear case is that the service is the equivalent of an email address: something nearly everyone has because you can’t function without it, even if it’s not their preferred means of communication.

    However, I say “extreme” for a reason: for many Facebook is much more than that. Here in Asia, for example, Facebook is LinkedIn: it is standard for an introductory business meeting to conclude with Facebook friend invites. Facebook is also the homepage for the vast majority of businesses: a page is much more discoverable and much easier to maintain, and most don’t even bother with a website.3 A significant amount of e-commerce happens on Facebook as well, and most celebrities and well-known bloggers post primarily or exclusively on Facebook (although Instagram is increasingly important as well). True, messaging services like WhatsApp and LINE increasingly handle one-on-one or private group conversations, but in country after country that I visit or research the social paradigm is Facebook + X social network; the X changes (and is often a Facebook property), but Facebook is the constant.4

    Still, the argument I just made is about the ongoing usefulness of having a Facebook account. What about engagement?

    What Drives Facebook Engagement?

    This is the central question facing Facebook, and a fascinating way to think about yesterday’s News Feed changes. The Verge has a good summary:

    Facebook has announced it’s twisting the knobs that control what content you see in your News Feed to favor more content from your close friends. In a post titled “News Feed FYI: Balancing Content from Friends and Pages,” Facebook said it’s making three changes. The first is that it won’t let people reach the “end” of a News Feed as easily, because it will be willing to show more content from the same publisher than it was before. Previously, you wouldn’t be likely to get two posts from the same Page. But the other two changes are more ominous for publishers, but potentially great for users who actually want to see content from their friends: “content posted by the friends you care about” will be “higher up in the News Feed.” Also, if a friend interacts with a post from a brand or publisher page, it will be less likely to show up in your News Feed.

    This is hardly the first major change to Facebook’s News Feed algorithm:5 consider, for example, this major update from December 2013:

    We’ve noticed that people enjoy seeing articles on Facebook, and so we’re now paying closer attention to what makes for high quality content, and how often articles are clicked on from News Feed on mobile. What this means is that you may start to notice links to articles a little more often (particularly on mobile)…

    While trying to show more articles people want to read, we also don’t want people to miss the conversations among their friends. So we’re updating bumping to highlight stories with new comments…With this update stories will occasionally resurface that have new comments from friends.

    This December 2013 change had a huge effect, crushing viral sites like Upworthy while bumping up traffic for sites like BuzzFeed, Business Insider, and The Guardian. What is perhaps more interesting, though, is that yesterday’s changes seem to run in the opposite direction: prioritizing friends, when 2013’s update prioritized news; and deprioritizing friends’ comments and likes, when 2013 bumped them up.

    Here’s the thing: I’m quite sure the 2013 changes weren’t arbitrary. Facebook is a very data-driven company, and all available evidence suggests that the changes had their intended effect: as I noted above Facebook is not only increasing users but also deepening the engagement of their existing users quarter after quarter. That’s why I’m curious just how important data was in yesterday’s changes compared to the personal preference of Facebook founder and CEO Mark Zuckerberg and his belief about what makes Facebook valuable.

    Facebook’s Vision and Potential

    Zuckerberg is quite clear about what drives him; he wrote in Facebook’s S-1:

    Facebook was not originally created to be a company. It was built to accomplish a social mission – to make the world more open and connected.

    I am starting to wonder if these two ideas — company versus mission — might not be more in tension now than they have ever been in the past. I’m increasingly convinced that Facebook has an absolutely massive business opportunity on its hands: to capture, almost completely, the imminent wave of advertising dollars deserting TV for digital. To do so, though, will require an embrace of Facebook’s status as the “homepage of the Internet” (on mobile in particular), and an abdication of sorts of the social interactions that built the company.

    For several years now the percentage of people’s attention devoted to digital has far outstripped the percentage of advertising devoted to the medium. Television and (especially) print, on the other hand, keeps a far greater share of advertising than it seems they deserve:

    via BuzzFeed
    via BuzzFeed

    A big reason for television’s dominance in particular is that it is simply the easiest way to reach a large audience. Twitter may offer interest-based targeting, for example, but your typical brand manager simply doesn’t have the time or expertise to optimize every dollar across a broad array of platforms. Efficiency is just as much a feature of advertising as is targeting capability, conversion tracking, or price.

    Still, as I wrote in Old-Fashioned Snapchat, advertisers care above all about attention, and there’s no question that television is losing it both to alternative video services like Netflix but also to digital services, especially Facebook. Last year the average user gave Facebook over 40 minutes of attention a day (and another 20 minutes to Instagram, a property capable of supporting a very Facebook-like advertising unit), and that number continues to grow. Given Facebook’s excellent targeting capabilities and aspirations for Atlas’s ability to provide conversion tracking (members-only), it’s not inconceivable that, at some point in the relatively near future, it is Facebook that is the default advertising medium, commanding dollars that exceed its already dominant share of attention. Still, this outcome depends on Facebook driving ever-more engagement, and I’m not convinced that more “content posted by the friends [I] care about” is the best path to success.

    What Matters to Users?

    Everyone loves to mock Paul Krugman’s 1998 contention about the limited economic impact of the Internet:

    The growth of the Internet will slow drastically, as the flaw in “Metcalfe’s law”–which states that the number of potential connections in a network is proportional to the square of the number of participants–becomes apparent: most people have nothing to say to each other!

    It’s worth considering, though, just how much users value what their friends have to say versus what professional media organizations produce. Again, as I noted above, Facebook made the 2013 decision to increase the value of newsworthy links for a reason, and in the time since, BuzzFeed in particular has proven that there is a consistent and repeatable way to not only reach a large number of people but to compel them to share content as well. Was Krugman wrong because he didn’t appreciate the relative worth people put on what folks in their network wanted to say, or because he didn’t appreciate that people in their network may not have much to say but a wealth of information to share?

    I suspect that Zuckerberg for one subscribes to the first idea: that people find what others say inherently valuable, and that it is the access to that information that makes Facebook indispensable. Conveniently, this fits with his mission for the company. For my part, though, I’m not so sure. It’s just as possible that Facebook is compelling for the content it surfaces, regardless of who surfaces it. And, if the latter is the case, then Facebook’s engagement moat is less its network effects than it is that for almost a billion users Facebook is their most essential digital habit: their door to the Internet.

    Facebook’s Choice

    I’ve written previously about publishers and the smiling curve, the idea that value is increasingly flowing to aggregators on the right and differentiated content creators on the left; publishers are being left in the cold.

    The Publishing Smiling Curve

    Facebook is slowly but surely building a bridge between the left and right sides of that curve: publishers are being invited into a revenue-sharing content-on-Facebook deal now, but what is to stop the program from extending to individuals? Same thing with Facebook’s video unit, which is on pace to attract more advertisers than YouTube. Were this to happen, it’s easy to see Facebook as a one-stop shop for even more users than today, and were that to happen, advertisers would inevitably follow to an even greater degree.

    This course, though, depends on Facebook giving users exactly what they want, or at least a good enough mix, in their news feed, and as I noted, I’m not convinced personal updates is enough. Moreover, while Facebook may view “the network” as their differentiator, the fact is that a lot of “friend” sharing is indeed moving to alternative networks like Snapchat and LINE and WhatsApp. With this News Feed update I am concerned that Facebook is limiting itself and committing to a battle — the private sharing of information — it can’t necessarily win.

    Consider Facebook’s smartest acquisition, Instagram. The photo-sharing service is valuable because it is a network, but it initially got traction because of filters. Sometimes what gets you started is only a lever to what makes you valuable. What, though, lies beyond the network? That was Facebook’s starting point, and I think the answer to what lies beyond is clear: the entire online experience of over a billion people. Will Facebook seek to protect its network — and Zuckerberg’s vision — or make a play to be the television of mobile?

    I will admit, I write this analysis with mixed feelings: from a strategic perspective, I think Facebook should go for it — be the dominant interaction model on mobile for every user on earth (outside of China). On the other hand, as an advocate and beneficiary of the open web, I do fear this future and wonder: What might be the broadband to Facebook’s dial-up?

    1. Pinterest is second, with 16% of social referral, and Twitter a distant third, with only 3% of social referrals ↩

    2. AOL — now Aol — still exists of course, but it’s basically another Yahoo; relatively high-traffic sites monetized through relatively undifferentiated advertising. That said, it is more profitable…thanks to dial-up! ↩

    3. As I’ve written previously, WeChat fills these roles in China, where Facebook is banned ↩

    4. That is why I was not surprised to see this month’s Pew Research report that showed that Facebook was still the dominant social networking service amongst teenagers, past scaremongering notwithstanding ↩

    5. Buffer’s blog has a running list  ↩

  • Twitter and What Might Have Been

    Twitter’s blog is generally a cheery place, with one big exception from 2012: the innocuously named Changes Coming in Version 1.1 of the Twitter API, and in particular, the section called “Changes to the Developer Rules of the Road” that attempted to put the kibosh on 3rd-party Twitter clients:1

    If you are building a Twitter client application that is accessing the home timeline, account settings or direct messages API endpoints (typically used by traditional client applications) or are using our User Streams product, you will need our permission if your application will require more than 100,000 individual user tokens.

    As you might expect, as with any decision by a big company to lock out 3rd-party developers, people were upset. Twitter’s move, though, cut far more deeply: it wasn’t just that people really loved their 3rd-party apps; rather there was a deep sense of betrayal because 3rd-party app developers were unusually responsible for Twitter’s success up to that point.

    A Brief History of Twitter and 3rd-Party Developers

    Twitter launched as an SMS service, but quickly the web app became the company’s primary focus. No surprise: at that time, the conventional wisdom was that websites would be the dominant interaction layer for a good long time. Of course that idea seems positively quaint: we now know that on phones, the platform with by far the largest share of both computers and attention, apps are far more important than the web, and the companies that recognized that shift early have benefited tremendously. Twitter is on that list, but not by their own doing.

    Just months after the first iPhone’s release Craig Hockenberry built Twitterrific, the first mobile Twitter client, and the use case was so compelling that all sorts of people who would normally not consider jailbreaking did it anyway just for his app. In fact, I would not be surprised if Hockenberry’s app had a direct impact on Apple’s decision to open up the App Store. The impact on Twitter was just as profound: it turned out that character-limited tweets and a timeline focused on the here-and-now were a brilliant match for a computer that was in your pocket and thus available in found moments and unusual happenings. Over the next few years Twitterrific and a host of competitors honed the Twitter user experience, and these developers — and users — pioneered many of Twitter’s most important features, making the company, by 2011, worth $8 billion and on the road to a 2013 IPO. And Twitter was slamming the door in their face!

    Twitter’s Mistake

    In fact, I think Twitter’s decision was absolutely justified: in April 2010 Twitter launched promoted tweets on their website, but as just noted, it was already clear that for most users the website was a secondary platform for consuming Twitter. If Twitter were to succeed in advertising it needed to place those promoted tweets and any other advertising inventory it might devise into Twitter apps. The problem, though, was that by definition Twitter had little control over how its content was presented in apps built by 3rd-party developers. And so, three days before the promoted tweets announcement, Twitter bought Tweetie, an app many (including myself) thought was the best 3rd-party Twitter app of all.

    This was the context for the aforementioned post: advertising works best at scale but 3rd-party apps were peeling away too much of Twitter’s audience. That is why the company made such a big mistake: they didn’t kill 3rd-party apps completely. The post noted:

    We will not be shutting down client applications that use those endpoints and are currently over those token limits. If your application already has more than 100,000 individual user tokens, you’ll be able to maintain and add new users to your application until you reach 200% of your current user token count (as of today) — as long as you comply with our Rules of the Road. Once you reach 200% of your current user token count, you’ll be able to maintain your application to serve your users, but you will not be able to add additional users without our permission.

    The funny/sad thing about this entire episode is that Twitter was clearly trying to bend over backwards for its 3rd-party developers: it was strategically stupid and financially unwise to let them continue to exist, but Twitter left this massive loophole open that limited growth but didn’t kill successful apps like Twitterrific or Tweetbot. And yet, the company was pilloried and tarred with a reputation for being especially unfriendly to developers, a reputation that strongly persists to this day.

    Meerkat and Datasift

    The reason this history matters is that Twitter has spent the last month cracking down once again. The first target, famously, was Meerkat, an app that allows you to stream live from your phone to anyone else with the Meerkat app. Within days of its release the app was a huge hit in tech circles, aided by its use of the Twitter graph: you signed in with your Twitter login and were immediately able to follow anyone you already followed on Twitter. Just two weeks after its launch, though, and on the eve of South-by-Southwest, Twitter cut them off.

    The second target was more surprising, but in retrospect — like the app situation — telegraphed by Twitter. A year ago Twitter acquired Gnip, at the time one of only four companies with access to the full Twitter firehose of every tweet passing through the service. And then, over the weekend, came the other shoe: Twitter is going to cut off Gnip-competitor Datasift and everyone else.

    Twitter’s Justification

    Both of Twitter’s decisions are correct, albeit for different reasons:

    • The most obvious justification for killing Meerkat was that Twitter was on the verge of launching Periscope, an app that does basically the exact same thing (Periscope has since launched). Why should Twitter be expected to give its direct competitor a leg up?

      Still, I think that Twitter’s decision would have been justifiable even if they weren’t about to launch Periscope. A fundamental part of Twitter, first established by Twitterrific on a jailbroken iPhone, was its sense of immediacy and incredible ability to broadcast information about what was happening right now from anywhere in the world to anywhere in the world. That, though, is exactly what makes Meerkat/Periscope so amazing: immediate access to what is happening anywhere in the world is exactly what they provide, but in a superior format: video. Twitter risked being unbundled, and given their floundering user growth, the company simply can’t afford to lose its claim on its most marketable attribute.2

    • Datasift is a little more troublesome. For one thing, in the same post where Twitter announced the 3rd-party app limitation, the company promoted data analysis as an opportunity for 3rd parties to build on the platform:

      Today on Twitter we see a broad and deep variety of individual developers and companies building applications using data and content from the Twitter API…with our new API guidelines, we’re trying to encourage activity in the upper-left, lower-left and lower right quadrants, and limit certain use cases that occupy the upper-right quadrant.

      dev_chart

      Datasift falls squarely in the lower-left bracket, which means Twitter’s decision is a complete reversal of their previous stated stance, a stance that led to investments worth hundreds of millions of dollars. It’s an ugly move.

      The problem for Twitter is the aforementioned user growth: it’s even uglier. Last quarter the company was celebrated for adding 13~16 million users, but I was far more troubled by the company’s explanation that it lost 4 million users because of iOS 8. What actually happened was that if you ever signed up for Twitter on iOS 7, Safari kept syncing your tweets whether or not you used the service. iOS 8 fixed that, which means those 4 millions users were only counted as active because of an iOS bug! Given that Twitter’s reported metric is monthly active users, how many of their 288 millions active users are actually humans reading their timeline as opposed to randomly updating apps or visits by an abandoned user to a page with an embedded tweet?

      The company has responded by doubling down on monetization, and that’s where this decision makes sense. Wall Street actually loved last quarter, because Twitter monetizes so well, and in the absence of user growth the company is clearly focused on growing revenue per user instead, and data is a low-hanging fruit.

    Still, even though all of Twitter’s decisions are understandable, I can’t shake the feeling that the company could have chosen a very different path.

    What Might Have Been

    It’s interesting to ponder why it is that Twitter is able to monetize effectively – and if they can keep it up – even as they are surpassed in active users by social networks like Instagram and (soon-if-not-already) Snapchat, never mind Facebook. I made this argument at the time of Twitter’s IPO offering:

    Whereas Google is valuable because it knows what I want, when I want to get it, Facebook knows who I am, and who I know. Ideally, they also know who and what I like, but it’s a much weaker signal. Twitter, on the other hand, knows exactly what I like and what I’m interested in. It’s obvious both from what I tweet about, but especially based on who I follow. If an advertiser wants to reach someone like me – and they certainly do, given my spending habits – Twitter is by far the best way to find me. Were Twitter able to consistently capture this signal and deliver effective ad units that caught their user’s attention, they could command some of the highest average revenues per user on the Internet.

    Indeed, I would argue that what makes Twitter the company valuable is not Twitter the app or 140 characters or @names or anything else having to do with the product: rather, it’s the interest graph that is nearly priceless. More specifically, it is Twitter identities and the understanding that can be gleaned from how those identities are used and how they interact that matters.

    If one starts with that sort of understanding — that Twitter the company is about the graph, not the app — one would make very different decisions. For one, the clear priority would not be increasing ad inventory on the Twitter timeline (which in this understanding is but one manifestation of an interest graph) but rather ensuring as many people as possible have and use a Twitter identity. And what would be the best way to do that? Through 3rd-parties, of course! And by no means should those 3rd-parties be limited to recreating the Twitter timeline: they should build all kinds of apps that have a need to connect people with common interests: publishers would be an obvious candidate, and maybe even an app that streams live video. Heck, why not a social network that requires a minimum of 140 characters, or a killer messaging app? Try it all, anything to get more people using the Twitter identity and the interest graph.

    As for monetization, Twitter actually already nailed it: that’s exactly what MoPub should have been focused on. I hoped at the time of the acquisition:

    I think it’s very likely MoPub is Twitter’s AdSense: Twitter has a great signal about its users — whom I follow is a great approximation for what I’m interested in. That’s even more valuable than whom I know. However, Twitter is not a great platform for any sort of display advertising; the targeting would have to be much more precise than what is possible with current technology for users to tolerate anything more than promoted tweets. MoPub solves this riddle; Twitter can serve up highly targeted ads everywhere but Twitter proper. It’s a great acquisition.

    Mostly right, I think, except for the focus on display ads: native advertising is exactly what works best on mobile, which is why Twitter bought native advertising company Namo last summer. Twitter’s offering is an attractive alternative for mobile publishers in particular, it’s just a shame that all those publishers aren’t building on top of Twitter’s identity, but then again, who can trust the company? The past is not dead…

    Twitter and Google

    Twitter’s story in many respects makes me think of Google: both companies started out benefiting greatly from openness and the power of both connecting users to what they were interested in and opening up powerful APIs to developers. The monetization model is even similar: note the AdSense reference above. Over time, though, Google has pulled more and more of its utility onto its own pages (and the revenue balance in the company has followed), just as Twitter focused on its own apps, and now Google is even starting to eat its best customers like travel websites and insurance agents (members-only), just like Twitter ate Datasift.

    Frankly, the arc of both companies is simultaneously understandable and saddening to me. I’ve loved them both for the ways they have connected me to truly new ideas and new people, and it’s frustrating to see the growth imperative push both companies to turn increasingly inwards. One does wonder if they might find salvation in each other.3

    1. The introduction originally read as follows:

      Clearly, Twitter would like you to forget one particularly controversial moment in its history:

      Screenshot 2015-04-15 at 8.43.34 PM

      In case you can’t make out the URL, the page in question is a blog post entitled “Changes coming in Version 1.1 of the Twitter API”, and the most upsetting news was in the section called “Changes to the Developer Rules of the Road”. From The Internet Archive:

      It turns out the blog post was moved, not deleted. I apologize for the mistake ↩

    2. That’s why I don’t really object to the celebrity pressure either. ↩

    3. This eternal rumor was repeated this week; for the record, I think it’s highly unlikely, first and foremost because Twitter is simply too expensive now. And, I might add, Twitter is doing (surprisingly) ok for now, at least in the stock market; missed potential doesn’t mean outright failure. The long-term danger of slowing user growth is real though: advertisers will want to consolidate which means the revenue growth not only won’t be sustainable but could actually decline ↩

  • It’s Not 1999

    The question of whether or not we are in a tech bubble has been raised regularly for years now; 2012, particularly Facebook’s acquisition of an app1 called Instagram for a ridiculous2 $1 billion, was a particular high point. The fact that Instagram is now valued at $35 billion suggests the 2012 doomsayers were just a bit off.

    Still, it’s possible to be off in timing but right in meaning; in retrospect Alan Greenspan was correct to, at the end of 1996, characterize what we now call the dot-com bubble as a period of “irrational exuberance”; the correction just took a few extra years to materialize, and it was all-the-more painful for having taken as long to arrive as it did. Might the tech industry be facing a similar reckoning?

    I don’t think so.3

    Changing Markets

    If you’ll forgive a brief diversion, a pet peeve of mine is when people analyze the mobile phone market, particularly iOS versus Android, through the lens of Apple versus Microsoft in the 80s and 90s. The issue is not the obvious differences — this time Apple was first, the absolute numbers are much larger, etc. — but rather the fact that many of these commentators simply have their facts wrong. Windows didn’t win because it was open or all the other nonsense that is ballyhooed about; it won because MS-DOS was the operating system for IBM PCs, and at a time when personal computers were sweeping corporate America, “no one got fired for buying IBM.”4 By the time the Mac arrived in 1984, the battle was already over: businesses, the primary buyers, were already invested in MS-DOS (and, over time, Windows), and not many consumers were buying PCs. Today, of course, the situation is the exact opposite: consumers vastly outnumber business buyers. Thus, the chief reason iPhone/Android is not Windows/Mac is because the market is fundamentally different.

    I tell this story because I think a similar mistake is made when comparing today’s funding environment to the dot-com era: it’s easy to look at numbers, whether that be valuations, revenue multiples, or simple counting stats, but any analysis is incomplete without understanding markets. In 1999 most consumer markets were simply not ready, whether it be for lack of broadband, logistics build-out, etc., while most enterprise opportunities were in selling licensed software to CIOs. And, in this latter market especially, the competition was other tech companies.

    Today, by contrast, many of the most valuable unicorns are consumer-focused companies like Uber or Airbnb. Moreover, these companies are competing not with other tech companies5 but rather with entirely new (to tech) industries like transportation or hospitality. And, even for more traditional pure software plays like Snapchat or Stripe the implications of mobile-everywhere means a whole lot more time — and contexts — to reach consumers. In short, the size of the addressable market for tech companies has exploded — why shouldn’t valuations as well?

    Changing Business Models

    Today’s startups also have very different business models than companies did in the dot-com era (to the extent they had business models at all, of course). The difference is the most stark when it comes to enterprise software: back in the late 90s enterprises bought software licenses that were usually paid for up-front. Thus, when a company closed a sale, they would get paid right away.

    Today, on the other hand, most enterprise startups sell software-as-a-service (SaaS) which is paid for through subscriptions. In the long run this is a potentially more lucrative business model, as the startup can theoretically collect subscription revenue forever, but it also means revenue is much slower to arrive as compared to the old software licensing model. For example, suppose you spend $1006 to acquire a customer who pays $35/year: in year one, for that customer, you will lose $65, but then profit $35 every year thereafter. It’s a great model, but it looks bad, especially when you consider growth:

    Year New Cust Growth CAC Total Cust Total Rev Annual Profit/Loss
    1 1 — $100 1 $35 ($75)
    2 2 100% $200 3 $105 ($95)
    3 3 100% $300 6 $210 ($90)
    4 6 100% $600 12 $420 ($180)
    5 12 100% $1200 24 $840 ($360)

    No company, though, can double forever, so watch what happens when the growth rate slips to, say, ~30%:

    Year New Cust Growth CAC Total Cust Total Rev Annual Profit/Loss
    6 7 29% $700 31 $1085 $38
    7 9 29% $900 40 $1400 $500
    8 12 30% $1200 52 $1820 $620
    9 15 29% $1500 67 $2345 $845
    10 20 30% $2000 87 $3045 $1045

    This is a simplistic example: there is no churn on one hand, and no decrease in CAC (which usually happens at scale) or increase in revenue/customer via add-on services on the other. The takeaway, though, is that particularly during a period of hyper-growth, SaaS companies need a lot of capital, and more pertinently, a lot more capital than a company that monetizes through up-front software licenses.

    There is a similar dynamic for many consumer companies, particularly companies that monetize through advertising. Advertising works at scale, which in today’s world means hundreds of millions of users; getting all of those users requires years of operating without revenue, which means a lot of capital. All of this is magnified for companies that operate in markets that include network effects: network effects translate into winner-take-all opportunities, which significantly increases the growth imperative, requiring, again, significant amounts of capital.

    Changing Nature of Capital

    The question, then, is where does all that capital come from? Traditionally, from one place: the public markets.

    There are multiple advantages to an IPO, for all of the various stakeholders in a startup:

    • The company gets additional capital to fuel growth, non-dilutive shares to use for acquisitions, and a bit of added prestige that can help with sales, particularly to enterprises
    • Founders and employees can finally be fully compensated (by selling shares) for their years spent building the company
    • Venture capitalists get a return on their investment that they can distribute to their limited partners

    There are downsides, though, as well. The run-up to an IPO is very difficult, and requires a lot of attention from senior management, the disclosure of a lot of information, and a large expense that has only increased because of recent legislation. The disclosure and expense continues, too, on a quarterly basis, which brings its own pressures and risks, including activist shareholders and SEC oversight.

    On the flip side, a number of IPO advantages have been peeled away:

    • The most important has been the emergence of a new type of capital: growth capital. Growth capital is less speculative than traditional venture capital; it seeks to make relatively larger investments for relatively smaller stakes in companies with provably viable businesses that are seeking to grow for all of the reasons listed above. Now an IPO is no longer necessary for growth
    • Secondary markets and special purpose vehicles that buy stock from founders and early employees give founders and early employees a way to realize some of their gains, again, without an IPO. This too removes a previous forcing condition for an IPO
    • Finally, more and more early investors have determined that doubling down on winners often provides a better return than spreading bets widely; indeed, an increasing number of venture funds are explicitly marketed to limited partners as a combination of venture and growth capital

    Just as Arthur Rock and Fairchild Semiconductor birthed venture capital, Yuri Milner and Facebook deserve the most credit for the development of growth capital.7 In 2009 Milner and Facebook shocked the Valley with a $200 million investment that valued the company at $10.2 billion. This investment was the growth capital archetype: a large amount of money in absolute terms for a surprisingly small stake in a provably viable business, and it paid off handsomely. Facebook would go on to further expand the growth capital market, first to private equity (Elevation Partners in 2010) and then to Wall Street (Goldman Sachs in 2011): Facebook was clearly a winner, and if Goldman Sachs’ clients wanted in, then growth capital was the answer. Since then nearly every huge startup has followed the same path, and Wall Street especially has responded: growth is no longer found by investing in IPOs, but in private companies that need the money but not the hassle of an IPO, and any qualms have been drowned in an environment where multiple countries are issuing negative rate bonds (a big contrast from 1999) — growth is hard to find!8

    What’s the Downside

    So to recount:

    • Today’s startups have massively larger markets than in 1999
    • Today’s business models require significantly more capital than ever before
    • Thanks to Facebook, funding this growth via Growth Capital has been established as a viable alternative to an IPO

    In short — and I’m not the first to say this — it’s less that valuations are unnaturally high than it is the fact that there is a completely new capital market — the growth market.

    There are, though, some downsides and new risks:

    • The most obvious downside is that the best growth opportunities are increasingly out of the reach of retail investors like you or I. Goldman Sachs Facebook Special Purpose Vehicle (SPV), for example, required a minimum $2 million buy-in
    • Companies like Facebook, Uber, or Palantir may be obvious winners, but the difficulty in investing in them creates a powerful sense of FOMO — Fear of Missing Out. This has the potential of pushing less sophisticated investors desperate for growth towards higher-risk companies
    • Relatedly, there is very little oversight around these investments, particularly when it comes to the disclosure of audited financial information. A lot of these growth investors are plunging in a bit blind

    Bill Gurley, an investor I greatly admire and have learned a lot from (his blog is a must-read), is worried the last two points in particular are leading to a risk bubble:

    All of this suggests that we are not in a valuation bubble, as the mainstream media seems to think. We are in a risk bubble. Companies are taking on huge burn rates to justify spending the capital they are raising in these enormous financings, putting their long-term viability in jeopardy. Late-stage investors, desperately afraid of missing out on acquiring shareholding positions in possible “unicorn” companies, have essentially abandoned their traditional risk analysis. Traditional early-stage investors, institutional public investors, and anyone with extra millions are rushing in to the high-stakes, late-stage game.

    Read that carefully: much of the media has adopted Gurley as the apostle of the “here we go it’s 1999 all over again” mantra, but that was a valuation bubble. Companies simply weren’t worth what they were priced at. Gurley is arguing that the private market with its limited information and oversight is producing something very different: investors putting too much money in companies without enough information or enough potential upside to justify the risk.

    To be fair, given that risk should be priced in, a risk bubble is a valuation bubble. The fact of the matter, though, is that it’s the word “bubble” that is mistaken. Go back to concern one: the lack of access for retail investors. It turns out this has its advantages, because the lack of liquidity is arguably a firewall against this truly being a bubble. For one, if everything goes sour, the folks taking a hit can very much afford it. More importantly, a bubble is less about companies than it is a bet on euphoria — the assumption that no matter how nonsensical your investment, there will always be someone on the other end ready-and-willing to buy. While many of these growth investors are arguably making such a bet on a future IPO that may or may not materialize, there is no question the combination of increased investor sophistication necessary to participate in this market and the lack of liquidity makes betting on euphoria a far more risky — and thus far more unlikely — outcome.

    Or, to put it another way, bubble talk is less 1999 than it is Y2K: a potential problem, and some people will lose money, but in all likelihood a far smaller deal than many in the media are making it out to be.

    1. Make sure you read this sentence with the proper amount of condescension; if you have forgotten, it was exceptionally heavy ↩

    2. This should be read with disgust ↩

    3. And this is where I desperately hope this article isn’t hung around my neck in a year or two ↩

    4. Moreover, it was IBM, not Microsoft, that made the PC open in order to get it more quickly to market, and it was a decision they came to dearly regret when “IBM-compatible” PCs, led by Compaq, undercut the IBM PC and came to own the growing market (and why were they “IBM-compatible”? Because they ran MS-DOS on Intel chips)  ↩

    5. Sorry Lyft ↩

    6. I.e. your customer acquisition cost (CAC) is $100 ↩

    7. Although Chris Sacca deserves special credit for coming up with the idea of special purpose vehicles that raise money in order to invest in a specific startup, an increasingly common vector for growth capital ↩

    8. As an aside, this is why I don’t take the relatively poor performance of many recent IPOs as a bubble indicator; increasing stock prices are about an increasing expectation of future growth, but if much of the growth has been realized, then why should we expect stocks to do anything more than hover? ↩