فصل 3

Chapter 3

Strategy and Business Models

Innovation, Transaction, or Hybrid

When Jack Ma was building Alibaba’s Taobao in China in 2003, he had a simple problem to solve: He needed to match buyers with sellers and figure out how to make a profit. Ma could have copied eBay’s internationally successful strategy where sellers paid eBay a fee. Instead, he decided to create a marketplace where neither sellers nor buyers would have to pay a fee for completing transactions. Although this strategy would evolve over time, in the early days active sellers on Taobao could choose to pay to rank higher on the site’s internal search engine, generating advertising revenue for Alibaba.

Now compare Jack Ma’s problem with Google’s challenge of building an ecosystem around Android. At the time of Android’s launch in 2008, Nokia, Apple, and BlackBerry dominated the smartphone business. Google needed to encourage software developers to create new games, applications, and content for its operating system, which was incompatible with its competitors. It needed to convince developers that this investment would be worth the risk. Google made the decision easier by giving the operating system away for free, licensing an open-source version, and providing software tools to allow external developers to develop new apps with less effort.

Both companies had to launch their platforms, but the specifics of how to launch and whom to attract required different decisions, strategies, and actions. These differences exist because all platforms are not the same; they do not all create value or operate in the same way. To select the right platform strategy and business model, managers and entrepreneurs should start with the value proposition they envision. If the value will come mainly from enabling third parties to build their own products or services that utilize and enhance the platform, then you should develop an innovation platform. If the value will come mainly from allowing different sides of a market to interact, then you should develop a transaction platform. Successful companies also have a natural tendency to grow and expand, which often leads platform businesses to adopt a hybrid strategy. For example, firms that start with a successful innovation platform tend to add a transaction side or a separate transaction platform (usually a marketplace). Firms that start with a successful transaction platform tend to add an innovation side or a separate innovation platform. Hybrids also differ to the extent that they integrate the two different types of platforms or keep them largely separate, much like a digital conglomerate.

In this chapter, we discuss the four strategic steps required to build an innovation platform or a transaction platform. The steps are the same, but we explain how execution and business models differ for each platform type. We also explain the different types of hybrid strategies and why hybrids are emerging as the most powerful and highly valued platform business model.

The Four Steps to Building a Platform

All firms must go through the same four steps when they try to create and sustain a successful platform business. Figure 3-1 summarizes the steps.

The first step is to identify the various market sides you want for your platform, and how to create value through them: the role different actors play (buyers, sellers, or complementors) and who specifically will take on these roles. Second is to launch the platform, which requires solving the chicken-or-egg problem of how to get started and then how to attract increasing numbers of users or complementors in order to generate strong network effects. Third is to design a business model that will monetize those network effects without depressing them. This monetization challenge involves identifying where the cash flow and profits will come from, and choosing which sides, if any, will benefit most from subsidies. Fourth is to establish and enforce rules of conduct for ecosystem governance: In other words, managers and entrepreneurs need to decide what behaviors to encourage or discourage on the platform, and how to enforce the rules. We will have more to say on platform and ecosystem governance in Chapter 6.

1. Choose the Market Sides of Your Platform

Choosing who should and should not participate on a platform is a strategically crucial decision. While it might appear obvious in retrospect, the choice of which sides to serve requires creative thinking. Some firms are overly ambitious and try to connect too many sides too early in their development. The result is often an overly complicated platform, which does not scale. One such example was Brightcove, a media platform that serial entrepreneur Jeremy Allaire founded in 2004. Brightcove started off with no fewer than four sides (content providers, advertisers, consumers, and syndicate affiliates). The broad platform quickly proved unworkable for many market participants. Other common mistakes include failing to identify the side of the market that will attract other sides, mispricing on the more attractive side, and entering a market too late. We discuss these and other causes of platform failure in Chapter 4.

INNOVATION PLATFORMS

Innovation platforms, such as Microsoft Windows, Google Android, Apple iOS, and Amazon Web Services, offer technological building blocks that third-party innovators use to develop new complementary products or services. The building blocks usually include tools and connectors that facilitate the creation of complements such as software applications for computers and smartphones. At least one side of an innovation platform always consists of complementors, and at least one other side always consists of end users.

The key for successful innovation platforms is to identify complementors that will stimulate demand for the platform by making new products and services that add significant value. Of course, we don’t always know in advance which companies or individuals will create those innovations. Innovation platforms (especially digital ones) often try to solve this problem by broadly exposing their application programming interfaces or APIs (meaning they release information and allow access to the platform’s internal instruction sets and communication protocols). They also actively encourage third parties to design complementary products or services, such as through sending out free software development kits and organizing developers forums, or creating incubators and venture funds that subsidize new developers of complementary innovations. Sometimes the platform company will build a few complements itself. Especially when a platform is new, delivering your own complements early in the process helps to stimulate demand for the platform’s newest version.

Even with ample interest from third-party complementors, hurdles remain. Complementors often need technical and financial support from the platform owner. An important challenge for some platforms, especially new platforms, is the difficulty of identifying precisely which complementors to support. For example, Numenta—an artificial intelligence platform founded in 2005 by Jeff Hawkins and Donna Dubinsky—struggled for ten years trying to figure out who were the most promising complementors. Despite an abundance of inquiries, it was never clear what might become the “killer app” for this platform and who would build it. Numenta licensed its technology to some organizations and the open-source software community, but it has remained small, with few active complementors and only around two dozen employees.

TRANSACTION PLATFORMS

Transaction platforms, such as Amazon Marketplace, Google Search, Facebook, Alibaba’s Taobao, Uber, Lyft, and Airbnb, are online marketplaces that enable the exchange of goods, services, and information. They can help people or organizations access and use assets such as cars or rooms, but also connect through social media, and attract advertisers. As our colleagues David S. Evans and Richard Schmalensee have emphasized, they act as matchmakers. For most transaction platforms, the choice of sides has been rather obvious. In the cases of eBay, Amazon Marketplace, and Etsy, the sides simply consisted of buyers and sellers of physical goods. For Upwork (formerly Elance-oDesk), the sides were buyers and sellers of freelance labor. For Airbnb, the sides consisted of individuals with accommodations to rent and people who want to rent accommodations.

While most transaction platforms start with just two sides, they often add other sides over time. The question of how many sides to add and when to add them is strategically important, and its answer may not be so obvious before the fact. For example, most of us take for granted that advertisers subsidize Internet search, but this was not always the case. Inktomi and AltaVista were first-generation search engines that did not rely on advertisers, nor did they monetize their search. Most of them failed as businesses. Google was initially no different: It provided search results for free and generated no revenue. Although Google developed a superior search technology, the management team’s crucial decision was to turn its search engine into a transaction platform by bringing advertisers on board in a clever way. Identification of the appropriate market sides and how best to engage with each side was essential to Google’s financial success.

Google positioned Internet searchers on one side and advertisers on the other. To users of the search function, Google’s high-quality PageRank algorithm provided a valuable service. It had an ingenious design that crawled the web and reverse engineered the links connecting web pages to each other. But Google could have chosen not to have two sides and instead monetize search directly by charging users a flat subscription fee or a per-search usage fee. If it had done this, then Google Search would have emerged as an entirely different (and probably much smaller and less profitable) business. By adding advertisers to the other side, Google Search turned into a global mass-market transaction platform. By allowing advertisers to have their ads prominently located on the screen next to the search results, Google created obvious value to advertisers. It also connected them intelligently using its AdWords technology with the very users who were searching a topic close to the object of their ads.

Social networks such as Facebook, Twitter, LinkedIn, Snapchat, Instagram, and Tinder provide another set of examples. Like Google Search, we consider them transaction platforms because they facilitate exchanges of information among users who otherwise would have difficulty connecting. These platforms also generate and rely on network effects. But how many market sides do social networks really need? They are one-sided platforms if all users are similar, and multisided if they can distinguish different categories of users. But the criteria for identifying sides differ for each type of platform and application. Sides can be men and women on Tinder when individuals use the platform to look for people of the opposite gender. For men looking for men, or women looking for women, Tinder is more like a one-sided platform. Some social networks, like eHarmony, charge for access, like social clubs in the physical world or buyer clubs like Costco. They could sell services, such as enhanced mail, which LinkedIn provides as a premium service. Or they could bring in an advertising complement side, as Facebook and Twitter have done. Other transaction platforms such as eBay in its early years and WhatsApp (now owned by Facebook) chose not to add an advertising side.

When bringing new sides on board, timing matters a lot. Bringing in advertisers too early might damage the user experience and depress growth of the user base, whereas a small number of users may not provide enough value to attract advertisers. This was why Facebook did not start with advertising. Mark Zuckerberg’s first task was to grow the number of users.

Sometimes firms experiment and change their strategy about adding sides. For instance, LinkedIn in 2007 envisioned a side of “experts” and planned to launch a research service in 2008. The aim was to provide access for a fee to institutions or individuals who were experts in specific areas. LinkedIn eventually dropped the plan because of limited interest.

1. Launch: Solve the “Chicken-or-Egg” Problem

Launching a platform and solving the chicken-or-egg problem is probably the most difficult challenge for platform strategists. When side A’s volume depends on side B, and side B’s volume depends on side A, how do you get started? Here again, innovation and transaction platforms need to approach this dilemma differently. Strategic choices generally fall into three categories: (1) Create stand-alone value for one side first, (2) subsidize one or both sides, and (3) sometimes bring two sides on board simultaneously.

INNOVATION PLATFORMS

Creating stand-alone value for one side—such as users—requires that the firm produce a strong product or service that does not initially need third-party complementary innovations. Third-party innovations may make the product even more valuable, and this is when a product can evolve into an innovation platform, provided that it has the right attributes. First, the product must be designed in such a way that it has “hooks” for outside firms to connect (like application programming interfaces for software platforms). Second, it must also be modular enough in design for outsiders to add significant innovations. Third, the company must facilitate easy access to the product’s core functionality, such as through inexpensive or free licensing terms.

One strategy for launching an innovation platform in a market where no platforms exist yet is to identify an industry-wide problem. Then offer your product as a solution to that problem, or at least as a “core” or essential ingredient for the solution. In prior writings, we called this strategy “coring.” There are several examples. In order to solve the problem of how to build an IBM-compatible personal computer during the 1980s, which IBM tried to control as a proprietary product and platform technology, Microsoft and Intel broadly licensed or sold the core ingredients, the MS-DOS operating system (followed by Windows) and the x86 microprocessor. Similarly, in the late 2000s, Google offered the Android operating system as a solution for handset makers that wanted to build smartphones with capabilities similar to Apple’s iPhone. ARM also made its microprocessor design the technology of choice for companies that wanted to build smartphones that used relatively little battery power.

When it comes to generating network effects, the chicken-or-egg problem for innovation platforms comes down to two questions: How can a platform owner make it attractive for potential customers to buy the platform even if there are few complementary applications? And how can a platform owner persuade complementors to invest in platform-specific innovations if there is uncertainty about the number of end users who are willing to buy the platform (and the complements)?

The general principle of “Get the engine running and build up some momentum” in order to launch applies to all platforms. What is unique about how complementors engage with innovation platforms is that complementors are not only adopting a platform: They also are acting as technology suppliers and innovators in their own right. When facing a decision to adopt an innovation platform and join the ecosystem, complementors must trust someone else’s technology (the platform) to develop their own new product or service. At the same time, the owner of the innovation platform must entice complementors to innovate on the platform so that the platform becomes increasingly useful and valuable.

Innovation platforms can solve this chicken-or-egg problem by developing or buying some of their own complements. They can also provide free or inexpensive tools and technological assistance to help accelerate third-party innovations. Apple, for example, launched the iPhone and the iPad with a few bundled applications that it developed in-house, including a web browser (Safari), Mail, Photos, Video, iTunes, Notes, Contacts, and Calendar. It also obtained a few other apps from key external content providers, such as Google Maps and the New York Times, ensuring the provision of early complements.

By maintaining secrecy prior to launch and then staging mega‒media events, Steve Jobs sought to stimulate maximum exposure for new products such as the iPhone and the iPad. This marketing strategy attracted interest not only from users, which constituted one side, but also from content providers and app developers, which constituted the other side.

TRANSACTION PLATFORMS

For most transaction platforms, solutions to the chicken-or-egg problem should be relatively straightforward: How can platform owners make certain there are enough buyers to attract sellers? And how can the platform deliver enough sellers to attract buyers?

When Brian Chesky and Joe Gebbia launched Airbnb in 2007, they decided that the first platform side to build up should be property owners with places to rent; in other words, build up the supply side. Their first challenge was to identify these property owners and rally them in large enough numbers to attract renters. Chesky and Gebbia had the clever idea to avoid starting from scratch, and instead used readily available information on property owners who wanted to rent out their properties. Where could they find this kind of information? A large number of owners had already posted their properties on a popular online classified website, Craigslist. Airbnb founders developed a piece of software that hacked Craigslist to extract the contact information. They also took advantage of newspaper ads and other public postings.

Identifying members for their first side was just the beginning. The founders then had to encourage the property owners to post their listings on the new platform. Airbnb’s value proposition was clear: While requiring no further investment from them, posting their property on Airbnb (in addition to Craigslist) would simply increase the quantity of the property owners’ exposure to possible renters. In addition to increased exposure, Airbnb also helped the property owners increase the quality of their exposure by hiring professional photographers to take photos similar to what a person would see for hotel listings. This decision was not only a differentiating factor from Craigslist-type postings or newspaper advertisements; it also increased the renters’ perception of the value they would get from the Airbnb platform. Renters began to see that they were getting hospitality services comparable to hotels, and usually for a significantly reduced price or in more convenient locations.

The way Airbnb attracted early members was ingenious and added value for both market sides, but this strategy also differed from how Airbnb expanded. For Airbnb to send photographers to each new member’s residence was not easily scalable and financially unsustainable. More important, it was not necessary to continue subsidizing professional photo shoots. By initially subsidizing professional photos, Airbnb set a high bar for the quality of property photography on the Airbnb website. This practice soon became the new norm: Property renters would invest on their own in professional photography to differentiate themselves from other renters. In effect, Airbnb initially incurred a cost that raised expectations within the ecosystem and later simply took advantage of competition among the property owners.

There are three lessons we can learn from Airbnb. First, for many transaction platforms, there may be no need to start from scratch. Instead, platforms should try to make use of existing groups and information, such as by aggregating and analyzing publicly available data. Second, once the platform identifies its different sides, it can offer a service that helps attract members to the platform. And third, platforms may contradict the conventional wisdom for new ventures that says, “Start as you mean to go on.” The launch phase can benefit from actions that just aim at getting started—even if they are not financially sustainable or practically scalable. In short, it is fine to start off in a direction that the platform cannot follow in the long term if it can ignite self-sustaining feedback loops along the way.

In order to launch, transaction platforms usually solve their chicken-or-egg problems in one of two ways: (1) Pick one side and build it up, and then, once that side is sufficiently populated, bring on another side; or (2) bring on both sides at once, little by little, in a zigzag fashion. Pick one side and build it up can work well because, once a single side is populated, it should attract the other side. Indeed, many platforms subsidize one side to get the ball rolling. The Yellow Pages, eBay, Etsy, Amazon Marketplace, Taobao, Facebook, Twitter, and Airbnb all subsidized one side of their platforms to get momentum started. The subsidy usually came in the form of no charge to be a user. A variation on this strategy consists of the platform subsidizing activities that help members become more attractive to another side, as Airbnb did at first when sending professional photographers to property renters.

Another tactic was offering stand-alone value to one side, much like the coring strategy we talked about for innovation platforms. If the platform could make a service valuable or even essential to one side’s members, irrespective of the number of users on the other side, then it becomes easier to start the ball rolling. OpenTable (which Priceline purchased in 2014 for $2.6 billion) did this when it signed up its first restaurants by offering a very useful computerized table management system and charging only a small monthly fee for the technology. Once restaurants were on board with this solution to their table management problem, it was easier to attract them as users of the reservation platform.

Operating a one-sided platform and growing it (prior to bringing another side on board) can also succeed if members of the first side find value in communicating or interacting with each other, as they do with the telephone as well as social networks. Once that first side gets big enough, then it can attract other sides. This is what Facebook did when it launched the social network within the Harvard student community as a one-sided platform, and then gradually expanded to other universities and then to anyone over age thirteen who wanted to join. Only when the user side was very large, numbering in the millions, did Facebook focus on bringing in other sides such as advertisers, application developers, and digital content partners. This two-step strategy combined tapping first into the power of direct network effects, which helped develop and expand a single side of the network. Then, in a later stage, other sides joined the platform when it was more likely to generate cross-side network effects.

When platforms try to bring on both sides at once, they often have to subsidize both sides at the same time. But subsidizing both sides at once is a costly and risky strategy. It makes most sense only if three conditions prevail: (1) The platform has extremely deep pockets. (2) The platform has a realistic chance at a winner-take-all-or-most outcome. (3) Once all or most competitors have exited, there are barriers preventing new players from entering and then customers from switching or multi-homing.

Still, there are two problems with this approach. First, as we discussed in Chapter 2, winner-take-all-or-most outcomes are relatively rare. Second, the platform owner is betting that users and ecosystem partners will still want to remain affiliated with the platform, even if prices eventually go up. WhatsApp, for instance, offered a free service, in effect subsidizing all of its users. WhatsApp’s owner, Facebook, hoped that, once it reached a billion users, it could charge its customers $1 per year. But despite the very modest fee, WhatsApp quickly discovered that even $1 could lead customers to switch, causing Facebook to abandon its plan.

1. Design Your Business Model

There is another problem with launch tactics that heavily subsidize one or two sides. Not only is this expensive, but some platforms get stuck in an “infinite launching loop.” Platform owners keep burning money in order to gain adoption from multiple sides, heavily subsidizing participants for years. Uber has been the obvious poster child for this approach. Uber’s management and its investors seem to be hoping for a winner-take-all-or-most outcome, where competitors eventually exit due to price competition and financial losses. Uber management also seems to believe that a large number of drivers and vehicles will provide new opportunities for related diversification (such as Uber Eats—delivering food) and eventually turn into high and stable streams of revenue and profit.

This optimistic scenario may well occur for Uber. After all, Uber has delivered a great value proposition for consumers: Conventional taxis are in short supply and relatively expensive in major cities and urban areas around the world. Uber’s very low prices and convenient way to match driver supply with rider demand through a smartphone app has led to explosive growth for the ride-sharing service. But even if a platform has enough funding to overcome the chicken-or-egg problem, it does not guarantee that the business will produce more revenues than its expenses. This is especially true if the company has high employee turnover (the downside of “contract workers”) and repeatedly incurs start-up costs as it expands into new geographic areas or new business ventures, some of which it may not understand well. As a result, even with a clear identification of market sides and very strong direct and indirect network effects, platform companies can still find it difficult to identify a profitable business model. WhatsApp in 2018 had 1.5 billion active users who loved the service, but WhatsApp has not yet made a dime. All platform businesses need to find a way to extract significant value from at least one market side and eventually turn that value—fueled by network effects—into increasing revenues and, eventually, a profitable operation.

INNOVATION PLATFORMS

Most successful innovation platforms generate profit in one of two ways: (1) They increase their customers’ willingness to pay for the platform itself, such as by adding new features and encouraging third parties to create complementary products and services that enhance the value of the platform. Or (2) they capture value as a portion of the sale of every complementary product or service sold for the platform, including complements they build themselves. In the PC world, Microsoft made it free and easy for developers to write new applications, and the price of DOS and then Windows rose accordingly. Microsoft also built Office and many other software applications that complemented Windows and made sure there was demand for each new generation of the operating system. In the video game world, Sony collected a licensing fee on every game sold for PlayStation. For smartphones and tablets, Google gave away the Android operating system based on the (correct) theory that the company could make money on mobile search.

The goal for every innovation platform is to create scale: Most digital platforms have relatively high fixed costs (the ongoing R&D requirements to create new features) and relatively low or zero variable costs (distributing software or data over an existing network). The key is to attract enough complementors to grow the ecosystem and help grow the number of users, which attracts more complementors, and then users, and so on. For a software platform, this dynamic can enable the platform to literally “print” money at scale. Microsoft, for example, spent $1 billion developing Windows XP, but it sold roughly 250 million copies at $60 per unit in its first year. Microsoft broke even in three weeks and “minted money” thereafter, or at least until it started to invest in the next operating system. Other innovation platforms, such as Nokia’s Symbian, failed to attract enough complementary applications to its platform after the entry of Apple’s iOS and then Google’s Android. This situation led to a rapid loss in market share (negative network effects), a subscale business, and eventual failure (sale of the phone business to Microsoft). The lesson is that successful innovation platforms must continue to attract new complements, achieve and retain scale, and remain competitive in functionality lest they be overtaken.

TRANSACTION PLATFORMS

All transaction platforms generate revenue through fees for business deals, advertisements, or services. However, their business models vary in terms of who gets charged, what gets charged, and which services are free or subsidized. Since they may not be selling a product or service directly, transaction platforms must understand the willingness to pay of their various market sides, as well as the difference between how much each side wants the other side to engage. In theory, the demand characteristics from each side should be the main driver for the business model. In practice, it is very difficult for managers and entrepreneurs to assess the level of demand in advance. We don’t know of any silver bullet to answer the question of how much one side should be willing to pay, but we can offer some principles and illustrate them with examples.

Transaction platforms tend to offer value to their various sides and ultimately generate profit in five ways: (1) matchmaking, (2) reducing friction in transactions, (3) complementary services, (4) complementary technology sales, and (5) advertising. We will discuss each of these approaches.

First, as matchmakers, transaction platforms help provide access to a large pool of users and identify suitable matches. These users can be buyers and sellers of goods or services as in online marketplaces such as Amazon Marketplace, Etsy, eBay, or Alibaba’s Taobao and Tmall. They can be users who want to exchange information with each other, as in social networks, or exchange financial authorizations and authentication data, as in payment platforms. Much of the value that transaction platforms provide comes from increasing the size of the pool and then increasing the likelihood of a better match.

The extent to which side members are prepared to pay for being matched with another side varies from market to market. This difference explains variations in the pricing decisions of different platforms. Airbnb charges room renters and room providers but does not charge when they are simply room seekers or room listers. Tinder, the location-based meeting app, launched in 2013 as fully free but changed its pricing model in 2015 to operate a freemium model. Upwork experimented and changed its pricing model in 2015, charging less to companies that used its services for the largest jobs, hoping to stimulate more recurring business. And as we shall see later, some platforms charge both sides but offer some degree of subsidy at least to one side.

One of the worst platform business models to date has been Uber, despite the obvious value it offers to users in terms of convenience and low prices. Uber lost some $4.5 billion in 2017 on gross revenues of $37 billion. It also has lost more than $11 billion cumulatively. We expect the company to finish 2018 with large (although diminishing) losses. The 2019 IPO will bring in more cash and help Uber expand, but it has to overcome several cost drivers before it can earn a profit.

To compete with taxis as well as other companies such as Lyft, Uber subsidized the cost of rides, keeping prices artificially low. Uber also paid many drivers a set fee in addition to per-ride compensation or had to pay other financial incentives to attract and keep drivers on the platform, including help with financing the cost of their drivers’ vehicles. In addition, Uber advertised heavily to attract new riders. The largest cost came from driver turnover, called the “churn rate.” This number was an estimated 12.5 percent per month, or 150 percent per year, which means that Uber had to replace all its drivers every eight months, on average. Assuming that Uber had 3 million drivers in 2018, it would need to find 375,000 new drivers each month just to replace its existing workforce, not to mention grow the business. With driver advertising and recruitment costs estimated at $650 per driver, in 2018 this expense alone could total over $240 million per month and nearly $3 billion for the full year! And there was more. Another big cost driver was high R&D expenses, such as more than $2 billion in 2017. These were mainly to develop self-driving vehicles—a big bet to reduce those enormous driver costs but not likely to have much impact in the next few years. Meanwhile, Uber was diversifying into related transportation services, hoping to leverage its existing drivers and logistics capabilities by delivering food (Uber Eats) and packages (UberRUSH). These activities still operated at a loss in 2018, though they may eventually be more profitable than ride sharing, which has become a commodity business with intense price competition. Uber also plans to enter yet another transportation service. It will address the shortage of truck drivers by matching fully loaded trailers from big companies with drivers from small and medium-size carrier companies. In other words, Uber heavily subsidized both sides of its market, and had a potentially fatal flaw—very high driver turnover, the downside of using contract workers rather than potentially more stable regular employees. Consequently, the company was forced to spend billions of dollars in venture capital and operating cash flow simply to maintain the status quo. Rapid growth in existing as well as new markets then required additional expenses and investments. Overall, Uber has grown mostly because of generous capital providers (largely sovereign wealth funds and SoftBank’s Vision Fund). Investors seem to be betting on a winner-take-all-or-most outcome where Uber outlasts both digital and conventional competitors, and then eventually raises prices or reduces costly subsidies.

Second, transaction platforms “reduce friction” to facilitate interactions among platform participants. Various books and articles have introduced this concept, including Invisible Engines (2006) by Evans, Hagiu, and Schmalensee, Matchmakers (2016) by Evans and Schmalensee, and Platform Revolution (2016) by Parker, Van Alstyne, and Choudary. This is a helpful concept, and we often see transaction platforms doing things like enabling the secure exchange of goods or money. Many transaction platforms also used the payment services of other transaction platforms, such as Google Checkout or PayPal, or provided a financial “escrow” service to sellers. Other friction-reducing or risk-reducing services included authentication (verifying users’ identities), insurance, and, for cryptocurrencies such as Bitcoin, currency exchanges and virtual wallets. Third, transaction platforms often create additional value for their members by providing complementary services. Sometimes they offer these services for free, but most of the time they charge. For example, Taobao did not charge sellers and buyers to sign up, but it charged sellers for obtaining a better ranking in its internal search engine. Amazon offered, for a fee, ancillary services such as goods consignment (for sellers) and delivery for its Amazon Marketplace seller-users. The “Fulfillment by Amazon” (FBA) was a popular bundle of services that Amazon offered its marketplace users. Those who signed up for FBA outsourced to Amazon the consignment, packaging, and shipping of their products to customers. Another significant part of FBA’s value for sellers is that it offered the possibility to have all their goods delivered with the Amazon Prime service, which provided expedited and free delivery. This valuable feature would often lead customers to choose certain products over other similar products.

Some platforms charge both sides for complementary services, although the services still need to be valuable, scalable, and difficult to copy in order for the companies to charge enough to make a profit. Consider Deliveroo, the online marketplace for British restaurants launched in London in 2013. As of 2018, this platform offered fast delivery (on average thirty-two minutes) of local restaurant-cooked food in over two hundred cities on three continents, avoiding the highly competitive United States. In a 2015 interview, Deliveroo’s cofounder, William Shu, explained that a big misconception about his business was that people assumed his company was a pure online marketplace. If that were true, then perhaps Deliveroo could earn a profit by matching drivers with restaurants and keeping expenses very low. However, Shu noted that Deliveroo was really a logistics service. The company worked with restaurants that did not have existing delivery services and provided them with drivers and access to a logistics platform built using machine learning algorithms to optimize routing. The business model charged both sides of the platform—the restaurants and the food consumers. Deliveroo took a commission from restaurants and charged a delivery fee to the customer. It relied on courier riders who most often rode their own bicycles to deliver the food. It did not consider the bicycle riders employees but as “independent contractors,” similar to how Uber categorized its drivers. Deliveroo’s couriers were not eligible for workers’ benefits such as sick pay and the national living wage, nor were they covered by a company insurance policy in case they had an accident.

Deliveroo’s employment practices (we discuss these further in Chapter 6) kept manpower costs low, but it had high expenses for technology as well as administrative staff and executives. As of 2018, there were still not enough economies of scale, operational efficiencies, or value added for the company to make a profit. The company had raised just under $1 billion but lost $176 million in 2016, the last time Deliveroo reported financial data. Food delivery may be a $100 billion market, but delivery by itself was not a very sophisticated or high-value-added service. Deliveroo could not charge a lot or raise prices easily because there were different ways for restaurants to deliver food. Fourth, some transaction platforms sell technology or other goods and services apart from transaction fees. This approach seemed far superior to delivery services. We can see this with OpenTable, which matched people who want to make reservations with restaurants who accepted reservations online. It charged $1 for every reservation made through its system and 25 cents for reservations made directly on the restaurant’s website, which amounted to a relatively large monthly fee for using the service. Some competitors, such as Yelp, also charged for the service and for specific reservations made through the system. Fifth, selling advertising is one of the most widely used business models for transaction platforms. It is particularly well suited to search engines like Google or social networks like Facebook or other types of transaction platforms with a social component. For example, TripAdvisor has user reviews of travel services and places, and these constitute an important part of the value that users obtain when searching on TripAdvisor. But to maximize value to the platform, advertising cannot be seen as an unwelcome “tax” on users that unilaterally degrades their experience. For advertising to play a positive role in the business model, it has to add value to both sides, even if not to the same extent. Google has mastered this approach with its AdWords auction, an effective and innovative method to best match the search user’s interest with the relevance of the advertisements users see when they search. The rise of ad blockers was also starting to threaten a number of advertising-based business models, inspiring those platforms to look for other means of generating revenue.

1. Establish and Enforce Ecosystem Rules

Ecosystem rules pose a basic question for anyone affiliated with a platform: Who should be allowed to do what? When Armin Heinrich posted his app called “I Am Rich” on Apple’s App Store in 2008, and sold it for $999.99, the challenges of governing a platform became apparent to Apple. The app itself did nothing except show a glowing red stone with the text: “I am rich, I deserv [sic] it, I am good, healthy & successful.” Eight people bought it, and Apple made $2,400 in revenues (which was mostly profit). Obviously a scam of sorts, Apple had to decide how to govern its platform: Should it take the money and run, or establish and enforce standards for its end users? Apple chose to delete the app and create relatively strict standards that all app developers had to meet. As the following section suggests, Apple was not alone in struggling with the challenges of ecosystem governance.

INNOVATION PLATFORMS

Innovation platforms need to ensure the quality of complements and clarify who can connect to and innovate on top of the platform. Good ecosystem governance encouraged lots of innovation and allowed complementors as well as users to benefit in a sustainable manner. By “sustainable” we mean that platform companies and their complementors need to be able to compete effectively with rivals and make a profit. Some innovation ecosystems, such as in the software industry, revolved around de facto technical standards. These platforms also existed on a spectrum of technology openness. On one extreme, they were proprietary and tightly controlled or partially closed to third-party complementors (such as Apple’s iOS). On the other extreme, they were “open source” and not owned by any one firm (such as with Linux and some versions of Android, itself a version of Linux). Indeed, the primary governance issues for innovation platforms relate to key elements of their strategies: the degree of openness and the extent to which the platform company competes with its complementors.

In general, participants in many successful innovation platforms tend to view the platform owners as the legitimate custodians of the technical standards and the ecosystem more broadly. In particular, complementors feel they are part of a “collective,” with a shared destiny. But legitimacy as a platform leader takes time to build and is easy to destroy. Innovation platform owners need to be careful not to hurt complementors’ incentives to innovate by carelessly trampling over their businesses. This concern limits to some extent the freedom of platform owners to enter and compete directly in complementary markets.

Innovation platforms often competed with each other and sometimes with their complementors. It is well-known that Lotus, WordPerfect, and Harvard Graphics dominated the markets for DOS and Windows versions of spreadsheet, word-processing, and presentation software during the 1980s, until Microsoft introduced Excel, Word, and PowerPoint, then bundled them in the Office suite from 1990. Accordingly, ecosystem rules tended to evolve over time along with changes in competition and cooperation between the platform leader and external complementors. We observed a similar pattern when Google bought Motorola Mobility in 2012. The acquisition generated anxiety and mistrust among mobile handset makers, which were using Android as an operating system. They were concerned that Google would favor Motorola handsets with the newest implementations of Android, putting Google’s partners at a disadvantage. Despite Google’s attempts to reassure its partners by making explicit statements that it would not play favorites, important Android licensees hedged their bets. Samsung started to push some phones with its own Tizen operating system, and LG introduced phones with webOS (initially developed by Palm). In 2014, Google resolved the tensions by selling Motorola Mobility to Lenovo. However, Google reentered the space in 2018 when it bought part of the handset engineering group of HTC, the Taiwan manufacturer that had been producing Google-branded Pixel phones. Governance of innovation platforms becomes even more complicated when platform competitors are also platform complementors. Google, for example, was a leading provider of applications for the iPhone, with apps ranging from search to maps. On early models of the iPhone, Apple prominently displayed Google Maps on its home screen, providing prime “real estate” to Google. By 2012, however, Apple increasingly viewed Google as a competitor and not just a complementor. This led Apple to invest heavily in an internally developed competing product, Apple Maps. The Google Maps app was still accessible and downloadable from Apple’s App Store, but Apple decided not to preload it onto the main iPhone screen. This decision reflected and reinforced a change in the relationship between Apple and Google, away from collaboration and toward more competition.

One of the specific ways innovation platforms govern their ecosystems is through the design of the technological interfaces between the platform and its complements, and in particular through the degree of openness of these interfaces. When platform leaders open up their interfaces (such as APIs), they send a clear signal to complementors that they are trying hard to improve complementors’ businesses. When they close or restrict their interfaces, they send the opposite message. When platforms change the openness of their APIs without communicating effectively about the impact on the ecosystem, they weaken the shared sense of the collective, and this can create ripples of discontent throughout the ecosystem.

Twitter provides a good example of what can happen when the platform sends mixed messages. We think of Twitter as a social media transaction platform, but for several years it also operated as a hybrid, with a lively innovation platform side to the business. Yet Twitter greatly disappointed its community of complementors with its decision to restrict or shut down access to its APIs in 2012. When Twitter first opened its APIs in 2006, the free and easy access stimulated a vibrant community of developers for Twitter-related apps, such as to enable users to view Twitter on their mobile devices (called Twitter “clients”). By 2010, Twitter was widely seen as a developer-friendly service and even launched a site for developers containing Twitter API documentation, platform status information, and forums. Some 75 percent of all traffic on the platform came through the Twitter API. But once Twitter developed its own mobile app, following its acquisition of Atebits, a start-up that had developed the popular Tweetie client application, it started to discourage external developers from creating different Twitter clients. Ryan Sarver, then director of the Twitter platform, announced that developers should not build new Twitter clients and existing clients needed to strictly adhere to Twitter’s rules. The announcement also stated that “we need to move to a less fragmented world, where every user can experience Twitter in a consistent way,” and mentioned changes to the terms of service related to Twitter APIs.

From 2012, Twitter also altered the rules of connectivity and in effect started closing off its public API. For example, in 2012, Twitter blocked Instagram’s “Find Friends” feature, no longer allowing Instagram to access Twitter followers’ graph data via an API. It also imposed a similar restriction on Tumblr and started to limit the number of users that outside services were allowed to access. But after poor market performance and growing disaffection among outside developers, from 2015 Twitter started to reverse its course. Cofounder Ev Williams even stated in July 2015 that the company had made a strategic error when it came to the Twitter API. In October 2015, cofounder and CEO Jack Dorsey stated at the 2015 mobile developer conference that Twitter wanted to “apologize” for the “confusing” and “complicated” relationship the company had created with developers, and claimed it wanted to “reboot” and have a great relationship in the future. Consequently, after shutting down Twitter API access to Politwoops, an app that monitored the Twitter accounts of politicians by using the Twitter API to collect and display tweets, Twitter reversed its decision in January 2016 and reinstated Twitter API access. Twitter then started to look for other ways to increase user engagement on its platform in order to increase the number of hours per day users spend on Twitter. Higher levels of Twitter usage would enable the platform to monetize user activity further by selling more (or more expensive) advertisements.

Innovation platforms (including hybrids) that become dominant in the market face a unique challenge, akin to a winner’s curse. As they become increasingly essential, the expectations of users and complementors grow. People come to expect innovation platforms to behave in responsible ways and pursue the overall good of the ecosystem, rather than simply maximize profits for the platform owner. This is why, not surprisingly, we sometimes treat powerful innovation platforms as if they were public utilities. We discuss this issue further in Chapter 6.

TRANSACTION PLATFORMS

The ecosystem rules for transaction platforms are similar to innovation platforms: Ultimately a transaction platform has to decide who can connect through the platform and what the various market sides can do on the platform. In addition, most transaction platforms have rules aimed at minimizing low-quality transactions, such as weeding out poor-quality goods and services, facilitating returns in case of unsatisfied customers, and fighting fraud.

Some transaction platforms impose and check credentials before members can perform an activity. We call this active limitation of who can engage on the platform “curation.” For example, after a number of incidents, including the sexual assault of a passenger in India by the driver, Uber attempted to improve driver qualifications by checking licenses and performing background checks. Didi Chuxing did the same thing in China after several sexual assaults and murders of passengers by drivers. Platforms may also make their rules explicit in the form of “community guidelines.” Uber’s guidelines included general principles such as “Respect each other,” “Give riders and drivers some personal space,” and “No physical contact with the driver or fellow riders” (with the explicit mention of “As a reminder, Uber has a no sex rule”), as well as policies against “discrimination.” In a similar effort to fight (and to be perceived as fighting) discrimination, Airbnb required all members to approve the “Airbnb Community Commitment,” agreeing “to treat everyone in the Airbnb community—regardless of their race, religion, national origin, ethnicity, disability, sex, gender identity, sexual orientation, or age—with respect, and without judgment or bias.” In practice, though, platforms varied greatly in the extent to which they curated members and activities. Freelancer.com, the online workplace platform established in 2009, connected over 24 million employers and freelance workers globally. It did not restrict the number of types of freelancers. By contrast, rival Upwork, founded in 1999, performed algorithmic curation and used multiple means to verify that freelancers are who they say they are, including authenticating email addresses and providing results of online skills tests. It also allowed job advertisers to include custom screening questions in their job posts and provided chat and videoconference tools for interviewing finalists. However, Upwork did not take responsibility for the quality of the work performed, and explicitly stated that it was ultimately the responsibility of the job advertiser to choose the freelancer. But the purpose of the curation was to improve the quality of work and build trust in its matchmaking functions.

Another way to increase the quality of engagement as well as users’ trust was to increase the transparency of performance data on platform members. Reviews and evaluations are ubiquitous features of transaction platforms such as TripAdvisor, Expedia, Airbnb, Uber, Amazon Marketplace, Upwork, and many others. They usually offer a simple interface that allows users to review and grade the performance of those who perform the services accessible through the platform. The platforms then compute the results for an overall score and make the scores and individual reviews visible and searchable. In this way, transaction platforms generally allow complete strangers to assess quickly whether they want to engage or transact with each other.

Reviews and evaluations, computed automatically through algorithms, play the role of social vetting of platform members. They created a searchable repository of past transactions and reviews. Allowing members of various sides to build a reputation creates value for users and provides incentives for users to return to the platform, making it “sticky,” instead of multi-homing with several platforms. Fake reviews and other undesirable postings continued to be a problem for many platforms, although they were investing in artificial intelligence tools to help identify fake or unauthorized reviews (such as those written by a restaurant or hotel owner, or by a product company to damage a competitor by submitting a negative review) as well as other inappropriate content. Authenticity of reviews was an especially critical part of the value that platforms provided to users. Some platforms, such as Facebook, Google’s YouTube, and TripAdvisor, increasingly used people as well as AI tools to verify problematic reviews and content. Upwork also displayed each freelancer’s “Job Success Score” and feedback on past projects.

In some cases, platforms themselves monitored the work performed by one side. For example, Uber’s rating system measured performance, and this had consequences: Drivers were “deactivated” or “taken off” the platform if their rating fell too low. Uber also encouraged drivers to work as much as possible and not turn down requests for rides; it even computed and posted a score for the drivers’ acceptance rate. Freelancer work platforms such as Upwork also required time-tracking apps for hourly jobs, which monitored the work performed by taking (with the approval of the freelancer) screenshots of the freelancer’s computer.

Although transaction platforms may provide tools for users to assess the quality or frequency of the service performed, they usually don’t take responsibility for the activities themselves. There are exceptions. For example, some platforms offered a guarantee or insurance for the quality of the service to buyers, and some online marketplaces organized product returns. But platforms differed in the extent to which they provided these customer services. Amazon Marketplace had a guarantee and return service. It also offered to mediate between buyers and sellers to enforce sellers’ rapid response to customer cancellation and return requests. Other online marketplaces such as the French Le Bon Coin did not.

Fraud prevention was another area where transaction platforms differed in how much responsibility they assumed. It remained a matter of debate to what extent firms such as eBay monitored and fought the sale of counterfeit goods; there were also differences across regions in how rigorously platforms applied their different rules. While eBay tended to be relatively strict (a policy we explore in Chapter 4), some of its competitors, such as Alibaba in China, historically took the opposite approach.

Many platforms take great care in limiting their liability. Airbnb stated in its June 2017 updated terms and conditions that, “as the provider of the Airbnb Platform, Airbnb does not own, create, sell, resell, provide, control, manage, offer, deliver, or supply any Listings or Host Services. Hosts alone are responsible for their Listings and Host Services. . . . Airbnb is not acting as an agent in any capacity for any Member.” And with regard to its role within potential disputes, Airbnb stated that, “while we may help facilitate the resolution of disputes, Airbnb has no control over and does not guarantee (i) the existence, quality, safety, suitability, or legality of any Listings or Host Services, (ii) the truth or accuracy of any Listing descriptions, Ratings, Reviews, or other Member Content (as defined below), or (iii) the performance or conduct of any Member or third party.” Finally, it stated that “Airbnb does not endorse any Member, Listing or Host Services.” Ultimately, governance policies and skills could make or break a platform. Problems ranging from Russian interference in U.S. elections via Facebook to stolen content uploaded to YouTube can bring the very legitimacy of a platform’s business into question.

Hybrids: Combining Transaction and Innovation Platforms

As we illustrated with Figure 1-2 in Chapter 1, some platform companies have adopted a hybrid strategy. They combine transaction and innovation functions within the same platform infrastructure or launch transaction and innovation platforms within the same firm. The most valuable platform businesses (and these rank among the most valuable companies in the world) are all hybrids: Microsoft, Apple, Amazon, Alphabet-Google, Facebook, Tencent, and Alibaba, among a few others.

The hybrid approach seems popular and valuable because it combines the best of both worlds: Transaction platform companies like Facebook or Tencent (with WeChat) can add an innovation platform function to access the innovative capabilities of third-party firms. They can make their social media or messaging activities more compelling with minimal in-house investment. Innovation platform companies like Apple or Google (with Android) can establish a separate transaction platform or store to distribute complementary innovations and digital content that make the platforms more valuable. As a side benefit, innovation platforms generate additional revenues from sales or transaction fees.

We also see two different types of hybrid strategies, although they exist more along a spectrum rather than with sharp distinctions. On one extreme, we have platform companies that add the second type of platform and connect the two in some meaningful way. The connections can be through cross-marketing to a common user base, relying heavily on digital technology, analytics, and unified customer databases. Or a firm might strengthen its innovation platform by establishing a marketplace or digital store to distribute applications and content. These investments make the innovation platform more valuable and have the potential to generate cross-side network effects, such as between end users and producers of complementary innovations like smartphone apps. We refer to connecting the two different types of platforms within the same firm as an integrated hybrid strategy.

On the other extreme, a platform company might add the other type of platform but without connecting the two technically or operationally. This approach we refer to as a conglomerate hybrid strategy. These organizations tend to resemble conglomerates in the non-digital world that own several distinct businesses under one organizational umbrella or holding-company structure. There is no tight coupling of the businesses such as through technology, marketing, or customers, although there may be some flow of funds across the different business units and a central staff, such as for administration or research. In the conventional economy, General Electric was the largest and most successful conglomerate in the second half of the twentieth century. Even so, it struggled in recent years due to competition from more efficient, focused businesses. Most research in management suggests that conglomerates with unrelated businesses will underperform firms that expand through related diversification. Nonetheless, big platform businesses such as Alphabet-Google, Amazon, Alibaba, and Tencent, as well as Yahoo in prior years (it is now owned by Verizon), tend to have both integrated and conglomerate-like investments in their portfolios.

INNOVATION + TRANSACTION

The main reason why innovation platforms add transaction platform capabilities (in other words, a marketplace) is to facilitate and control the distribution channel of complements. Transaction features create value for the complementors by offering them a distribution infrastructure. At the same time, the companies can capture a larger part of the value created. Innovation platforms are relatively useless without applications and digital content, and the best and cheapest way to increase applications is to encourage third-party developers to innovate and then make distribution easy through integrated app stores.

Before the Internet, few companies combined innovation and transaction platforms. But in the last two decades, several innovation platforms added a transaction dimension soon after launching. Apple added its App Store about eight months after launching the iPhone. Google launched its app store (Google Play) about a year after it started licensing Android. Learning from these other examples, Amazon launched its app store for the Alexa home-speaker AI device along with introduction of the new platform. (See Chapter 7.) We should note that the Palm personal digital assistant (PDA), a precursor of the smartphone popular in the late 1990s, also launched an app store accessible through the web, so Apple was not first. App stores and digital content stores can generate significant additional revenue for innovation platforms and complementors. Apple and Google took a 30 percent cut from apps sold on their respective app stores. In 2017, Apple’s App Store revenues reached $11.5 billion, while app developers received payments amounting to $26.5 billion. Although Google also charged 30 percent in 2018, Google began its app store taking only a 20 percent cut in order to provide a bit more incentive for developers to write apps for Android.

TRANSACTION + INNOVATION

The main reason why transaction platforms add an innovation platform side to their business (in other words, they open their APIs and some customer data to outside firms) is to stimulate innovation by third parties. More apps or features generally make the transaction platforms a more compelling experience for users and create additional opportunities for monetization, such as to sell more advertisements or take different types of transaction fees. The data generated by the transaction platform on user behavior also becomes a valuable asset that the platform company and third parties such as advertisers can leverage to better understand user behavior and needs, or to design marketing strategies aimed at generating cross-side network effects among the platforms.

We mentioned before that Facebook made itself more compelling by enabling access to millions of applications and websites. This openness can also backfire, as we saw in the case of Cambridge Analytica as well as later revelations on how much data Facebook collected from users (often without explicit permission) and then provided to advertisers and app developers. In general, though, as Microsoft and other companies discovered long ago, becoming a hybrid with an innovation platform function allows transaction platforms to add new capabilities or features without high internal R&D costs. Instead, they can access the external supply of creativity and software engineering skills available worldwide. But sometimes adding the second platform is more an act of desperation. Snapchat, for example, has struggled with competition from Facebook’s Instagram. In 2018 it opened up its APIs to encourage third parties to build complementary innovations in the hope that some new apps would make Snapchat a more compelling experience for users and a better draw for advertisers. Besides Facebook and Snapchat, a less obvious transaction-to-innovation hybrid example was Expedia, the travel services platform. When Expedia established an affiliate program under the banner of “Your Business. Our Technology,” it opened up a set of APIs. The APIs allowed outside companies to add hotel, flight, and car booking capabilities to their own applications. The Expedia APIs also made it possible for external developers to build applications that used Expedia’s transaction capabilities, which supported over thirty currencies and ten different types of credit card payments. In addition to offering access to its transaction technology, the Expedia APIs made it possible for third parties to access the platform’s “rich content,” such as over 11 million images of property, 6.5 million room images, 100 characteristics per room, and 380,000 geography destinations, all supported in over thirty-five languages.36 INTEGRATED VS. CONGLOMERATE HYBRIDS

Some hybrid companies integrate their two types of platforms more closely than others. Apple, for example, tightly connected its main innovation platform (the iOS operating system for the iPhone and iPad) with its transaction platform (the App Store) and online store (iTunes), as well as other services (iCloud, iBooks, etc.). Icons for the App Store and iTunes come bundled on the iPhone boot-up screen, and product registration includes registration for apps and content, with some additional information requests, such as for a credit card. Tight integration has multiple benefits: The App Store provides a distribution channel allowing iPhone app developers to access the vast pool of iPhone and iPad users. The integration creates value for the app developers who want to sell their apps. Providing complementors and content providers with a global distribution channel increases their incentives to remain active within the Apple ecosystem and keep developing new apps and content for Apple. The existence of the App Store keeps end users interested in buying new iPhones and iPads, since a continuing stream of new apps and content makes the devices increasingly versatile and attractive. Benefits go in the other direction as well: The App Store generates increasingly large revenues and profits from the distribution of software, digital services, and digital content, which Apple hopes will one day encompass a very significant portion of the company’s business.

Hybrid platform strategies provide other strategic benefits when they work well. They raise barriers to entry for innovation platform competition because it takes time for competitors to build a portfolio of apps and a developer ecosystem. Hybrid strategies also weaken competitors by giving the platform owner power to exclude rivals, although they need to be careful not to run afoul of antitrust regulation. (We return to this theme in Chapter 6.) For example, Android is open source and free to license, but Google imposed tight restrictions on smartphone makers who wanted to offer access to Google Play (the largest marketplace for Android apps). Smartphone makers needed to obtain a license for Google Mobile Services, which meant accepting a stringent (and ever-growing) list of device-centric requirements, with an obligation to continuously validate their devices against “Google’s compatibility test suite.” Google Mobile Services included extremely popular apps such as YouTube, Gmail, Google Maps, and Google Docs. Smartphone makers who did not use the Google-approved version of Android had a hard time passing these compatibility tests and finding alternative services.

Google also has done more than simply tie Android apps to its Android operating system and services. In the early days of Android, vendors could introduce some variations into the free and open-source Android code and build their own apps. This led to some separate and incompatible ecosystems and allowed handset makers such as Samsung and Xiaomi to capture more value for themselves. However, Google wanted to prevent these “forks” in the Android software and their incompatible applications. No doubt Google also wanted to prevent other companies from drawing away advertisements or software sales. Consequently, Google moved the APIs for apps development from the Android operating system to the Google Play store itself. Since 2012, most developers now build their Android apps to be compatible with the “Google Play Services APIs.” Google can also update this API layer independently of operating system updates (which smartphone manufacturers, not Google, control).

Making the Google Play Store into the applications platform was an elegant solution to the problem of Android fragmentation and potential loss of revenues for Google. Some of these tactics have run afoul of European antitrust regulators, however, and Google responded by changing some of its policies. Nonetheless, having both types of platforms and integrating them closely was a competitive advantage that Google had versus its rivals. Amazon is another case of a hybrid platform company with different levels of integration, albeit with economies of scale and scope even in seemingly unrelated businesses. The Amazon online store (new products that Amazon is selling or reselling) and Amazon Marketplace (which sells third-party goods, new and used) both offer the same types of goods. As users see only one interface screen, the network effects from the marketplace business directly reinforce the online store business. The two businesses also run on the same digital infrastructure, powered by Amazon Web Services (AWS), and collect and analyze customer transaction data. At the same time, AWS is a separate transaction and innovation platform for companies that want to rent cloud services and utilize bundled features and software engineering tools as an applications development environment. Then we have the Kindle and Alexa devices, which are separate transaction and innovation platforms, although both connect users to the Amazon online store and the Amazon marketplace to buy digital books and other goods.

The top hybrid companies in China also expanded from one platform type to the second type and operated their platforms and other businesses with different degrees of integration. Many Chinese users do not use personal computers or credit cards very frequently, so Chinese companies in some sense have leapfrogged these technologies. They have made the smartphone and applications like WeChat into a single platform for many activities and services. Tencent provides a good illustration. It started out in 1998 with a messaging platform (QQ) for Chinese users of personal computers, and generated income from advertising and a premium messaging service. From 2004, it expanded into developing and hosting online games for a fee. When mobile devices became popular, Tencent evolved QQ into WeChat, which added social media functions to the messaging app. WeChat today has over a billion subscribers. Tencent also evolved WeChat into a popular innovation platform. Tencent as well as third-party firms use WeChat APIs to build a wide variety of apps, offering services such as electronic payments as well as video games and ride sharing. (Tencent, as well as Alibaba, are investors in Didi Chuxing, which acquired Uber’s Chinese ride-sharing operations.) Tencent also closely integrated WeChat with its game development and hosting platform. In terms of the business model, WeChat makes money by investing the cash that users deposit in their electronic payment accounts. It also earns fees from video games and sending customers to its partners such as Didi Chuxing and JD.com, a major Chinese online shopping platform in which Tencent invested in 2014.

Key Takeaways for Managers and Entrepreneurs

In this chapter, we discussed the differences between innovation and transaction platforms and how they followed the same steps in building their platforms: Identify platform market sides, solve the chicken-or-egg problem to launch and generate network effects, find a workable business model, and establish governance rules to determine who is allowed to do what through the platform. We also discussed hybrid platform strategies that combine innovation and transaction platforms or functions within the same firm. So, when it comes to selecting your platform strategy and business model, what are the key takeaways for managers and entrepreneurs?

First, and most obvious, is the need to understand the strengths and weaknesses, as well as the potential costs and benefits, of the two different types of platforms. For example, successful innovation platforms are rare because they sit in the middle of enormous ecosystems with thousands, millions, and even billions of participating firms and individual users. The world can support only so many of these powerful innovation platforms, even though the platform concept may be cheap to introduce (such as when a product firm opens up its APIs to third-party firms). More often, innovation platforms are very expensive and risky to build from scratch (such as to create a new mass-market operating system or cloud-computing infrastructure). About 60 to 70 percent of start-up firms and billion-dollar unicorns were transaction platforms. Why? It seems easier and cheaper for entrepreneurs to create transaction platforms because they are relatively simple to build on a technical level and they can provide value in many different markets, ranging from sharing cars and rooms to sharing household tools and pets. Innovation platforms, by contrast, require the creation of a core technology that can serve as a foundation for other firms to build complementary products and services. This type of platform was most common in markets for software and hardware system technologies. Consequently, we mainly hear about innovation platforms for computers, smartphones, consumer electronics, video game development, and cloud-based hosting and application development environments. Nonetheless, the fact that transaction platforms like Amazon, Facebook, Tencent (WeChat), Uber, and Airbnb were adding innovation platform functions suggested that managers and entrepreneurs should take a much broader view of where they can build useful innovation platforms.

Second, innovation platforms can enable “open innovation” in a variety of settings. They can be an effective way for companies to enhance the value of their products and services with relatively small in-house investments compared to the potential benefits from thousands or even millions of third-party innovations. Not surprisingly, despite their relative rarity, we see new innovation platforms emerging. IBM was trying to turn its Watson AI technology into a new consulting service as well as an innovation platform by building partnerships with application developers at companies and universities, especially for health care applications. General Electric opened up its Predix operating system to other firms, encouraging them to build products and services for the Internet of things. (We explore this case in Chapter 5.) We also have some open general-purpose technologies emerging as potentially new innovation platforms. Blockchain is a good example. This was once associated with Bitcoin, the cryptocurrency (also a kind of transaction platform technology). Various firms were starting to use blockchain software to track different types of transactions over the Internet, including shipments of food as well as transfers of money and confidential documents. Third, managers and entrepreneurs in many more industries probably need to give serious thought to combining innovation and transaction functions—adopting a hybrid strategy. It is clear to us that hybrids are the next phase in the evolution of platform thinking. They push even further the logic of what makes a platform such a powerful strategic weapon. Platforms are all about leveraging complementarities among existing assets and organizational capabilities. Sometimes the goal is to take advantage of user engagement to make customers more attractive to other market sides, such as advertisers or application developers. At other times, the goal is to leverage a shared asset by making it available to third parties for their own innovations.

But while both innovation and transaction platforms create value through increasing the use of existing assets, hybrids take leverage a step further. This is especially the case when hybrids use their expertise in digital technology to facilitate integration between various software-based services and allow the use or reuse of software modules and user data. Not surprisingly, hybrids seemed to work best when companies designed them so that the transaction and innovation platforms were integrated and took advantage of each other, technologically and strategically. Digital conglomerates with weakly connected or unrelated businesses should not have any particular advantages over traditional conglomerates. Yahoo comes to mind since it evolved from a web directory service into a hodgepodge of Internet properties (search, email, shopping, news, sports and financial information, etc.) selling mostly generic ads, with different user bases and minimal network effects. However, as we have seen at Amazon, Google, Alibaba, and Tencent, hybrid firms can create new types of “relatedness.” They can centralize customer data and analytics for cross-platform marketing and advertising. They can also take advantage of enormous scale and scope economies by using the same digital infrastructure to power and connect multiple platforms and other online businesses such as digital stores.

We also need to remember that strong network effects are difficult to generate and healthy business models tend to elude most platform ventures. The most common mistakes managers and entrepreneurs make when launching new platforms is the subject of the next chapter.