Of the three core techniques of blitzscaling, the first and most foundational is to design an innovative business model capable of exponential growth.
The story of entrepreneurship in the Internet era is a story of this kind of business model innovation.
Think back to the dot-com era, which stretched roughly from the IPO of Netscape in 1995 until the NASDAQ began to crash in 2000. During this period, enormous numbers of start-ups and pretty much every established company tried to build great Internet businesses, yet nearly all of them failed. The problem was, most of them simply tried to cut and paste existing business models onto the new online medium. You can’t transplant a heart from one species into another and expect it to thrive.
If you had asked stock market analysts in 1995 which companies were best positioned to dominate the Internet, most would have pointed to existing giants like Microsoft and Time Warner, which invested millions in Internet businesses like MSN and Pathfinder. Others would have mentioned “pure play” dot-com start-ups like eToys, which combined proven business models like the “category killer” store with the new online medium.
Yet when the wreckage of the dot-com crash cleared, the most successful companies still charging full steam ahead were the few start-ups that were designed around totally new business models, such as Amazon, eBay, and Google.
Walmart should have dominated online retail, yet Amazon emerged and practically wrote the bible for e-commerce, including consumer reviews, shopping carts, and free shipping. Newspapers and phone book companies should have been able to transfer their information businesses to the online world, but Yahoo! and then Google stepped up to the plate. They built the search engines that indexed the world’s information, and Google developed the business model that made it worth more than all traditional media companies combined.
In contrast, and much to their misfortune, start-ups that relied purely on technology innovation without any real business model innovation largely went bust. Companies like eToys that tried to “Amazon” various markets, but without Amazon’s front- and back-office innovations, crashed and burned once the financial markets began to demand profits rather than just expensive revenue growth. Even Netscape, whose Netscape Navigator mainstreamed Web browsing, and whose IPO kicked off the dot-com boom, was forced to sell itself off to AOL. Netscape engineers invented JavaScript, SSL, and all kinds of cool technology for the Internet that are still used today, but Netscape accepted the status quo when it came to using tried-and-true business models rather than developing new ones that were enabled by its own technology innovation. Unfortunately for Netscape, its competitor Microsoft already understood those business models all too well and knew exactly how to use its economic might and resources to pull their levers. In the first “browser war,” Microsoft preinstalled its Internet Explorer on all new Windows computers, then gave away its Web server software, Internet Information Server (IIS), which effectively destroyed Netscape’s business model.
Could Netscape have succeeded with a different strategy? We believe so. Consider that one of the ways that Netscape monetized its Navigator browser was to sell the sponsorship of its Net Search button to the Excite search engine for $5 million. Netscape believed that the browser itself was the key, while search was simply a sideline. It was left to two pairs of Stanford graduate students, Jerry Yang and David Filo (Yahoo!) and Larry Page and Sergey Brin (Google), to prove that search was a much bigger business. Google’s innovative model of selling text ads next to search results via an automated marketplace allowed it to build a franchise so dominant that it later withstood a series of frontal assaults by Microsoft, including a marketing program in which Microsoft essentially paid people to use its Bing search engine.
The same story has been repeated in multiple waves since. Facebook and LinkedIn dominate social networks even though AOL, Microsoft (Hotmail), and Yahoo! (Yahoo! Mail) controlled most consumer online identities when those social networks first emerged. Alibaba beat eBay in China. Uber outflanked the taxi companies. Airbnb has more room listings than any hotel company in the world.
These success stories are technology companies, sure. But as we’ve seen, technological innovation alone is insufficient—even when its impact on the future is huge. Services like Craigslist, Wikipedia, and IMDb (the Internet Movie Database) were early, influential Internet innovators, but they still never became massively (financially) valuable on their own.
The real value creation comes when innovative technology enables innovative products and services with innovative business models. Even though the business models of Google, Alibaba, and Facebook might seem obvious—even inevitable—after the fact, they weren’t widely appreciated at the time they launched. How many people in 1999 would have realized that running tiny text ads next to the equivalent of an electronic card catalog would lead to the world’s most valuable software company? Or that setting up an online shopping mall for China’s emerging middle class would lead to a $100 billion business? Which of you in 2004 would have predicted that letting people see what their friends are talking about by staring at a tiny screen on a handheld computer would become the dominant form of media? Great companies and great businesses often seem to be bad ideas when they first appear because business model innovations—by their very definition—can’t point to a proven business model to demonstrate why they’ll work.
To really understand why these business models succeed, we need to clearly define what we mean by “business model” in the first place. Part of the problem is that the term can be interpreted in so many different ways. The great management thinker Peter Drucker wrote that business models are essentially theories composed of assumptions about the business, which circumstances might require to change over time. Harvard Business School professor and author Clay Christensen believes that you need to focus on the concept of the “job-to-be-done”; that is, when a customer buys a product, she is “hiring” it to do a particular job. Then there’s Brian Chesky of Airbnb, who said simply, “Build a product people love. Hire amazing people. What else is there to do? Everything else is fake work.”
As Andrea Ovans aptly put it in her January 2015 Harvard Business Review article, “What Is a Business Model?”, it’s enough to make your head swim! For the purposes of this book, we’ll focus on the basic definition: a company’s business model describes how it generates financial returns by producing, selling, and supporting its products.
What sets companies like Amazon, Google, and Facebook apart, even from other successful high-tech companies, is that they have consistently been able to design and execute business models with characteristics that allow them to quickly achieve massive scale and sustainable competitive advantage. Of course, there isn’t a single perfect business model that works for every company, and trying to find one is a waste of time. But most great business models have certain characteristics in common. If you want to find your best business model, you should try to design one that maximizes four key growth factors and minimizes two key growth limiters.
The most basic growth factor to consider for your business model is market size. This focus on market size may sound obvious, and it’s right out of Pitch Deck 101 for start-ups, but if you want to build a massive company, you need to begin with the basics and eliminate ideas that serve too small of a market.
A big market has both a large number of potential customers and a variety of efficient channels for reaching those customers. That last point is important; a market consisting of “everyone in the world” might seem large, but it isn’t reachable in any efficient way. We’ll discuss this in greater depth when we look at distribution as a key growth factor.
It’s not easy to judge the size of a market, or what pitch decks and venture capitalists often refer to as TAM (total available market). Predicting TAM and how it will grow in the future is one of the main sources of uncertainty in blitzscaling. But predicting it correctly and investing accordingly when others are still paralyzed by fear is also one of the main opportunities for unexpectedly high returns, as we’ll see in the cases of Airbnb and Uber.
Ideally, the market itself is also growing quickly, which can make a smaller market attractive and a large market irresistible.
In Silicon Valley, the competition for venture capital exerts a strong pressure on entrepreneurs to focus on ideas that are going after big markets. Venture capital firms might raise hundreds of millions or even billions of dollars from their investors—limited partners like pension funds and university endowments—who are seeking above-market returns to compensate them for taking a chance on privately held companies rather than simply investing in the Coca-Colas of the world. To deliver these above-market returns, venture capital funds need to at least triple their investors’ money. A $100 million venture capital fund would need to return $300 million over the typical seven- to ten-year life of a fund to achieve an above-market internal rate of return of 15 to 22 percent. A $1 billion fund would need to return $3 billion. Since most venture capital investments either lose money or barely break even, the only realistic way that venture capitalists can achieve these aggressive goals is to rely on a small number of incredibly successful investments. For example, Benchmark Capital invested $6.7 million in eBay in 1997. Less than two years later, eBay went public, and Benchmark’s stake was worth $5 billion, which is a 745 times return. The specific fund that made that investment, Benchmark Capital Partners I, took $85 million from investors and returned $7.8 billion, for a 92 times return. (The initial investors in Facebook did even better, but were individuals rather than firms.)
Given the desire for home runs like eBay, most venture capitalists filter investment opportunities based on market size. If a company can’t achieve “venture scale” (generally, a market of at least $1 billion in annual sales), then most VCs won’t invest, even if it is a good business. It simply isn’t large enough to help them achieve their goal of returning more than three times their investors’ money.
When Brian Chesky was pitching venture capitalists to invest in Airbnb, one of the people he consulted was the entrepreneur and investor Sam Altman, who later became the president of the Y Combinator start-up accelerator. Altman saw Chesky’s pitch deck and told him it was perfect, except that he needed to change the market-size slide from a modest $30 million to $30 billion. “Investors want B’s, baby,” Altman told Chesky. Of course, Altman wasn’t telling Chesky to lie; rather, he argued that if the Airbnb team truly believed in their own assumptions, $30 million was a gross underestimate, and they should use a number that was true to their convictions. As it turns out, Airbnb’s market was indeed closer to $30 billion.
When evaluating market size, it’s also critical to try to account for how lower costs and product improvements can expand markets by appealing to new customers, in addition to seizing market share from existing players. In 2014, Aswath Damodaran, a professor of finance at NYU’s Stern School of Business, estimated that Uber was probably worth roughly $6 billion, based on its ability to ultimately win 10 percent of the global taxi market of $100 billion, or $10 billion. According to Uber’s own projections, in 2016 the company processed over $26 billion in payments. It’s safe to say that the $10 billion market was a serious underestimate, as the ease of use and lower cost of Uber and its competitors expanded the market for transportation-as-a-service.
As Aaron Levie, the founder of the online file storage company Box noted in a tweet in 2014, “Sizing the market for a disruptor based on an incumbent’s market is like sizing a car industry off how many horses there were in 1910.”
The other factor that can lead to underestimating a market is neglecting to account for expanding into additional markets. Amazon began as Amazon Books, the “Earth’s Biggest Bookstore.” But Jeff Bezos always intended for bookselling to serve as a beachhead from which Amazon could expand outward to encompass his massive vision of “the everything store.” Today, Amazon dominates the bookselling industry, but thanks to relentless market expansion, book sales represent less than 7 percent of Amazon’s total sales.
The same effect can be seen in the financial results of Apple. In the first quarter of 2017, Apple generated $7.2 billion from the sale of personal computers, a category the company pioneered and once dominated. That’s a great number to be sure, but, over that same financial quarter, Apple’s total revenue was a whopping $78.4 billion, which meant that Apple’s original market accounted for less than 10 percent of its total sales.
My Greylock colleague Jerry Chen, who helped Diane Greene scale VMware’s virtualization software into a massive business, likes to point out, “Every billion-dollar business started as a ten-million-dollar business.”
But whether you are creating a new market, expanding an existing market, or relying on adjacent markets to get to those “B’s” that investors want (baby), you need to have a plausible path to get from here to there. This leads us to one of my favorite growth factors to discuss with entrepreneurs: distribution.
The second growth factor needed for a strong, scalable business is distribution. Many people in Silicon Valley like to focus on building products that are, in the famous words of the late Steve Jobs, “insanely great.” Great products are certainly a positive—we’ll discuss the lack of product quality as a growth limiter later on—but the cold and unromantic fact is that a good product with great distribution will almost always beat a great product with poor distribution.
Dropbox is a company with a great product, but it succeeded because of its great distribution. In an interview for Reid’s Masters of Scale podcast, founder and CEO Drew Houston said that he believes that too many start-ups overlook the importance of distribution:
Most of the orthodoxy in Silicon Valley is about building a good product. I think that’s because most companies in the Valley don’t survive beyond the building-the-product phase. You have to be good at building a product, then you have to be just as good at getting users, then you have to be just as good at building a business model. If you’re missing any of the links in the chain, the whole chain is broken.
The challenge of distribution has become even greater in the “mobile first” era. Unlike the Web, where search engine optimization and e-mail links were broadly applicable and successful distribution channels, mobile app stores offer little opportunity for serendipitous product discovery. When you go to Apple’s or Google’s app store, you’re searching for a specific product. Few people install apps just for the hell of it. As a result, the business model innovators who have succeeded (e.g., Instagram, WhatsApp, Snap) have had to find creative ways to get broad distribution for their product—without spending a lot of money. These distribution techniques fall into two general categories: leveraging existing networks and virality.
New companies rarely have the reach or resources to simply pour money into advertising campaigns. Instead, they have to find creative ways to tap into existing networks to distribute their products.
When I was at PayPal, one of the major vehicles for distribution of our payment service was settling purchases on eBay. At the time, eBay was already one of the largest players in e-commerce, and by the beginning of 2000 already had ten million registered users. We tapped into this user base by building software that made it extremely easy for eBay sellers to automatically add a “Pay with PayPal” button to all of their eBay listings. The amazing thing is that customers did so even though eBay had its own rival payments service, Billpoint! But sellers were required to add Billpoint manually to each of their listings; PayPal did it for them.
Many years later, Airbnb was able to perform a similar feat by leveraging the online classified service Craigslist. Based on a suggestion from Y Combinator’s Michael Seibel, Airbnb built a system that allowed and encouraged its hosts to cross-post their listings to the much-larger Craigslist. Hosts were told, “Reposting your listing from Airbnb to Craigslist increases your earnings by $500 a month on average,” and were allowed to do so by clicking a single button. This took serious technology skills—unlike many platforms, Craigslist doesn’t have an application programming interface (API) that allows other software to interact with it—but it was technology innovation for the purposes of distribution innovation, not product innovation. “It was a kind of a novel approach,” Airbnb founder Nathan Blecharczyk said of the integration. “No other site had that slick an integration. It was quite successful for us.”
Leveraging an existing network can have downsides, of course. What the existing network gives (or unknowingly allows to be taken), the existing network can also take away. Zynga, the leading social games company, achieved great success leveraging Facebook for distribution, but had to dramatically reengineer its distribution model after Facebook decided to stop allowing people playing Zynga games to post their progress to their Facebook friends. (Disclosure: I am a member of Zynga’s board of directors.) Zynga founder Mark Pincus was farsighted enough to build a strong enough franchise to survive the change.
In contrast, so-called content farms like Demand Media that leveraged Google’s search platform to generate website traffic and advertising revenues never recovered after Google tuned its algorithms to deprioritize content from what it called “junk” websites.
Despite these dangers, leveraging existing networks can be a critical part of a business model, especially if these networks can provide a “booster rocket” that is later supplemented with virality or network effects.
“Viral” distribution occurs when the users of a product bring more users, and those users bring additional users, and so on, much like an infectious virus spreads from host to host. Virality can either be organic—occurring during the course of normal usage of the product—or incentivized by some kind of reward.
After launching LinkedIn, the team and I devoted significant time and energy to figuring out how to improve organic virality; that is, how to make it easier for existing users to invite friends to use the service. One way we did this was to refine what have become some of the standard tools of virality, such as address book importers. For example, we built software that allowed LinkedIn to connect to our users’ Outlook contacts, which made it very easy for them to invite their most important connections.
But equally important was an unanticipated source of virality. As it turned out, users wanted to use their LinkedIn pages as their primary professional identity on the Internet. Having a page like this to point others to—with all the details of their professional life together in one place—generated value not only for the user, but for the people viewing the page, and it made viewers realize that they should get their own LinkedIn profile. As a result, we added public profiles as a systematic tool to boost both the member value proposition and our viral growth rate.
At PayPal, we combined organic and incentivized virality. The payment product was inherently viral; if someone e-mailed you money using PayPal, you had to set up an account to get paid. But we enhanced this organic virality with monetary incentives. If you referred a friend to PayPal, you got $10, and your friend got $10. This combination of organic and incentivized virality allowed PayPal to grow 7 to 10 percent per day. As the PayPal network grew, we reduced the incentives to $5 and $5, then finally eliminated them altogether.
Incentives don’t have to be monetary; like PayPal, Dropbox used a similar combination of organic virality (as users share files with nonusers) and incentivized virality (Basic account holders get 500 MB of extra storage per user they refer; Pro account holders get 1 GB) to grow. Even though Dropbox invested in partnerships with leading PC makers like Dell, Drew Houston credits virality with driving the company’s rapid growth, helping it double its one hundred thousand users at launch to two hundred thousand users just ten days later, then skyrocket to one million users just seven months after that.
If your distribution strategy focuses on virality, you also have to focus on retention. Bringing new users in through the front door doesn’t help you grow if they immediately turn around and leave. According to Houston, Dropbox discovered this truth the hard way, when activation rates revealed that only 40 percent of the people signing up were actually putting files in their Dropbox and linking them to their computers. In an interview for my Masters of Scale podcast, Drew described a scene reminiscent of the television show Silicon Valley (but with a happier ending):
What we did is we went on Craigslist and offered $40 to anyone who’d come in for half an hour—a poor man’s usability test. We’re like, “All right, sit down. This is an invitation to Dropbox in your e-mail. Go from here to sharing a file with this e-mail address.” Zero of the five people we tested succeeded. Zero of the five even came close. This was just stunning. We’re like, “Oh my God, this is the worst product ever created.” So we made a list of like eighty things in this Excel spreadsheet, then just sanded down all these rough edges in the experience, and watched our activation rate climb.
Virality almost always requires a product that is either free or freemium (i.e., free up to a certain point, after which the user has to pay to upgrade—Dropbox, for example, offers 2 GB of free storage). We can’t recall a single instance of a company that grew to a massive scale by leveraging the virality of a paid product.
One of the most powerful distribution innovations is to combine both strategies. Facebook was able to do this by harnessing the organic virality of a social network (where users invite other users to join them) and leveraging existing networks centered around campuses by rolling out the product on a college-by-college basis. We’ll discuss Facebook’s rollout strategy in greater depth when we consider network effects.
One of the key growth factors that entrepreneurs often overlook is the power of high gross margins. Gross margins, which represent sales minus the cost of goods sold, are probably the best measure of long-term unit economics. The higher the gross margin, the more valuable each dollar of sales is to the company because it means that for each dollar of sales, the company has more cash available to fund growth and expansion. Many high-tech businesses have high gross margins by default, which is why this factor is often overlooked. Software businesses have high gross margins because the cost of duplicating software is essentially zero. Software-as-a-service (SaaS) businesses have a slightly higher cost of goods sold because they need to operate a service, but thanks to cloud providers like Amazon, this cost is becoming smaller all the time.
In contrast, “old economy” businesses often have low gross margins. Growing wheat is a low-margin business, as is selling goods in a store or serving food in a restaurant. One of the most amazing things about Amazon’s success is that it has been able to build a massive business based on retailing, which is generally a low-margin industry. And even Amazon now relies heavily on its high-margin SaaS business, Amazon Web Services (AWS). In 2016, AWS accounted for 150 percent of Amazon’s operating income, which means that the retail business actually lost money.
Most of the valuable companies we’re focusing on in this book have gross margins of over 60, 70, or even 80 percent. In 2016, Google had a gross income of $54.6 billion on sales of $89.7 billion, for a gross margin of 61 percent. Facebook’s gross income was $23.9 billion on sales of $27.6 billion, for a gross margin of 87 percent. In 2015, LinkedIn’s gross margin was 86 percent. As we’ve already discussed, Amazon is the outlier, with a 2016 gross income of $47.7 billion on $136 billion in sales, for a gross margin of 35 percent. Yet even Amazon’s gross margins are greater than those of a “high margin” traditional company like General Electric, which in 2016 had a gross income of $32.2 billion on sales of $119.7 billion, for a gross margin of 27 percent.
High gross margins are a powerful growth factor because, as noted below, not all revenue is created equal. The key insight here is that even though gross margins matter a great deal to the seller, they are irrelevant to the buyer. How often do you consider the gross margin involved when you make a purchase? Would you ever choose Burger King over McDonald’s because Whoppers are lower margin than Big Macs? Typically, you focus solely on the cost to you, and the perceived benefits of the purchase. This means that it’s not necessarily any easier to sell a low-margin product than a high-margin product. If possible then, a company should design a high-gross-margin business model.
Second, high-gross-margin businesses are attractive to investors, who will often pay a premium for the cash-generating power of such a business. As the prominent investor Bill Gurley wrote in his 2011 blog post, “All Revenue Is Not Created Equal,” “Investors love companies where, all things being equal, higher revenues create higher profit margins. Selling more copies of the same piece of software (with zero incremental costs) is a business that scales nicely.” Appealing to investors makes it easier to raise larger amounts of money at higher valuations when the company is privately held (we’ll delve into the details of why this is so important later on), and lowers the cost of capital when the company is publicly traded. This access to capital is a key factor in being able to finance lightning-fast growth.
It’s important to note the difference between potential gross margin and realized gross margin. Many blitzscalers, such as Amazon or the Chinese hardware makers Huawei and Xiaomi, deliberately price their products to maximize market share rather than gross margins. As Jeff Bezos is fond of saying, “Your margin is my opportunity.” Xiaomi explicitly targets a net margin of 1 to 3 percent, a practice it credits Costco for inspiring. All other factors being equal, investors almost always place a much higher value on companies with higher potential gross margins than companies that have already maximized their realized gross margins.
Finally, most of a company’s operational challenges scale based on revenues or unit sales volume, not gross margin. If you have a million customers who generate $100 million per year in sales, the cost to serve those customers doesn’t change whether your gross margin is 10 percent or 80 percent; you still need to hire enough people to respond to their support requests. But it’s a lot easier to afford good customer support when you have $80 million in gross margin to spend rather than $10 million.
Conversely, it’s a lot easier to sell and service 125,000 customers who generate $12.5 million per year in sales and $10 million in gross margin than it is to have to sell and service a million customers who generate $100 million in sales to achieve that same $10 million in gross margin. That’s eight times as many customers and eight times the revenues, which means eight times as many salespeople, customer service representatives, accountants, and so on.
Designing a high-gross-margin business model makes your chances of success greater and the rewards of success even greater. As we’ll see in a later section, high gross margins have helped even nontech businesses, such as the Spanish clothing retailer Zara, grow into global giants.
Market size, distribution, and gross margins are important factors in growing a company, but the final growth factor plays the key role in sustaining that growth long enough to build a massively valuable and lasting franchise. While the past twenty years have driven improvements in the first three growth factors, the rise in Internet usage around the world has pushed network effects to levels never before seen in our economy.
The increasing importance of network effects is one of the main reasons that technology has become a more dominant part of the economy.
At the end of 1996, the five most valuable companies in the world were General Electric, Royal Dutch Shell, the Coca-Cola Company, NTT (Nippon Telegraph and Telephone), and ExxonMobil—traditional industrial and consumer companies that relied on massive economies of scale and decades of branding to drive their value. Just twenty-one years later, in the fourth quarter of 2017, the list looked very different: Apple, Google, Microsoft, Amazon, and Facebook. That’s a remarkable shift. Indeed, while Apple and Microsoft were already prominent companies at the end of 1996, Amazon was still a privately held start-up, Larry Page and Sergey Brin were still a pair of graduate students at Stanford who were two years away from founding Google, and Mark Zuckerberg was still looking forward to his bar mitzvah.
So what happened? The Networked Age happened, that’s what.
Technology now connects all of us in ways that were unthinkable to our ancestors. Over two billion people now carry smartphones (many of them made by Apple, or using Google’s Android operating system) that keep them constantly connected to the global network of everything. At any time, those people can find almost any information in the world (Google), buy almost any product in the world (Amazon/Alibaba), or communicate with almost any other human in the world (Facebook/WhatsApp/Instagram/WeChat).
In this highly connected world, more companies than ever are able to tap into network effects to generate outsize growth and profits.
We’ll use the simple layman’s definition of network effects in this book:
A product or service is subject to positive network effects when increased usage by any user increases the value of the product or service for other users.
Economists refer to these effects as “demand-side economies of scale” or, more generally, “positive externalities.”
The magic of network effects is that they generate a positive feedback loop that results in superlinear growth and value creation. This superlinear effect makes it very difficult for any node in the network to switch from an incumbent to an alternative (“customer lock-in”), since it is almost impossible for any new entrant to match the value of plugging into the existing network. (Nodes in these networks are typically customers or users, as in the canonical example of the fax machine, or the more recent example of Facebook, but can also be data elements or other fundamental assets valuable in a business.)
The resulting phenomenon of “increasing returns to scale” often results in an ultimate equilibrium in which a single product or company dominates the market and collects the majority of its industry’s profits. So it’s no surprise that smart entrepreneurs strive to create (and smart investors want to invest in) these network effects start-ups.
Several generations of start-ups have tapped these dynamics to build dominant positions, from eBay to Facebook to Airbnb. To accomplish these goals, it’s critical to develop a rigorous understanding of how network effects work. My Greylock colleague Simon Rothman is one of the world’s premier experts on network effects from building eBay’s $14 billion automotive marketplace. Simon warns, “A lot of people try to bolt on network effects by doing things like adding a profile. ‘Marketplaces have profiles,’ they reason, ‘so if I add profiles, I’ll be adding network effects.’ ” Yet the reality of building network effects is a bit more complicated. Rather than simply imitate specific features, the best blitzscalers study the different types of network effects and design them into their business models.
On his industrial organization of information technology website, the NYU professor Arun Sundararajan classifies network effects into five broad categories:
Direct Network Effects: Increases in usage lead to direct increases in value. (Examples: Facebook, messaging apps like WeChat and WhatsApp)
Indirect Network Effects: Increases in usage encourage consumption of complementary goods, which increases the value of the original product. (Example: Adoption of an operating system such as Microsoft Windows, iOS, or Android encourages third-party software developers to build applications, increasing the value of the platform.)
Two-Sided Network Effects: Increases in usage by one set of users increases the value to a different set of complementary users, and vice versa. (Example: Marketplaces such as eBay, Uber, and Airbnb)
Local Network Effects: Increases in usage by a small subset of users increases the value for a connected user. (Example: Back in the days of metered calls, certain wireless carriers allowed subscribers to specify a limited number of “favorites” whose calls didn’t count against the monthly allotment of call minutes.)
Compatibility and Standards: The use of one technology product encourages the use of compatible products. (Example: within the Microsoft Office suite, Word’s dominance meant that its document file format became the standard; this has allowed it to destroy competitors like WordPerfect and fend off open-source solutions like OpenDocument.)
Any of these different network effects can have a major impact; Microsoft’s ability to tap into multiple network effects with Windows and Office contributed greatly to its unprecedentedly durable franchise. Even today, Windows and Office remain dominant in the PC market; it’s simply that other platforms like mobile have achieved similar or greater importance.
A key element of leveraging network effects is the aggressive pursuit of network growth and adoption. Because the impact of network effects increases in a superlinear fashion, at lower levels of scale, network effects actually exert downward pressure on user adoption. Once all your friends are on Facebook, you have to be on Facebook too. But conversely, why would you join Facebook if none of your friends had joined yet? The same is true for the first user of marketplaces like eBay and Airbnb.
With network effects businesses, you can’t start small and hope to grow slowly; until your product is widely adopted in a particular market, it offers little value to potential users. Economists would say that the business has to get past the “tipping point” where the demand curve intersects with the supply curve. Companies like Uber subsidize their customers in an attempt to manipulate the demand curve to reach that tipping point faster; the bet is that losing money in the short term may allow you to make money in the long term, once you’re past the tipping point.
One challenge that this approach produces is the (eventual) need to eliminate the subsidies in order to make the unit economics work. When I was at PayPal, one of the things we did to encourage adoption was to proclaim that the service would always be free. This meant eating the transaction costs of accepting credit card payments. I wish I could say we had a grand plan. We had hoped that we could make up for the credit card transaction fee subsidy by making money off the float—the funds being kept in PayPal. Unfortunately, this came nowhere close to offsetting the fee subsidies, and the company was hemorrhaging money. So we switched PayPal from “always free” to “ACH always free” and started charging fees to accept credit card payments. Fortunately, we already had a loyal following, and our customers accepted the change.
When the business can’t change the economics of the product (free services like Facebook can’t lower their prices), it can instead sway the expectations of potential users. The value users place on the service when deciding whether or not to adopt it depends on both the current level of adoption and their expectations for future adoption. If they think others are going to jump on board, the perceived value of the service increases, and they become more likely to adopt it.
This technique is reflected in one of the most influential business books of all time, Geoffrey Moore’s Crossing the Chasm. Moore argues that technology companies often run into problems when they try to transition from a market of early adopters to the mainstream—the proverbial “chasm.” He recommends that companies focus on niche beachhead markets, from which the company can expand outward using a “bowling pin” strategy in which these markets help to open up adjacent markets. This strategy is even more important for network effects businesses.
A company can also reshape the demand curve by designing the product to be valuable to the individual user regardless of network adoption. At LinkedIn, for example, we discovered that public LinkedIn profiles had some value independent of the user’s network, since they served as an online professional identity. This gave people a reason to join LinkedIn even if their friends and colleagues hadn’t done so yet.
In addition to supporting network effects, the high connectivity of the world we live in today also makes it easier to reach the tipping point where network effects kick in, and to sustain those network effects and the market dominance they produce.
First, the Internet has driven the cost of discovery for products and services lower than ever. Unlike in the past, when companies needed to offer goods in retail stores or broadcast advertising in order to be visible to customers, today buyers can find whatever they’re looking for on Amazon or other online marketplaces like Alibaba, in app stores, or, when all else fails, by Googling. Because products and services that are already popular will almost always come up first in search results, companies with a competitive advantage can quickly grow to the point where the increasing returns of network effects produce a winner-take-most or winner-take-all market. This also explains why the growth factor of distribution is as or more important to company success as the product itself—without distribution, it is difficult to reach the tipping point.
After network effects take hold, the efficiencies enabled by the Networked Age make it easier to sustain the pace of rapid growth. In the past, rapid customer growth inevitably led to rapid organizational growth and to dramatic increases in the overhead required to coordinate a large number of employees and teams. Today’s networks allow companies to sidestep these traditional growth limiters, such as when Apple used Foxconn to get around the potential limitation of its manufacturing infrastructure (more on this in the next section). The more you can remove those limiters, the more dominant a network effects–driven business can grow. This is why companies like Google that have surpassed the $100 billion mark in annual revenues are still growing at over 20 percent per year.
Finally, the remarkable profitability of these companies gives them the financial resources to expand into new fields and invest in the future. The S-curve of innovation argues that the rate of adoption of every innovation eventually slows as the market saturates. However, companies like Apple have mastered the strategy of investing in new products that let them hop onto additional S-curves. Apple hopped from music players to smartphones to tablets, and it is no doubt spending some of its vast profits chasing the next S-curve. The premium that the public markets grant these companies also helps them use mergers and acquisitions (M&A) to jump these curves, much as Facebook did with Instagram, WhatsApp, and Oculus, and Google did with DeepMind.
Of course, network effects don’t apply to every company or market, even if they are superficially similar—as many companies and their investors discovered to their chagrin during the dot-com bust, the Great Recession, and the funding slowdown of 2016. This is why the best entrepreneurs try to design innovative business models that leverage network effects. One of the reasons that Google is Google and Yahoo! is now part of AOL (which in turn is owned by Verizon) is that Google focused on AdWords (a marketplace with strong network effects) while Yahoo! tried to become a media company (a traditional model based on economies of scale).
Much of Silicon Valley’s historical success in building giant companies can be traced to its cultural emphasis on business model innovation, which results in the creation of network effects–driven businesses. The irony is that many people in Silicon Valley couldn’t define a network effect or what caused it if asked. Yet simply because so many entrepreneurs are trying so many different business models, they can end up stumbling into powerful network effects. Craig Newmark simply started e-mailing his friends about local events in 1995; almost twenty-two years later, network effects have kept Craigslist a dominant player in online classifieds despite operating with a skeleton crew and making seemingly no changes to the website design during that entire period!
This is where an emphasis on speed also plays an important role. Because Silicon Valley’s entrepreneurs focus on designing business models that can get big fast, they are more likely to incorporate network effects. And because the fierce local competition forces start-ups to grow so aggressively (i.e., blitzscale), Silicon Valley start-ups are more likely to reach the tipping point of network effects before start-ups from less aggressive geographies.
One of the motivations for this book is to help entrepreneurs from around the world emulate these successes by teaching them how to systematically design their businesses for blitzscaling. When you design your business model to leverage network effects, you can succeed anywhere.
Building key growth factors into your innovative business model is only half the battle. It is fiendishly difficult to grow an amazing business, in part because it is fiendishly easy to run smack into obstacles that limit your growth. A key component of business model innovation is designing around these growth limiters.
Product/market fit enables rapid growth, while the lack of it makes growth expensive and difficult. The concept of product/market fit originates in Marc Andreessen’s seminal blog post “The Only Thing That Matters.” In his essay, Andreessen argues that the most important factor in successful start-ups is the combination of market and product.
His definition couldn’t be simpler: “Product/market fit means being in a good market with a product that can satisfy that market.”
Without product/market fit, it’s impossible to grow a start-up into a successful business. As Andreessen notes,
You see a surprising number of really well-run start-ups that have all aspects of operations completely buttoned down, HR policies in place, great sales model, thoroughly thought-through marketing plan, great interview processes, outstanding catered food, 30" monitors for all the programmers, top tier VCs on the board—heading straight off a cliff due to not ever finding product/market fit.
Unfortunately, it’s far easier to define product/market fit than it is to establish it!
When you start a new company, the key product/market fit question you need to answer is whether you have discovered a nonobvious market opportunity where you have a unique advantage or approach, and one that competing players won’t see until you’ve had a chance to build a healthy lead. It’s usually difficult to find such an opportunity in a “hot” space; if an opportunity is obvious to everyone, the chance that you’ll be the one who succeeds is exceedingly low.
Most nonobvious opportunities arise from a change in the market that the incumbents aren’t willing or able to adapt to. In many cases, this can be a disruptive technological innovation, but it can also be a change in the law or financial regulations, the rise of a new group of customers, or any other major shift. For example, Charles Schwab was able to build his eponymous financial empire by leveraging the deregulation of brokerage commissions to launch a discount brokerage.
Frequently, you won’t be able to fully validate product/market fit before you commit to building a company. But you should try. As authors and entrepreneurs, we’re huge fans of Eric Ries and his lean start-up methodology. It is an excellent process for systematically tackling risk. But the fact is that most start-ups don’t follow that process; instead, their chosen experiment is “Do we succeed or run out of money?”
The best way for a small, resource-strapped team to assess potential strategies is to leverage what we dubbed “network intelligence” in our previous book, The Alliance. Even a small group of founders is likely to have a huge collective personal network of smart people with relevant knowledge or experience. Initiate a conversation, inviting them to challenge your idea and tell you what else you should consider.
Of course, even the best network intelligence won’t guarantee that you’ve actually found product/market fit during this design phase. The only way to truly prove product/market fit is to get the product into the hands of real users. But entrepreneurs can and should do their research, and try to design their business model to maximize their chances of achieving product/market fit as quickly as possible.
Designing a scalable economic model isn’t enough if you can’t scale up your operations to meet demand. Too often, entrepreneurs dismiss the challenges of operational scalability by saying, “Managing explosive growth is a high-class problem.” High-class problems are still problems; it may feel better for your ego to be wrestling with the issues of growth rather than simply trying to avoid missing payroll, but both can still kill your company. Rather than dismiss these challenges, the wisest innovators design operational scalability into their models.
A significant number of operational issues arise simply because of human limitations. As much as we might wish that we and our colleagues could work tirelessly and seamlessly, regardless of the scale of the organization, the fact is that growth causes us to trip over a wide array of issues.
If you are leading a small founding team with four members, you have to worry about your direct relationship with the three other cofounders, plus their direct relationships with one another. Combinatorial mathematics tells us that this means you need to manage the relationships between six pairs of individuals ([4*3] / 2). Now imagine that you hire two employees, for a total team size of six. Now you need to manage the relationships between fifteen pairs ([6*5] / 2). You increased the team size by 50 percent, but the number of relationships you need to manage went up by 150 percent. The math just gets more daunting from there. And that only considers the relationships of individual pairs of team members, not the relationships between any three members, any four members, and so on.
One approach is to design a business model that requires as few human beings as possible. Some software companies employ business models that allow them to achieve massive success with minimal numbers of employees. The founders of WhatsApp, Jan Koum and Brian Acton, designed a clever business model that addressed some of the key growth factors (their messaging service leveraged both classic network effects and the existing distribution network of telephone address books to grow faster) but also managed to skirt around issues of operational scalability. WhatsApp had a freemium business model; the service was free for a year, after which it cost $1 per year. This low-friction model essentially eliminated the need for people working in functions like sales, marketing, and customer service, allowing WhatsApp to grow to five hundred million monthly active users by the time of its acquisition by Facebook, with a staff of just forty-three employees, a ratio of over ten million active users per employee!
Another approach is to find ways to outsource work to contractors or suppliers. Airbnb’s strategy for photographing its hosts’ rooms offers an instructive example. Early on, Airbnb’s founders discovered that one of the key factors that increased the chances of renting a room on Airbnb was the quality of the photographs of that room. It turns out that most of us aren’t professional photographers, and our poorly composed, poorly shot cell phone pictures don’t do a good job of conveying the awesomeness of our living spaces. So the founders took to the road, visiting hosts and taking photographs for them. Obviously, personally visiting every host was hardly a scalable solution, so the the task was soon outsourced to freelance photographers. As Airbnb grew, the strategy shifted from the founders managing a short list of photographers, to an employee managing a large group of photographers, to an automated system managing a global network of photographers. Founder Brian Chesky describes this strategy succinctly: “Do everything by hand until it’s too painful, then automate it.”
Ultimately, even with clever business models and automation, nearly every massively successful company requires thousands or even tens of thousands of employees. Smart techniques can delay the reckoning, but not forever. Later on, we’ll discuss some of the management innovations that allow companies to handle this kind of organizational growth and scale.
The other main challenge of operational scalability comes from the strain of scaling up the nonhuman infrastructure of the business. It doesn’t matter how much demand you generate if your infrastructure can’t handle it. Infrastructure limitations can even be fatal to a company’s ambitions. Consider the examples of the social networks Friendster and Twitter.
While many have forgotten it now, Friendster was the first (pre-Facebook) online social network to break through into the mainstream (disclosure: I was an early investor in Friendster). Launched in March 2003, Friendster rode viral growth to millions of users within months. Before the year was out, Friendster-mania was such a cultural phenomenon that founder Jonathan Abrams appeared on the late-night television program Jimmy Kimmel Live! But Friendster’s massive growth brought massive headaches, especially on the infrastructure side. Despite a talented technology team, Friendster’s servers couldn’t handle the growth, and it became common for Friendster profiles to take up to forty seconds to load. By the beginning of 2005, a faster new entrant, MySpace, was generating more than ten times the number of pageviews as Friendster, which never recovered. MySpace, of course, ultimately lost the consumer social networking war to Facebook, which is a story we’ll discuss in detail later in this book.
Twitter came close to melting down in the same way, but managed to recover in time to build a massive business. When Twitter began its rise in the late 2000s, it became infamous for its “Fail Whale,” a whimsical error message that appeared whenever its servers couldn’t handle the load. Unfortunately for Twitter, the Fail Whale made fairly regular appearances, especially when big news hit, such as the death of the recording artist Michael Jackson in 2009 (to be fair, Twitter was hardly the only website that had these issues when the King of Pop passed away) or the 2010 World Cup. Twitter invested serious resources into rearchitecting both its systems and its engineering processes to be more efficient. Even with this strenuous effort, it took several years to “tame” the Fail Whale; it wasn’t until after Twitter made it through the 2012 US presidential election night without melting down that the company’s then–creative director Doug Bowman announced that the Great Blue Whale had been put to death.
One of the main reasons for the very large increase in the growth of valuable Web companies that we’ve seen in recent years is Amazon’s cloud offering, Amazon Web Services (AWS), which has helped many such businesses navigate around infrastructure limitations. Dropbox, for example, was able to scale up its storage infrastructure much more quickly and easily because it used AWS storage, eliminating the need to build and maintain its own arrays of hard disks.
AWS reflects one of the ways that Amazon has made operational scalability a competitive advantage. Web services like AWS tap into what Harvard Business School professor Carliss Baldwin and former Harvard Business School professor Kim Clark refer to as “the power of modularity.” As Baldwin and Clark describe in their book, Design Rules, Vol 1: The Power of Modularity, this principle makes it possible for a company like Amazon and its customers to build complex products out of smaller, standardized subsystems. But the power of modularity goes beyond just software development and engineering. By building easy-to-integrate subsystems like payments and logistics, Amazon makes its entire business more flexible and rapidly adaptable.
The equivalent to AWS on the hardware side is China. Hardware start-ups are able to manage infrastructure limitations and scale much more quickly by tapping into Chinese manufacturing capabilities, either directly or by working with companies like the custom manufacturing design firm PCH. The smart thermostat maker Nest, for example, had only 130 employees when it was acquired by Google for $3 billion, largely because it had outsourced all of its manufacturing to China.
In contrast, Tesla Motors has seen its growth held back by infrastructure limitations. Due to the complexities of its manufacturing process, Tesla’s production rates have lagged behind those of other automakers, the result being that its award-winning vehicles are almost always sold out, with back orders measured in months and even years. Demand generation is not a problem for Tesla; meeting that demand is.
Whether by design or not, the business models of rapidly growing companies often follow proven patterns that tap into growth factors and bypass growth limiters. These patterns will be described in more detail below, but here it bears noting that these high-level patterns are principles rather than exact recipes. Simply adopting any of these particular patterns isn’t enough to ensure an innovative business model, but understanding them does provide an entrepreneur with a set of good role models.
It is also worth mentioning that not all patterns are created equal. Some common business models follow proven patterns, but nonetheless don’t seem to produce $100 billion businesses or even $10 billion businesses. Take open-source software, which has been wildly successful as a pattern for spreading software products like Linux. Open source, which means offering free, community-created software that users can modify, arose to prominence during the dot-com era and has been an integral part of the world’s technology stack ever since.
The story of open-source software fits the pattern of business model innovation. Open-source software serves a large market, has powerful distribution via open-source software code repositories, benefits from the network effects of standards and compatibility, and neatly avoids many of the human limitations on operational scalability by tapping into a distributed community of volunteer contributors rather than building a large organization of employees.
Yet even the most successful open-source business, Red Hat, has a market capitalization of “only” about $15 billion, and that’s after being in business for two decades. The empirical evidence suggests that open source is a pattern that is valuable for engagement but not for building a massively profitable business.
In order for a pattern to be proven, it must be able to demonstrate that multiple massively valuable businesses follow it. Based on that criterion, we’ve assembled the following list of proven patterns to help inspire your own business model innovation.
Google and Facebook are largely software businesses that focus on electronic bits rather than material atoms. Bits-based businesses have a much easier time serving a global market, which in turn makes it easier to achieve a large market size. Bits are also far easier to move around than atoms, so bits-based businesses can more easily tap into distribution techniques like virality, and their ability to be highly networked provides more opportunities to leverage network effects. Bits-based businesses tend to be high-gross-margin businesses because they have fewer variable costs.
Bits also make it easier to design around growth limiters. You can iterate more quickly on software products (many Internet companies release new software daily) than on physical products, making it faster and cheaper to achieve product/market fit. And bits-based businesses, as we saw with WhatsApp, can get away with far fewer employees than most of their atom-based counterparts.
Back in 1990, the futurist George Gilder demonstrated his prescience when he wrote in his book Microcosm, “The central event of the twentieth century is the overthrow of matter. In technology, economics, and the politics of nations, wealth in the form of physical resources is steadily declining in value and significance. The powers of mind are everywhere ascendant over the brute force of things.”
Just over twenty years later, in 2011, the venture capitalist (and Netscape cofounder) Marc Andreessen validated Gilder’s thesis in his Wall Street Journal op-ed “Why Software Is Eating the World.” Andreessen pointed out that the world’s largest bookstore (Amazon), video provider (Netflix), recruiter (LinkedIn), and music companies (Apple/Spotify/Pandora) were software companies, and that even “old economy” stalwarts like Walmart and FedEx used software (rather than “things”) to drive their businesses.
Despite—or perhaps because of—the growing dominance of bits, the power of software has also made it easier to scale up atom-based businesses as well. Amazon’s retail business is heavily based in atoms—just think of all those Amazon shipping boxes piled up in your recycling bin! Amazon originally outsourced its logistics to Ingram Book Company, but its heavy investment in inventory management systems and warehouses as it grew turned infrastructure limitations from a growth limiter to a growth factor. On the retail side, merchants pay Amazon to manage their inventories and logistics for them, while the massive computer systems that Amazon built to operate its retail business gave it the capabilities to launch its AWS business (which is a high-margin, bits-based business!).
Platform economics predates the Networked Age, and even the Industrial Age. Trade-oriented principalities like the Republic of Venice provided a welcoming ecosystem for merchants, complete with currency and the rule of law, as well as taxes to harvest the value of the platform. Technology platforms like Microsoft Windows demonstrated the power of being the chosen platform on which businesses were built back when the World Wide Web was still a glimmer in Tim Berners-Lee’s eye (Sir Berners-Lee wrote his proposal for a global hypertext system in 1989). Yet despite the proven value of platforms in the pre-Internet era, the Networked Age has made them vastly more powerful and valuable.
Rather than being limited like the Republic of Venice to a specific geography, today’s software-based platforms can achieve global distribution almost immediately. And since transactions on today’s platforms are conducted through application programming interfaces (APIs) rather than person-to-person negotiations, they proceed swiftly, seamlessly, and in incredible volumes, all with barely any human intervention.
If a platform achieves scale and becomes the de facto standard for its industry, the network effects of compatibility and standards (combined with the ability to rapidly iterate and optimize the platform) create a significant and lasting competitive advantage that can be nearly unassailable. This dominance lets the market leader “tax” all the participants who want to use the platform, much as levies were imposed in the bygone Republic of Venice. For example, the iTunes store takes a 30 percent share of the proceeds whenever a song, a movie, a book, or an app is sold on that platform. These platform revenues tend to have very high gross margins, which generate cash that can be plowed back into making the platform even better. Amazon’s merchant platform, Facebook’s social graph, and, of course, Apple’s iOS ecosystem are great examples of the power of platforms.
“Free” has an incredible power that no other pricing does. The Duke behavioral economist Dan Ariely wrote about the power of free in his excellent book Predictably Irrational, describing an experiment in which he offered research subjects the choice of a Lindt chocolate truffle for 15 cents or a Hershey’s Kiss for a mere penny. Nearly three-fourths of the subjects chose the premium truffle rather than the humble Kiss. But when Ariely changed the pricing so that the truffle cost 14 cents and the Kiss was free—the same price differential—more than two-thirds of the subjects chose the inferior (but free) Kisses.
The incredible power of free makes it a valuable tool for distribution and virality. It also plays an important role in jump-starting network effects by helping a product achieve the critical mass of users that is required for those effects to kick in. At LinkedIn, we knew that our basic accounts had to be free if we wanted to get to the million users we theorized represented critical mass.
Sometimes you can offer a product for free and still be profitable; in the advertising-driven business model, a large enough mass of free users can be valuable even if they never pay for your service. Facebook, for example, doesn’t charge its users a dime, but it is able to generate large amounts of high-gross-margin revenue by selling targeted advertising. But sometimes a product doesn’t lend itself to the advertising model, as is the case with many services used by students and educators. Without third-party revenue, the problem with offering your product to users for free is that you can’t offset your lack of sales by “making it up in volume.”
Here is where the innovation of freemium comes in. The venture capitalist Fred Wilson coined the term in a 2006 blog post (based on a suggestion from Jarid Lukin), but the business model itself predates the term, having its origin in the “shareware” model for selling software in the 1980s. The free product was a tool for discovery and gaining a critical mass of users, while the paid version of the software allows the business to extract value from those users once its value is clear. Dropbox is one of the premier examples of a successful freemium business—by giving away 2 GB of storage, Dropbox attracted a massive user base, a reasonable percentage of which decides to pay for the value and convenience of additional storage.
Marketplaces represent one of the most successful business model patterns, with the dot-com era’s Google and eBay and today’s Alibaba and Airbnb standing out as examples of important, valuable companies that follow this pattern. One reason marketplaces are powerful is because they often tap into two-sided network effects. While it is difficult to create a successful marketplace from a cold start, the first marketplace that does manage to achieve liquidity—the ability for buyers and sellers to quickly and efficiently find a counterparty to conduct a transaction—becomes very attractive to both sides of the market. As buyers and sellers pour in, the marketplace becomes even more attractive to both parties, triggering a positive feedback loop that makes it very hard for new entrants to win any market share.
Marketplaces also offer key advantages beyond the obvious network effects. By creating a liquid market where buyers and sellers both participate, the dynamic forces of supply and demand price their transactions better than any human judgment could. The more efficient the prices in a marketplace, the more value it creates, because that means more transactions that might create value actually occur. In contrast, in illiquid markets, sellers often misprice their products, resulting in fewer sales and less value creation than optimal.
The best example of the benefits of efficient market pricing is probably Google’s AdWords advertising marketplace. AdWords allows anyone to bid on targeted keywords, in any quantity, so even the smallest businesses can tap into global distribution. Contrast this to the traditional advertising market, in which large clients spend millions of dollars paying advertising agencies to run expensive thirty-second television ads during coveted programming like the Super Bowl broadcast. Google’s system also measures advertising quality; ads targeted at its audience to generate the most paid click-throughs are favored. The net effect is that consumers are shown the most effectively targeted ads, without the overhead of a middleman like Don Draper and his three-martini lunch. Google also increases its own gross margin, because, unlike commercials during a television broadcast, search-based ad space is virtually unlimited and costs Google next to nothing.
Although marketplaces, even local ones, have always been a powerful business model, the changes ushered in by the Networked Age have made them potentially more valuable than ever. But unlike a local market with its size constraints—think of an old-fashioned bazaar in the center of a populous city—online marketplaces tap a global market. And by connecting buyers and sellers instead of holding inventory or managing logistics (and thus dealing in bits rather than atoms), online marketplaces avoid many of the growth limits of human or infrastructure scalability.
When Salesforce.com first launched its on-demand customer relationship management product, there were many legitimate questions about this new software-as-a-service (SaaS) model. Selling software as a subscription, delivered via the Internet, represented a major departure for enterprise software vendors. The previous model of selling permanent licenses for on-premise software and charging for maintenance provided more cash up front than monthly or annual subscriptions. The personnel required to support the model were also different; selling and supporting on-premise software required field salespeople and sales engineers to install pilot deployments, while the new SaaS model required additional staff to provide 24/7 data center coverage and support.
As it turns out, of course, SaaS eventually became the dominant business model for enterprise software. The cash flow disadvantages and required personnel shifts were real concerns, but mainly for existing players in the market. New SaaS businesses like Salesforce.com and Workday were designed and built around the new model, giving them a major advantage over existing players who tried to convert their on-premise software businesses to subscription ones.
Subscription Internet services have been successful because the sales and delivery model provides a larger market size and better distribution than traditional packaged software. Due to the cost and overhead of the extensive field operations required to support on-premise software, traditional enterprise software licenses had to be in the six- or seven-figure range simply to make the model work. This meant that software vendors focused on the needs of only the largest customers.
In contrast, Salesforce.com and other SaaS vendors can sell software licenses in any quantity, not only to Fortune 500 companies, but also to midmarket and small to medium-sized businesses, significantly enlarging their potential market. Internet delivery and self-service allow new forms of distribution that weren’t possible in the packaged software world, such as Dropbox’s viral incentive of additional free storage for referring new customers.
Nor is the pattern of Internet subscriptions limited to enterprise software. The dominant players in both music (Spotify, Pandora) and video (Netflix, Hulu, Amazon) also enjoy lower overhead and greater distribution by using the subscription business model.
Another, less obvious benefit to this model is that once a subscription business achieves scale, the predictability of its revenue streams allows it to be more aggressive with long-term investments, since it isn’t obliged to maintain large cash balances to weather short-term variations in the business. This financial firepower can represent a major competitive advantage. For example, Netflix, which announced plans to invest $6 billion in original content for its streaming service in 2017, has exploited its direct subscription model to outspend classic television networks, which have to rely on less robust revenue streams like payments from cable providers and advertising sales.
One of the emerging patterns that build on new platforms and services is the business of selling digital goods. Sitting at the intersection of “bits rather than atoms” and platforms, digital goods are intangible products that, arguably, have no intrinsic value—but they can still make for a profitable and scalable business. For example, the messaging service LINE derives significant revenues by selling “stickers”: images that are incorporated into the text of smartphone messages. In 2014, its first year of operation, LINE’s sticker business generated $75 million in revenue. That figure grew to $270 million in 2015, which represented over a quarter of LINE’s total revenues. Not bad for an intangible product with no intrinsic value!
Digital goods have also become a key business model in the video game industry, with in-game purchases of digital items that can help players advance in the game or advertise their status. Market-wide revenue from in-app purchases are projected to outstrip paid-app downloads in 2017, $37 billion to $29 billion.
In addition to enjoying the advantages of any bits-based business, digital goods tend to have nearly 100 percent gross margins, since they are purely digital and usually do not add significantly to infrastructure or overhead costs.
One of the most underrated and underappreciated proven patterns is the news feed. Facebook’s powerful network effects allow the site to attract its users, but its innovation of the news feed has made it a world-class business. Yet Facebook is hardly the only feed-centric success story. Companies like Twitter, Instagram, and Slack have all built multibillion-dollar market values around the news feed pattern.
The power of the news feed comes from its ability to drive user engagement, which in turn drives both advertising revenue and long-term retention. As Facebook has demonstrated, a news feed with sponsored updates is the most effective way to monetize proverbial Internet “eyeballs.” Facebook’s News Feed’s dominance of the online advertising market is only exceeded by Google’s AdWords, and AdWords starts with the significant built-in advantage of capturing active consumer intent rather than simply the desire to be amused. For example, how many people visit Facebook with the intention of going shopping? The magic of the news feed model has been its ability to monetize bored people catching up on what their friends are doing.
Of course, effective use of the news feed model requires a lot of sophisticated technology. Facebook doesn’t just insert sponsored updates at random. The company knows your interests better than you do, based on all the items you’ve ever clicked on, liked, or otherwise engaged with. It can carefully target the advertisements it shows you based on your individual habits and the context of what surrounds them in your feed. This targeting ability explains why Facebook succeeded in monetizing this model when other feed-based products like RSS readers failed.
This pattern is so powerful that Twitter, whose product is essentially one long news feed, is still an important Internet company despite barely changing its product in nearly a decade (going from 140 characters to 280 characters doesn’t count). Twitter is a business that scaled massively because of the power of business model innovation, not product or technology innovation.
Underlying the proven patterns of business model innovation are larger principles that can help refine those patterns or even create new ones. These principles aren’t themselves business models, but they often power the technological innovation that enables business model innovation.
Moore’s Law is the fundamental principle that puts the “Silicon” in Silicon Valley, and has powered the worldwide ascent of the technology industry. Moore’s Law is named after its codifier, Intel cofounder Gordon Moore, who coined the term in a paper he wrote in 1965, observing that the number of transistors that could be crammed onto the surface of a silicon chip appeared to double each year. While Moore revised his eponymous law in 1975 to a doubling of transistors every twenty-four months, the industry has since settled on a broad consensus of eighteen months. Today, Moore’s Law no longer refers specifically to transistor density; rather, it predicts that computing power tends to double every eighteen months. In recent years, this growth in computing power has been driven by the transition to multicore, multithreaded computing. Perhaps in the future, Moore’s Law will be met by quantum computing, optical chips, the use of DNA, or something even more impossible to foresee. The point is, it appears that the true limit to Moore’s Law is human engineering ingenuity, not solid-state physics.
Moore’s Law matters because the relentless increase in computing power that it predicts acts as a constant source of technological innovation, which, as we have seen, can help enable business model innovation. For many years, the power of Intel’s central processing units (CPUs) was measured by their “clock rate”—the number of times per second that the CPU could perform an operation. While clock rate is no longer a good measure of computing power, it is still a good metaphor for how Moore’s Law drives the world of computer technology: each tick of the clock enables new technologies, driving faster and faster innovations.
Increasing computing power allowed the shift from gigantic mainframes to smaller minicomputers to personal computers, all the way to today’s smartphones and wearables. We’ve seen similar increases in things like network bandwidth, allowing the Web to shift from text to images to audio to video, and in the future, 3-D and virtual reality (VR). Yet today’s smartphones aren’t simply smaller versions of IBM mainframes—remember, technology innovation enables business model innovation.
The best entrepreneurs don’t just follow Moore’s Law; they anticipate it. Consider Reed Hastings, the cofounder and CEO of Netflix. When he started Netflix, his long-term vision was to provide television on demand, delivered via the Internet. But back in 1997, the technology simply wasn’t ready for his vision—remember, this was during the era of dial-up Internet access. One hour of high-definition video requires transmitting 40 GB of compressed data (over 400 GB without compression). A standard 28.8K modem from that era would have taken over four months to transmit a single episode of Stranger Things. However, there was a technological innovation that would allow Netflix to get partway to Hastings’s ultimate vision—the DVD.
Hastings realized that movie DVDs, then selling for around $20, were both compact and durable. This made them perfect for running a movie-rental-by-mail business. Hastings has said that he got the idea from a computer science class in which one of the assignments was to calculate the bandwidth of a station wagon full of backup tapes driving across the country! This was truly a case of technological innovation enabling business model innovation. Blockbuster Video had built a successful business around buying VHS tapes for around $100 and renting them out from physical stores, but the bulky, expensive, fragile tapes would never have supported a rental-by-mail business.
(As hard as it may be for some readers to comprehend, when we were in college, we would often drive to a Blockbuster Video store on a Friday or Saturday night, pay a couple of bucks to rent a VHS tape of a movie, and use a landline telephone to call Domino’s to order a pizza before popping the videotape into a VCR that was connected to a twenty-five-inch standard-definition cathode-ray tube.)
DVD technology allowed Netflix to create a completely new business model. Rather than renting out individual movies and being charged exorbitant late fees if they failed to return the VHS tape in time, Netflix customers paid $20 per month for a subscription to “unlimited” movies—provided they checked out just one movie at a time. This allowed Netflix to eliminate Blockbuster’s widely loathed late fees and capture the powerful and certain revenue stream from the proven model of a subscription service. Netflix took off, and even went public as a DVD-by-mail service.
But Hastings never lost sight of his ultimate vision for Netflix—on-demand television delivered via the Internet. And as Moore’s Law continued to work its magic, making computers ever more powerful and Internet bandwidth ever greater and cheaper, Netflix bided its time, waiting for streaming video to become viable.
“When we first started raising money in 1997, we thought we’d be mostly streaming in 5 years,” Hastings told us when he visited our Blitzscaling class at Stanford. “In 2002, we had no streaming. So we thought that by 2007, it would be half our business. In 2007, we were still nowhere. So we made the same prediction. And this time we were wrong the other way—by 2012, streaming was 60% of our business.” It may have taken longer than Hastings expected, but Moore’s Law eventually came through for him.
Today, Netflix is synonymous with television on demand delivered via the Internet, and it has created an entirely new category of “binge watching.” As of 2017, 53 percent of American adults say that their household has access to Netflix, and the service is growing rapidly across the rest of the world. Netflix has used the financial power of its subscription model to become one of the premier sources of original video content, from television shows like Stranger Things, to movies like Beasts of No Nation, to events like comedian Dave Chappelle’s comeback stand-up comedy specials.
Traditional television commissions large numbers of pilot episodes, the majority of which never make it to series, trying to produce optimistically named “Must See TV” to appeal to a broad audience, which has to be convinced to tune in every single week. In contrast, the on-demand model allows Netflix to cater to many different audiences rather than program a small number of thematic channels, as cable television does. Broadcast television succeeded by providing the same thing to all its viewers—a model driven by the technological innovation of broadcasting content via wireless signals and later coaxial cable. Netflix succeeds by providing a carefully personalized experience to each of its many viewers, giving it a huge advantage over its traditional television competitors. Moreover, Netflix produces exactly what it knows its customers want based on their past viewing habits, eliminating the waste of all those pilots, and only loses customers when they make a proactive decision to cancel their subscription. The more a person uses Netflix, the better Netflix gets at providing exactly what that person wants. And increasingly, what people want is the original content that is exclusive to Netflix. The legendary screenwriter William Goldman famously wrote of Hollywood, “Nobody knows anything.” To which Reed Hastings replies, “Netflix does.” And all this came about because Hastings had the insight and persistence to wait nearly a decade for Moore’s Law to turn his long-term vision from an impossible pipe dream into one of the most successful media companies in history.
Moore’s Law has worked its magic many other times, enabling new technologies ranging from computer animation (Pixar) to online file storage (Dropbox) to smartphones (Apple). Each of those technologies followed the same path from pipe dream to world-conquering reality, all driven by Gordon Moore’s 1965 insight.
Blitzscaling companies use automation. If they have the ability to perform a task (which is a big if), computers are almost always faster, cheaper, and more reliable than human beings. Furthermore, computers continue to get faster and cheaper, doubling in power every eighteen months according to Moore’s Law, as opposed to human beings, who evolve over the course of millions of years according to Darwin’s principle of natural selection.
In 2014, the journalist Jan Vermeulen compared the original Apple II (introduced in 1977) with the then state-of-the-art iPhone 5S. He found that in the intervening thirty-seven years, Apple’s products had become 2,600 times faster in terms of clock speed (from a 1 MHz single-core CPU to a 1.3 GHz dual-core CPU) and had 16,384 times the amount of RAM. That’s three to four orders of magnitude of improvement in the span of a single human generation. And that massive delta doesn’t even take into account that the Apple II was a desktop computer with a bulky cathode-ray tube monitor, and the iPhone 5S was a portable supercomputer that people carried in their pockets.
The same year that the Apple II was introduced, Joe Bottom set a world record by swimming the 50-meter freestyle in 23.74 seconds, for a brisk pace of just under 7.6 km/h (4.7 mph). If human swimming speed had increased as quickly as the computing speed in Apple’s products, the world record in 2014 would have been 19,700 km/h (12,250 mph)—not quite enough to achieve orbital velocity, but about twenty-five times the speed of the average commercial jetliner. The actual human world record for the 50-meter freestyle in 2014 was 20.91 seconds, for a more modest 11 percent improvement.
That’s the power that automation taps into.
The power of automation applies not just to direct-to-consumer products like the iPhone but also to internal processes and capabilities. Think of the value that automation creates by increasing the productivity in Amazon’s warehouses, or by making it easier to keep Google’s server farms running 24/7.
At a higher level of abstraction, successful scale-ups place more emphasis on adaptation than optimization. Rather than the giant assembly lines of Detroit automakers, which trace their origins to Henry Ford’s Model T, the current generation of Silicon Valley companies practice continuous improvement, whether through an emphasis on speed or the constant experiments and A/B testing of growth hacking. This emphasis makes sense in an environment where companies need to seek product/market fit for new and rapidly changing products and markets. Consider how Amazon expanded into new markets like AWS rather than simply honing its retail capabilities, or how Facebook has been able to adapt to the shift from a text-based social network accessed via desktop Web browsers to an image- and video-based social network accessed via smartphones (and soon, perhaps, VR).
My friend Peter Thiel has written eloquently about the power of being a contrarian in his book Zero to One.
Whenever I interview someone for a job, I like to ask this question: “What important truth do very few people agree with you on?”
This question sounds easy because it’s straightforward. Actually, it’s very hard to answer. It’s intellectually difficult because the knowledge that everyone is taught in school is by definition agreed upon. And it’s psychologically difficult because anyone trying to answer must say something she knows to be unpopular. Brilliant thinking is rare, but courage is in even shorter supply than genius.
Being contrarian is often critical to the process of creating a massively valuable technology company. As we’ve discussed, key growth factors like distribution and network effects tend to provide disproportionate rewards to a company that is the first in its space to achieve critical scale. Being contrarian and right gives you a huge advantage because you get a head start on achieving scale.
If your company is pursuing an opportunity that nearly everyone agrees is very attractive, you’re likely to have a difficult time distancing yourself from your army of competitors. But if your company is pursuing an opportunity that conventional wisdom ignores or disdains, you will probably have the time you need to refine your business model innovation into a well-oiled machine. Amazon pursued e-commerce when most people didn’t think consumers would feel comfortable using credit cards online. Google launched its search engine when most people thought search was a mature commodity. And Facebook built its social network when most people believed social networking to be either useless, a market dominated by MySpace, or both.
As we’ve already seen, most great ideas look dumb at first. Being contrarian doesn’t mean that dumb people disagree with you; it means that smart people disagree with you! Remember what happened when Brian Chesky, Joe Gebbia, and Nathan Blecharcyzk tried to pitch Airbnb? Investors like Paul Graham literally couldn’t imagine why people would ever use the service. This doesn’t happen because investors are dumb; most venture capitalists and angel investors are smart, and most smart, successful people would probably agree that investing in proven ideas is better than investing in unproven ones.
The problem is that, by definition, business model innovation involves trying something that is new, and thus unproven!
In this book, we’ve tried to lay out a set of tools, principles, and patterns that you can use to design, invest in, or evaluate an innovative business model. Many venture capitalists like to brag that they are masters of “pattern matching”—but here we must caution not all pattern matching is helpful. The bad kind of pattern matching is what B- and C-grade investors love—the Hollywood high-concept pitch. The movie Speed was famous for its high-concept pitch: “Die Hard on a bus.” And if you’re the first person to make the connection, you might succeed. Speed was in fact a commercial success, mostly because it did in fact live up to its description. But the success of Speed led to a raft of derivative and inferior movies, ranging from Steven Seagal’s Under Siege (“Die Hard on a ship”) to Steven Seagal’s Executive Decision (“Die Hard on a plane”). When an investor funds “Uber for Pets,” that’s bad pattern matching.
The good kind of pattern matching involves understanding what medical science terms “the mechanism of action.” Speed works because confining the action to a bus that has to stay at a certain speed or higher to avoid setting off a bomb creates built-in dramatic tension—especially given the famously bad traffic in Los Angeles. Airbnb works because it has a large market, because travelers spreading awareness from city to city creates virality, and because it follows the proven pattern of an online marketplace.
To help you get a feel for applying our principles of business model innovation, let’s practice by analyzing some of today’s great businesses and how they follow those principles.
When we started LinkedIn in 2002, the recent dot-com bust had led most people to consider the consumer Internet industry to be dead. The last thing venture capitalists were willing to do was provide millions of dollars to fund rapid growth. Despite this fact, I thought there was a big opportunity available, and was able to guide LinkedIn through the start-up growth phase until we could raise the capital to really blitzscale.
This is the story of how it happened.
The key insight behind LinkedIn was that the Internet was shifting from anonymous cyberspace to an extension of the real world, and thus your online identity was an extension of your real identity. Readers of my generation might remember the famous New Yorker cartoon with the caption “On the Internet, nobody knows you’re a dog.” I didn’t think this kind of anonymity would work in a professional context, hence the need for a professional online identity. And though our thesis was contrarian at the time, my cofounders and I were fairly confident that the market of “all white-collar professionals” was sufficiently large to represent a major opportunity.
In order to raise money to scale LinkedIn, we had to find a way to prove our distribution strategy. Unfortunately, investors thought of us as “Friendster for business relationships,” which was bad pattern matching and made about as much sense to them as “Tinder for business relationships” would to today’s VCs. Instead, we had to find a way to use the money and reputation I had acquired by helping build PayPal to get LinkedIn to the point where people would invest.
The first step was to assemble a small, super-scrappy team. We got our first office by squatting in the building of a friend’s failing start-up. “Just clean up after yourselves so we can get the lease deposit back, and you can use it for three months,” he told me. I leveraged my reputation to secure a small investment, but I knew we needed to show significant progress in distribution before we could raise our next round. Since we didn’t have the capital to pay for traditional marketing, we implemented a number of techniques similar to what people today call “growth hacking” to get to one million users, which allowed us to raise money from Greylock.
Our core distribution strategy was organic virality, much as it had been at PayPal. Our users would invite their contacts via e-mail because it helped them build their networks and keep track of their key connections. But the initial level of virality simply wasn’t enough. We couldn’t offer PayPal’s kind of financial incentives, so instead we built things like the e-mail address book importer so that we could increase the number of invitations and let our users know when their contacts joined the service.
Gross margins were important because it became apparent that our user growth was always going to be surpassed by that of the leading consumer social networks. At this point, MySpace had eclipsed Friendster, and Facebook was quickly gaining on MySpace—and all of them had far more users than LinkedIn. Our argument was that our professional users were far more valuable, but to prove that argument we had to demonstrate our ability to earn significant high-margin revenues.
The first business model pattern we tried was a freemium subscription service. The free LinkedIn.com service limited the number of requests a user could send to friends of friends (InMails), and when users hit those limits, they would be offered the chance to upgrade to a premium subscription. This subscription revenue was enough to get us to cash-flow profitability, but it wasn’t growing fast enough to be truly compelling.
The key inflection point came when we discovered that companies were willing to pay for the ability to scan LinkedIn profiles to find the best job candidates. So we offered it to companies as an enterprise subscription product, and once we proved that this new model was a source of significant high-gross-margin revenues, we had the confidence to blitzscale.
The long-term value of LinkedIn was always intended to come from network effects. As a professional social network, LinkedIn leveraged both direct and two-sided network effects, as well as becoming a standard format for presenting one’s professional identity. The direct network effects come from the fact that each additional LinkedIn user makes the network slightly more valuable to all other LinkedIn users. The two-sided network effects occur because more users attract more corporate employers, while more employers increase the value of LinkedIn as a passive job-hunting tool. Finally, by becoming an integral part of most people’s professional online identities, LinkedIn has become a standard that has largely replaced the traditional résumé. Just one of these network effects would probably be enough to create first-scaler advantage; all three working together built a massive strategic moat that protected the LinkedIn business from any new entrants, and even from attempts by consumer networks like Facebook to take away the professional market.
Finding product/market fit for our enterprise product was the key inflection point in the business. How did we do it? We focused on getting market feedback as quickly as possible. We hired a salesperson, gave him some mock-ups of an enterprise product, and sent him to visit potential customers. It turned out that they all wanted to buy it!
Blitzscaling LinkedIn presented two major operational scalability challenges, beyond the obvious one of supporting a global social network with hundreds of millions of users. First, to support the business, we actually had to develop, maintain, and update two different products. Without the consumer product, companies wouldn’t see the value of our enterprise product. Without the enterprise product, we couldn’t make enough money to build a great business. We had to do both. It’s hard to find an engineering expert who would recommend fracturing your product and engineering group to work on two largely separate products, but that’s precisely what we did, despite the inefficiency and messiness.
Second, we had to rapidly scale a salesforce while we were still developing the product they were selling. This took a lot of hard work on the part of LinkedIn’s CEOs, Dan Nye and then Jeff Weiner, and their teams. But where we could, we also used technology to help alleviate scaling constraints. Our “Merlin” tool helped make our salespeople more productive (and thus scalable) by automating much of their manual work. Merlin would analyze usage patterns and tell each salesperson which companies to call, how they were already using LinkedIn, and even create a personalized sales deck for each individual prospect!
Jeff Bezos’s original vision for Amazon was to take advantage of unlimited digital shelf space to run a store where a customer could buy literally anything. Amazon began with books because this represented a large enough market with a product amenable to e-commerce (durable, fairly standard sizes, readily available through wholesale distributors). Since then, Amazon has steadily expanded from books into many other verticals, and today very nearly lives up to Bezos’s vision of an “everything store” (though you still can’t buy automobiles on Amazon…yet). Retail is a truly gargantuan market and Amazon has captured an almost unthinkable portion of it and even made its market much bigger by launching Amazon Web Services. Now, in addition to being “the everything store,” Amazon also provides much of the Internet’s computing power, bandwidth, and storage (including for other dominant companies like Netflix).
Amazon was one of the first companies to fully grasp the possibilities of the Internet as a distribution platform in creating the first successful affiliate program, Amazon Associates, which incentivizes individuals and owners of other websites to refer customers to Amazon in exchange for a share of the revenues generated. This allows Amazon to turn everyone else’s websites and online communications into a powerful distribution channel. Even today, if you see a book title on the Internet, or in a tweet or an e-mail signature, and you follow that link, you’ll probably find yourself on Amazon’s website via an affiliate link.
Amazon actually scores fairly poorly on this growth factor, though this is largely a function of the industry rather than being specific to Amazon. Retail is a relatively low-margin business, and Amazon’s devotion to offering low prices further hurts margins. Even today, Amazon’s retail business isn’t profitable (though it probably could be if the health of the company required it; for example, Amazon’s core North American operations are profitable—it’s just that its profits are outweighed by the losses generated by Amazon’s efforts in Asia).
Yet even within Amazon’s retail business, we detect signs that these low gross margins are actually part of a long-term strategy that can generate high gross margins, even on retail sales. It’s no secret that Amazon dominates e-commerce; in 2017, analysts like Slice Intelligence reported that Amazon accounted for 44 percent of US e-commerce sales in 2016, and predicted the figure would be even higher in the future. But what is often overlooked is that Amazon’s retail business consists of two very different units. The first is Amazon’s traditional retail operation, in which Amazon buys products from suppliers and sells them to its customers. The second, far less well-known unit is Amazon’s marketplace, which lets third-party sellers sell their products on Amazon. Those third-party sellers store their inventory in Amazon warehouses and pay Amazon to deliver their products to their customers. If you’ve ever shopped on Amazon, you’ve probably bought a product from a third-party seller; Jeff Bezos has said that almost 50 percent of units purchased on Amazon come from them. Because this marketplace business doesn’t require tying up Amazon’s capital in inventory (it ties up the third-party sellers’ capital instead), its gross margins likely resemble high-margin eBay’s more than it does low-margin Walmart’s. As Benchmark’s Matt Cohler notes, “I sometimes wonder if Amazon’s owned-inventory business is just a marketing loss leader and a capital-intensive competitive moat.”
Where Amazon is already tapping into high gross margins is with its AWS business. Remember, 150 percent of its operating margins in 2016 came from AWS, which accounted for $12.2 billion in revenue and over $3 billion in operating income. The high gross margins of AWS allow Amazon to invest heavily in maintaining its lead over its competitors. Indeed, AWS is estimated to hold over 40 percent of the market for cloud computing infrastructure, more than its three biggest rivals—Microsoft, Google, and IBM—put together!
Amazon is relatively weak on network effects. One customer’s use of Amazon doesn’t make it more valuable for another customer, with the possible exception of Amazon’s product reviews. Yet whatever direct network effects exist because of product reviews pales in comparison to the impact of network effects on something like Facebook. Amazon technically is a marketplace with two-sided network effects, thanks to its third-party sellers, but one side is largely missing: Amazon sellers are attracted by Amazon’s massive customer base, but Amazon’s customer base is largely indifferent to those sellers. Amazon does benefit from scale effects, and explicitly uses the “flywheel” framework of author and strategy guru Jim Collins. Brad Stone summarized this approach in his book on Amazon, The Everything Store:
Lower prices led to more customer visits. More customers increased the volume of sales and attracted more commission-paying third-party sellers to the site. That allowed Amazon to get more out of fixed costs like the fulfillment centers and the servers needed to run the website. This greater efficiency then enabled it to lower prices further. Feed any part of this flywheel, they reasoned, and it should accelerate the loop.
Yet as impressive as Amazon’s flywheel is, when compared with the powerful superlinear effect of most network effects, it is merely linear or sublinear. Fortunately, Amazon does benefit from strong network effects in one of its units.
Most of Amazon’s network effects, like most of its gross margins, come from its AWS business. The AWS platform benefits from both indirect network effects and compatibility and standards. The success of AWS encourages developers and development products like Docker to rely on it as their infrastructure of choice, which makes AWS even more successful (while the emergence of AWS as a standard makes it easier for services built on the platform to connect via API).
Amazon has rarely struggled with product/market fit in its core business. For the most part, because it was tapping into an existing—and thriving—retail market, Amazon was able to leap into hypergrowth almost immediately. Even AWS met with rapid uptake, helped by Amazon’s savvy decision to lead with its simplest product, S3 (Simple Storage Service), before expanding to more complicated ones. It is important to remember that Amazon has had many failures outside its core business. Amazon’s powerful core retail operations didn’t allow it to take over auctions or payments from eBay or PayPal, and its Fire Phone was a costly and fruitless attempt to take on Apple and Android.
Amazon has managed operational scalability so well that it might be the best in the world at this task.
On the human side, Jeff Bezos has been able to guide Amazon with a strong and steady hand while allowing business leaders like Andy Jassy, the CEO of AWS, or Jeff Wilke, the global head of the consumer business, to run large portions of the company. This delegation has allowed Amazon to grow to over 541,900 employees as of 2017, making it one of the ten largest employers in the United States.
On the infrastructure side, Amazon has deftly shifted from minimizing infrastructure spending, as it did during its early years by using techniques such as outsourcing logistics to book distributors like Ingram, to becoming one of the world’s great infrastructure companies. Amazon is so good at infrastructure that its fastest-growing and most profitable business (AWS) is all about allowing other companies to leverage Amazon’s computing infrastructure. Amazon also makes money by offering Fulfillment by Amazon to other merchants who envy its mastery of logistics, which ought to strike fear into the hearts of frenemies like UPS and FedEx. In addition to its eighty-six gigantic fulfillment centers, Amazon also has at least fifty-eight Prime Now hubs in major markets, allowing it to beat UPS and FedEx on performance by offering same-day delivery of purchases in less than two hours. Amazon has also built out “sortation” centers that let it beat UPS and FedEx on price by shipping small packages via the United States Postal Service for about $1 rather than paying FedEx or UPS around $4.50.
Google’s market size was dramatically underestimated at the outset. When Google came on the scene, many considered it “yet another search engine” in a market that was already dominated by companies like Yahoo! and Lycos. Even in the unlikely event that Google was able to capture a significant share of the search market, it would still be a niche player in comparison to, say, Yahoo!, which was a portal with major properties like Yahoo! Mail and Yahoo! Finance.
Observers failed to realize two things. First, Google’s business model innovation—the relevance-based, revenue-maximizing, self-service advertising system of AdWords—allowed it to generate far more revenue per search than its predecessors. Second, the importance of search was growing at a faster rate than the Internet as a whole. As the Internet grew and the amount of content increased at a superlinear rate, so did the difficulty of filtering and finding relevant information, making search increasingly important. Combine that effect with the rapid growth of the Internet itself, and the result was a massive market.
Google has astutely expanded the market since then by leveraging the power of its business model to make and monetize key acquisitions like Android, Google Maps, and YouTube.
Google’s technology receives most of the credit for the company’s success, and it is impressive. However, this means that Google’s skillful use of the distribution growth factor is often overlooked.
To go from “yet another search engine” to “the last search engine” (as my old friend Peter Thiel put it in his 2014 Stanford lecture “Competition Is for Losers”), Google had to leverage a series of existing networks and partners. For example, Google’s bold deal to power AOL’s search results helped the company grow its search business by orders of magnitude. Later, other distribution bets like the Firefox partnership, the acquisition of Android, and the creation of the Chrome browser all paid off and helped maintain Google’s distribution dominance.
Google also found ways to leverage small partners as well, with its AdSense program for Web publishers feeding more raw traffic into the AdWords machine.
Google is a phenomenally profitable company, with an enviable margin of 61 percent in 2016. But this profitability didn’t happen by accident or luck; the credit belongs to Google’s AdWords business model. As we discussed in our section on business model patterns, the advertising-supported media model hasn’t worked for the Internet. Yet when Google first emerged, this was the dominant business model being pursued by major players like Yahoo! and Lycos. Google adopted the self-service advertising auction model of Overture, added its own refinement of selecting ads based on considerations of relevance and quality as well as bid prices, and pursued a business model of capturing purchase intent rather than just gathering eyeballs. This purchase intent proved to be far more valuable per unit of traffic, enabling Google to earn fat margins.
Google has since used the financial power of its gross margins to place big bets that other companies might shy away from, such as investing in Android and Chrome, two products that were going up against dominant competitors (Apple’s iOS in mobile phone software and Microsoft and Firefox in Web browsers). Google has also used its margins to fund radical experiments like X (formerly Google X) and Waymo (self-driving cars). These bets may or may not pay off, but even if they fail, Google’s margins give it the ability to recover quickly and keep going.
Google has leveraged network effects quite a bit in its major business lines, though not, ironically enough, in its core search product!
The mobile traffic app Waze is a classic example of a direct network effect. Waze harnesses each user’s location to create a more accurate model of traffic conditions, while also letting drivers easily report events such as traffic accidents, speed traps, and stopped cars on the side of the road. Then Waze makes all that data public to everyone using the app. In other words, the more Wazers on the road, the more accurate that road information becomes. Each additional user creates value for all the previous users.
The Android mobile operating system is a classic example of indirect network effects. Its broad adoption by end users increases the incentives for developers to create Android versions of their applications. The increased availability of useful apps encourages more people to use devices that run on the plaftorm.
YouTube is a classic example of two-sided network effects. YouTube brings together video creators and consumers—the more content is created, the more people show up to consume it. The more consumers who show up, the more incentive there is to create content.
Finally, Google’s G Suite provides a great example of the power of compatibility and standards (ironically enough, much like Microsoft Office, its archrival) as well as local network effects. When users share Google Docs or Google Sheets with others, they lock in anyone who wants to collaborate on those documents to do the same. This is especially common in individual networks like a project team or school. Once some of the school’s teachers start using Google Docs for homework assignments, the pressure builds for all of them to standardize on Google Docs, and for children and parents to adopt it as well. Chris speaks from experience here.
Google got the product/market fit for its core search and AdWords product incredibly right. Even from the start, Google’s search results were better than those of its competitors. But many people don’t realize that it actually took Google a long time to find the right product for the right market. Google started off trying to sell enterprise search appliances, a tool that sits inside a corporate data center, indexing content stored on a company’s servers, then offering a Google search box to find items within that content. Next, Google tried the advertising-supported model by running DoubleClick ads; ironically enough, Google would later buy DoubleClick. Fortunately, Google found product/market fit by refining Overture’s advertising auction model. Google’s AdWords product was so much better at monetizing search through its self-service, relevance-driven, auction system that by the time those competitors managed to play catch-up, Google had amassed the financial resources that allowed it to invest whatever was necessary to maintain product superiority.
Google doesn’t always get product/market fit right (and if it had run out of money before hitting upon AdWords, the search business might have died before ever achieving that fit). This is a reflection of its very intentional product management philosophy, which relies on bottom-up innovation and a high tolerance for failure. When it works, as in Gmail, which was a bottom-up project launched by Paul Buchheit, it can produce killer products. But when it fails, it results in killed products, as demonstrated by projects like Buzz, Wave, and Glass. To overcome this risk of failure, Google relies on both its financial strength (which comes from its high gross margins, among other things) and a willingness to decisively cut its losses. For example, when Google bought YouTube (which had clearly achieved product/market fit), it was willing to abandon its own Google Video service, even though it had invested heavily in that product.
Other massively successful companies take a very different approach. In contrast to Google, where new ideas can come from anywhere in the company and there are always many parallel projects going on at the same time, Apple takes a top-down approach that puts more wood behind fewer arrows. Apple keeps its product lines small and tends to work on a single major product at a time. One philosophy isn’t necessarily better than the other; the important thing is simply to find that product/market fit quickly, before your competition does.
Unsurprisingly for an engineering-driven organization, Google excels in operational scalability. For one thing, its heavy investment in its own tools and infrastructure has allowed its engineering organization to fine-tune its infrastructure for high performance as the company has grown.
Google has innovated in people scalability as well. While most of Google’s people management practices are smart but relatively straightforward—for example, Google uses smaller teams to work on new products and larger teams to sustain and grow existing products—Google has invested heavily in people analytics and data to determine things like the optimum number of interviews per candidate (no more than five) and to improve practices for recruitment, performance reviews, and so on.
Market size is one of the key reasons that many failed to appreciate the potential value of Facebook in its early days. At the time, the elevator pitch for Facebook would have been “social network for college students.” This description, which combined a new and unproven product category with a specific (and narrow) audience, made Facebook sound like a niche product. But by the time I invested in Facebook, Mark Zuckerberg’s vision was far broader and more valuable. Mark wanted Facebook to be the default way that people stayed in touch with their friends, which was and is an enormous market. Of course, even when Mark pitched his broader vision, many investors didn’t believe him, much to their later regret.
Facebook excelled at distribution. As noted earlier, Facebook’s early focus on college students, which caused some to dismiss it as a niche product, was actually part of an extremely successful distribution strategy. To achieve incredible virality, Facebook would deliberately delay launching at a college campus until over 50 percent of the students had requested it so that local critical mass was reached almost immediately.
Facebook further benefited from leveraging existing friend networks to expand outward from its original college user base. As users experienced the benefits of staying connected via Facebook, they naturally wanted to add their off-line friends to the network.
Like Google, Facebook started its life without an effective revenue model. But once it discovered the value of sponsored posts within a news feed, Facebook was able to become wildly profitable. About 90 percent of Facebook’s revenue today comes from advertising sales, and the company achieves an astounding 87 percent gross margin.
This gross margin allows Facebook to invest heavily in talent and technology. It has also allowed Mark Zuckerberg to make canny (and expensive) acquisitions, like Instagram and WhatsApp, to become a dominant player in mobile as well as desktop social networks, and also long-term future bets like Oculus.
We’ve already talked about how Facebook leverages classic direct network effects (the more users that join the platform, the greater the value of Facebook to every other Facebook user) and local network effects (once it becomes the dominant social network at a college, it becomes extremely difficult for any other player to pry away Facebook’s users).
Facebook also experiences some helpful indirect network effects thanks to its platform services, such as the Graph API (which allows developers to leverage the Facebook social graph of users and their relationships) and Facebook Connect (which allows users to log in to a Web service using Facebook rather than create a new account for that service).
Facebook achieved product/market fit for its core consumer experience almost immediately, hence its rapid growth. However, part of what makes Facebook a great company and Mark Zuckerberg a great CEO is that Facebook has been able to achieve product/market fit in additional and less obvious areas at other points in the company’s history.
Many people forget how Facebook struggled with the transition from desktop to mobile. Facebook’s initial mobile product provided a slow, suboptimal experience, and adoption of that product was accordingly slow. Fortunately for Facebook, Mark Zuckerberg saw that the market was going mobile and put a moratorium on new feature development in order to focus the entire team on building a new, far superior mobile product. In parallel, he also moved quickly and decisively to acquire Instagram and WhatsApp; when they were announced, both acquisitions were considered pricey, but in hindsight they were clearly bargains. Today, Facebook has over 1.7 billion active mobile users each month, and mobile advertising accounts for 81 percent of the company’s advertising revenue. Over 56 percent of Facebook users access the service exclusively via mobile.
Equally important was Facebook’s ability to achieve product/market fit for its advertisers. When Facebook began, the conventional wisdom was that user-generated content like Facebook would never be able to attract advertisers, who would not want their brands appearing with poor-quality or even inappropriate content. Google’s search model was what worked in online advertising. Facebook was able to overturn the conventional wisdom by developing algorithms to block inappropriate content, and by learning from Twitter’s sponsored update model and incorporating ads into the Facebook News Feed. The news feed model has been especially effective for monetizing mobile usage. In a return to what worked in the print world, advertisements are intermixed with content, and as you page through the magazine or scroll through the feed, you encounter advertisements as part of your normal flow, as opposed to the interruptions of pop-up or takeover ads, or the easily ignored static placement of the traditional banner ad. Yet Facebook’s News Feed is even better for advertisers than a magazine, because Facebook’s core social actions (clicking, liking, sharing) train users to engage with whatever appears in the News Feed, including advertisements!
How did Facebook successfully overcome the growth limiter of operational scalability? On the technology side, one of the philosophies that helped Facebook become successful was its famous motto “Move fast and break things.” This emphasis on speed, which came directly from Mark Zuckerberg, allowed Facebook to achieve rapid product development and continuous product improvement. Even today, every new software engineer who joins Facebook is asked to make a revision to the Facebook codebase (potentially affecting millions or even billions of users) on his or her first day of work. However, as Facebook’s user base and engineering team grew to a massive size, Mark had to change the philosophy to “Move fast and break things with stable infrastructure.”
While this new motto might seem self-contradictory, Mark explains that it focuses on a higher-level goal. “The goal is to move fast,” Mark told me. “When we were smaller, being willing to break things allowed us to move faster. But as we grew, the willingness to break things actually started slowing us down, because increasing complexity made it harder and harder to fix things once they broke. By taking the extra time to focus on stable infrastructure, we reduce the impact and time to recover from breaking things, so that we can actually move faster.”
If you believe you’ve designed a business model that can support massive growth and value creation, the next step is to decide on your strategy. That’s where strategy innovation comes in.