5

Government as a Platform

The irony was not lost on me that I was late for my meeting at Waze, stuck in traffic in Tel Aviv.1 To make matters worse, when I finally did arrive, and I finally did get to asking Gai Berkovich about his company’s ability to grow its free data-sharing program with cities, I seemed to have put my foot in my mouth. “In two years,” I asked, “how many cities will the Connected Citizens Program be in? One hundred?” Waze had gone from zero partners to fifty over the program’s first two years.

“I think that’s the wrong question,” Berkovich replied. “How many cities won’t we be in?”

The Connected Citizens Program (CCP) was an extension of the core Waze navigation product that more than sixty-five million commuters used each month in 2017 for routing and real-time traffic updates.² Waze called those users “Wazers” and described them as the world’s largest community of drivers. The basic idea for CCP was that Waze would share with cities traffic data it was gleaning from its users (the app passively collected speed and congestion data as drivers used it to navigate their commutes), and cities would share with Waze planned special events, street closures, street construction, etc. Win-win. CCP had a religious origin of sorts: The impetus for the program was a visit by Pope Francis to Rio de Janeiro in 2013 and a suggestion by Rio’s mayor that his staff reach out to Waze to learn how to incorporate the company’s data into the city’s planning for the momentous visit. Rio was the company’s first data-sharing partner, and nine others joined in October 2014. Now, two years after that, I was wondering how many more were coming, and Berkovich, the chief operating officer at Waze, was chiding me (with a grin) for being skeptical.

My doubt wasn’t totally unfounded. There was much interest in CCP at Google (Waze’s parent company) and in the world, but there weren’t many resources dedicated to it. More than eighty cities and groups were asking to join the program. More than fifty had. But each new partner had needs: how to join the program, how to make sense of the data, and how to turn that sense into action. The cities didn’t have the spare talent—not the data analysts or data visualists or traffic engineers—to do all that. And neither, it seemed, did Waze have the people to support them. Inside the behemoth that was Google (and then Alphabet), there was something close to seventy thousand employees at the time, and yet CCP was supported by just three of them.3 Within Google, where the culture was to build things and then build them “10x” bigger, these three people were expected to grow the program to five hundred cities—or, in Berkovich’s conception, to every city on the planet. I felt not totally stupid wondering how?

I wasn’t the only one wondering. Paige Fitzgerald was curious, too, and she had more than a passing interest. Fitzgerald was CCP’s program manager. She led and made up one-third of the scant team. She was seeking more full-time help on the project, but Di-Ann Eisnor, CCP’s founder and the head of growth at Waze, had demurred. For the time being, Fitzgerald borrowed resources from inside Google where she could. “We basically operated on goodwill,” she recounted. Fitzgerald considered whether the time had come to charge for the free program. Maybe then the team would have the resources she felt they needed. “We can’t work hand-in-hand if we scale,” she had told me in the summer of 2016 in the face of the growth demands. “Our biggest challenge is just trying to hang on for dear life.”

If my doubts about the ambitions for CCP were well founded, they were more broadly founded, too. Hanging on for dear life was a professional hazard of program management in public work. CCP wasn’t the first time a sleek new program had been thought up, and even tried, in order to solve a public problem. But too often we observe that the press releases, the ribbon cuttings, and the celebratory launches are followed by . . . not much. Can we solve public problems anymore? is a question in two parts: Can we come up with new ideas? And can we try them? But then comes a third part. In the wake of too many experiments that lead nowhere, or “shiny objects” that deliver acclaim but no results, or that do deliver but are starved of resources nonetheless, we also have to ask ourselves: Can we scale those new solutions so that they make a transformative difference in people’s lives?

If I’d faced as much traffic heading home, maybe I would have had the time to fully recognize the rationale for Berkovich’s confidence. He was facing the age-old question of how to take a new solution and make it a big one, and he didn’t seem fazed. Maybe I wouldn’t have attributed his self-assurance to mere ambition. It took me some time—and some visiting of other tech companies with similar appetites for growth, and a rereading of the history of government ventures that accomplished things at large scale—before I recognized the thing (and then worried about it, too) that gave Berkovich confidence. It wasn’t blind ambition he was channeling. It was “platform thinking.”4

The Drivers’ Wikipedia

Waze began as a community project in 2006 called Free Map Israel, led by Ehud Shabtai.⁵ Shabtai was joined in creating the company by Amir Shinar and Uri Levine as they built what one of them described as “the drivers’ Wikipedia.”⁶ Over its first ten years, Waze developed into a free, real-time, crowdsourced navigation app. Waze could be used anywhere in the world but required that there be enough people providing data to make it useful. In the spring of 2016, while CCP was getting off the ground in just a handful of cities, 186 metropolitan areas globally each had more than twenty thousand active Wazers using the main app.⁷

In addition to the Wazers, the company had relied on more than 360,000 volunteer map editors around the world who updated maps and verified local edits to improve Waze’s accuracy. One journalist commented on Waze editors and their “almost missionary zeal” to improve the quality of the app. He described a top editor, Jesse May, who spent “hours on the computer every night, making sure the world that Waze presents to drivers is the world that is really out there. ‘It’s my hobby . . . Instead of playing Doom or watching the boob tube, I’m going to do something that’s more beneficial.’ Sometimes he will even drive out of his way to check on a user report in person. ‘My wife just shakes her head.’”⁸

In November 2012, Waze began monetizing the app by offering resellers and advertisers a web interface to advertise based on location. The next year, Google bought the company for a reported $1.3 billion amidst rumors of a counteroffer from Facebook and interest from Apple.9 Waze, with its community of sixty-five million users, had something of interest to these tech giants. It also had something in common. Like Facebook’s core offering, like Google’s search, like Apple’s iOS for its iPhone, the Waze app was more than a product. It was a platform.

Platforms Bring Users Together

A definition is a useful starting point for understanding how Waze got to where it did, how CCP might, and what that has to tell us about how we might scale efforts to solve public problems. Platforms “bring together individuals and organizations so they can innovate or interact in ways not otherwise possible, with the potential for nonlinear increases in utility and value.”¹⁰

Consider two types of platforms. The first, innovation platforms, “consist of common technological building blocks that the owner and ecosystem partners can share in order to create new complementary products and services.”¹¹ Apple’s iOS is a platform of this type. An entire ecosystem of app developers has grown up around the smartphone operating system, all adding value to the phone itself. Google’s Android is, too. So is Microsoft Windows for the PC, and Amazon Web Services for the cloud. Transaction platforms, the second type, “are largely intermediaries or online marketplaces that make it possible for people and organizations to share information or to buy, sell, or access a variety of goods and services.”¹² Google’s search is a platform of this type, allowing advertisers to target searchers. So is Facebook’s social network, allowing individuals to share information with each other. So are Amazon’s Marketplace, Tencent’s WeChat, and many other platforms. Airbnb and Uber, which we came across in chapter 4, are transaction platforms. So is Waze.

These platforms all bring users together, but there is something very particular about them that gives them their astronomical growth: They bring users together in ways that create value for other users. I found the case of Waze so illuminating because with one use of the app, this power is easy to see. If I were the first user of Waze, the app would be of no value to me. But as soon as other drivers started downloading and using it on their way to work or home or school, it became more valuable. The more users there are, the more traffic information there is, which makes the app’s route suggestions better. And because of this value, more and more people join and fewer and fewer leave, which compounds the usefulness further. The power is in the network of users. These effects have come to be called network effects.¹³

Not all products and services have these network effects. If I buy a Ford car, it’s not as if another buyer of a Ford creates much value for me in my car. I’m mostly indifferent, really, as to whether others buy a Ford or a Toyota or any other brand. Not so for Waze or for any of the other platforms we came across. The same effects are at play for Facebook and Twitter. Also for Airbnb: The more guests, the more value for hosts. The more hosts, the more value for guests, who can choose among a wider array of apartments. The more guests and, consequently, the more guest ratings, the more reliable the platform is for other guests. Same for Lyft and Uber. Drivers bring passengers. Passengers attract drivers. Passengers make it better for other passengers with their ratings. Users add value for other users. (See figure 5-1.)

FIGURE 5–1

These effects are in play both for users of similar types and for those across the platform. When the effects involve users on the same side of a platform adding value for each other, they are called “same-side” or “direct” network effects. Wazers making the app more useful for each other is an example of direct network effects. When it is users on different sides of a platform adding value for each other, these are called “indirect” or “cross-side” network effects. When more iPhone users attract more app developers, that’s a cross-side network effect. Some platforms have both. Same-side network effects are at play when Facebook users create content for each other. Cross-side network effects are in action when developers build apps that work on Facebook’s platform.

Network effects can be positive—the addition of the second user or partner adds value for the first—but they can also be negative, where users degrade the value of the platform for others. More users on Airbnb means more ratings, but it also means more competition for that prized rental for a vacation weekend. If fraudulent reviews swamped Amazon’s Marketplace, there would be large negative network effects at play.14 The same goes when Facebook fills up with haters and harassers, and the rest of us leave.

The same goes when your commute to work grinds to a halt on a highway. The next time you want to pound your fist on the steering wheel and curse being late for your 9:00, feel free to blame “the platform.” Roads have giant network effects—positive and negative ones—and are a great reminder that platforms existed long before Amazon and Facebook and Microsoft and even before fax machines and telephones. A road that connects one person to nothing is of no value (unless it is solitude you seek). But as soon as a friendly neighbor or a useful business pops up alongside, the road’s value has increased for everyone. Waze and the roads its users ride on demonstrate that it’s not just the tech companies that build platforms that leverage network effects. Governments do too.

Government as a Platform

Tim O’Reilly is one of the chroniclers of the internet and has been writing about government as a platform for at least a decade. He drew on the roadway analogy: “The Federal-Aid Highway Act of 1956, which committed the United States to building an interstate highway system, was a triumph of platform thinking, a key investment in facilities that had a huge economic and social multiplier effect. Though government builds the network of roads that tie our cities together, it does not operate the factories, farms, and businesses that use that network: that opportunity is afforded to ‘we the people.’”¹⁵

O’Reilly observed a similar phenomenon in many other aspects of public life. “The launch of weather, communications, and positioning satellites is a similar exercise of platform strategy. When you use a car navigation system to guide you to your destination, you are using an application built on the government platform, extended and enriched by massive private sector investment. When you check the weather—on TV or on the internet—you are using applications built using the National Weather Service (or equivalent services in other countries) as a platform.”16 He contrasted this approach with “vending machine government” and explained what he meant: “We pay our taxes, we expect services. And when we don’t get what we expect, our participation is limited to protest—essentially, shaking the vending machine. Collective action has been watered down to collective complaint . . . What if, instead of a vending machine, we thought of government as the manager of a marketplace? . . . In the vending machine model, the full menu of available services is determined beforehand. A small number of vendors have the ability to get their products into the machine, and as a result, the choices are limited, and the prices are high.” A marketplace, “by contrast, is a place where the community itself exchanges goods and services.”¹⁷ O’Reilly suggested that if we look hard enough and long enough, we might find this strategy in action all across public stuff. “The idea of government as a platform applies to every aspect of the government’s role in society.”¹⁸

Where could we look to find these platforms? We could start in our own neighborhoods. We would, for example, find them in community policing and neighborhood watches. These approaches were getting new looks in 2020 in the wake of the Black Lives Matter movement and calls to reduce funding to police departments or to reimagine public-safety approaches altogether. In the vending-machine model, how do cities and towns provide public safety? When crime goes up, more police officers are promised; more services are delivered by the government. And in a platform model? The community is invited in to play a role. The city, instead of (or in addition to) providing more patrols, provides what? Maybe it’s a folding table and some chairs in a community space on a Wednesday evening. Not too long ago, it might have been access to an email list to share information. Nowadays, it is access to open data on everything from the status of streetlight repairs to the rate of car break-ins to camera feeds. But whatever the tools, the idea is the same—connect people so that they can share information (a transaction platform) or so that they can create “complementary products,” such as a teen basketball league (an innovation platform). And all the while, exploit the fact that more users make other users better off. If more of my neighbors join and share, my family is safer. (Except when neighbors start using the meetings to pester other neighbors, or when community policing is a guise for vigilantism or worse, or when digital versions of the neighborhood watch become platforms for racial harassment, as they have on more than one occasion, and then, decidedly not.)

We could also look to national government. In 2010, at the urging of and under the leadership of one of India’s famed internet entrepreneurs, Nandan Nilekani, the country undertook an initiative to provide Indians with a universal identification. (It was akin to an American Social Security number.) Indians didn’t have a national ID, which meant they didn’t have a way to prove their identity, which meant that getting access to government services like subsidized food or to private ones, like a bank account or mobile phone, was difficult. That, Nilekani observed, left a public system rife with fraud and abuse, marginalized the poor, and hampered the economy.19 Nilekani assumed a senior government post, set up shop in Bangalore, and issued the team its own sort of “what cities won’t we be in” mandate: reach 600 million Indians in five years, and ultimately all 1.2 billion of them.

Achieving that scale that speedily would be extraordinary. It took Facebook eight years to reach a billion users around the world and WeChat just over seven to get its first billion, and neither required a user to sign up in person.20 Each twelve-digit, randomly generated ID number would be linked to all ten fingerprints of a person and to iris scans of both eyes. That meant collecting 2.4 billion iris scans and 12 billion fingerprints across more than four thousand cities and towns. India’s people speak at least 121 different major languages, and actually many more, and every person would have to be spoken with.²¹ India had at the time something close to 75 million homeless people, a population on its own that would have made it the twentieth biggest country in the world, and they all had to be reached.²² On top of that, Nilekani and his government made this seemingly impossible task even impossibler: they made the program voluntary.

Then Nilekani did the thing that tells you much of what you need to know about his plan for getting there. He gave the ID the name Aadhaar. Aadhaar means “foundation” in Hindi. He branded the thing as a platform, and then he built it that way. “There are two big things we have to do. We have to constantly focus on enrolling the un-enrolled and make sure there are useful applications for the enrolled,” he said as they got underway, alluding to two sides of the platform.²³ On the one side: 1.2 billion people if they could be enticed. And on the other: government agencies and businesses that had subsidized food and fuel and the private mobile phones and bank accounts they wanted. “Think about it like a flywheel that has to be put in motion so it gains acceleration. The real strategic challenge is how do we create a virtual flywheel which spins faster and faster, and gains so much momentum that it is unstoppable,” Nilekani had added.²⁴

How do you? It’s a chicken-or-egg problem that all platform launchers face, and it 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?” There are three broad approaches: “(1) Create stand-alone value for one side first, (2) subsidize one or both sides, and (3) sometimes bring two sides on board simultaneously.”25 Three platform scholars offer up a list of eight maneuvers for dealing with the dilemma: ²⁶

1. “Follow the rabbit.” “Use a non-platform demonstration project to model success.”
2. Piggyback. “Connect with an existing user base from a different platform and . . . recruit those users to participate in your platform.”
3. Seeding. “Create value units that will be relevant to at least one set of potential users.”
4. Marquee. “Provide incentives to attract members of a key user set onto your platform.”
5. Single side. “Create a business around products or services that benefit a single set of users; later convert the business into a platform business.”
6. Producer evangelism. “Design your platform to attract producers, who will induce their customers to become users.”
7. Big-bang adoption. “Use one or more traditional push marketing strategies to attract a high volume of interest.”
8. Micromarket. “Start by targeting a tiny market that comprises members who are already engaging in interactions.”

Amazon used the first strategy when it opened its traditional business (selling goods it held in inventory) to its Marketplace business and allowing other small businesses and vendors to participate. PayPal used the second strategy (among others), as it got started and grew. The authors explained, “PayPal’s marketing team . . . simulated customer demand on eBay by creating a bot (an automated software tool) that bought goods on the site and then insisted on paying for these transactions using PayPal. Noting this apparent growth in demand, many eBay sellers signed up for the PayPal service—which in turn made PayPal even more visible and attractive to consumers . . . within three months, PayPal’s user base grew from 100,000 to one million.”27 Adobe pursued a seeding strategy and launched the PDF document-reading tool “by arranging to make all [US] federal government tax forms available online. The size of this instant market was huge, encompassing any individual or business that might need to pay US taxes. Adobe induced the IRS to cooperate by suggesting that millions of dollars in printing and postage costs could be saved. Taxpayers, in turn, got fast, convenient access to documents that everyone needs . . . Impressed by the value provided, many adopted Adobe as their document platform of choice.”²⁸ Nintendo used the marquee strategy by signing up major video-game designers to release games on its new consoles. OpenTable, the reservation platform, adopted a single-side strategy by first using the app as a seating inventory software for restaurants, and only later opening it up to restaurant guests. Kickstarter turned its creators into evangelists by giving them the tools to host and manage their campaigns, and thus attracting crowdfunders to the site. Twitter tried the big-bang approach with success when it launched at South by Southwest in 2007. Facebook fell into the last strategy when it started at Harvard College.²⁹

What did Nilekani do? He started by evangelizing himself: “I visited 25–26 state governments. I went to Raipur, Lucknow, Patna . . . I met chief ministers and chief secretaries, made presentations, got everybody on board, met banks, insurance companies and oil companies. We had talks at industry associations. I spread the message.” Nilekani and his team built partnerships with registrars to enroll residents. They went to big organizations that already interfaced with huge populations, including federal agencies, state governments, the India Post, and telecom companies, and sought to piggyback on their enrollments. They built a training system for operators who enrolled residents and then improved it when it was too slow. They ran pilot projects to test authentication for banks and telecoms, and they got regulators to agree to accept Aadhaar “as proof of both identity and address, enabling investors to . . . buy mutual funds.” They even had a budget to spur application development by state governments and federal ministries. They gave awards to district collectors who developed interesting apps. They supported business-plan competitions for private Aadhaar entrepreneurs. They did all of this with the notion of getting that flywheel started, so that the growth would then occur on both sides, exponentially, and also with the notion of tamping down the doubts. “The best way to deal with [opposition] is to show [Aadhaar’s] usefulness . . . to give unmatched benefits . . . that is what will win the debate.”30

It looked for a while like this wouldn’t be enough, like Aadhaar would miss its goal. In 2012, three years into the program and two years since the first person had been signed up, 190 million had enrolled.³¹ But that meant more than 200 million to go to meet the 600 million target in just two years’ time. The team hit those numbers and more. By December 2014, more than 700 million had signed on. By December 2016, there were well over a billion. By the end of 2017, the team had hit 1.2 billion. By the summer of 2019, they were providing almost 1 billion authentications a month.³² They’d provided more than 32 billion overall. They’d achieved their “which cities won’t we be in” goal.

What could Waze do? It was at fifty cities, with every other one to go. Paige Fitzgerald, the CCP program manager, was hanging on for dear life, yet meanwhile the program seemed to be a boon for Google. Waze’s parent company had come under fire for tax avoidance, antitrust violations, privacy violations, and censorship—the kitchen sink of tech scrutiny—but every time CCP came to town, it generated positive headlines for the company. A student of mine, a former Googler, observing that CCP was worth “$100 million” in good PR to the company, said that she would ask for a $5 million budget were she Fitzgerald. I agreed that in terms of PR, in terms of the groundwork it laid for Waze’s next big effort in cities (its carpool service), CCP was worth—if not $100 million or even $5 million—at least two or three more engineers and a few program administrators. What Waze could do was fund them. Provide the resources. From Waze. From Google. Di-Ann Eisnor, CCP’s founder, told me, “We put a video of CCP up at TGIF, the Google weekly meeting. Sergey [Brin] pulled us aside and said, ‘I love this.’” So, find the money? Isn’t that what you do to support programs that work? Build the case with data, go to the higher-ups, get the goods?

Or . . . what Waze could do was what Waze always did and what Google had done with Search and YouTube. Turn CCP, the program, into CCP, the platform. Connect people on and/or across a platform. Create and strengthen ways for them to add value for each other. Get the flywheel going by getting one side started, or the other, or both.

And that’s what they did. “Waze built a community portal so the cities could help each other,” Fitzgerald explained. There the cities exchanged case studies and analytical tools they had created on their own. Waze held summits in New York, in Paris, in Mexico City, where transit leaders shared tools and techniques. The CCP team piggybacked onto an existing platform, inviting super-Wazers to help in the cities where they lived. Fitzgerald and her small team—still three full-timers—created some marquee wins that would attract others to the platform. They turned their best cities into evangelists. “We couldn’t continue to hand-hold all of these partners,” Fitzgerald confessed. “We focused on high-profile partners as well as innovative partners who were engaged, who would create tools that could be shared with other partners as well . . . We had those summits and we figured out even giving out just speaking spots was a big motivator for people to present their work. What we chose to highlight, what we knew was key, was cities building an analytics solution that could be shared with other cities.” “That’s what enabled us to scale,” Fitzgerald reported back, two years after I’d asked how fast CCP would be able to grow. In 2019, the program was in one thousand cities. Not every city on the planet (there are more than four thousand with populations in excess of one hundred thousand), but a long way from fifty, and with practically the same tiny team.33

What Would We Do without Each Other?

Platform thinking underlay the One Fund Boston, too. I didn’t know to call it that at the time. For weeks after I brought home that first black bag of money, the contributions poured in. What John Hancock, the marathon’s signature sponsor, had started with a $1 million contribution was echoed in numbers large and small. I had opened some of the envelopes early on, before the volunteers swelled in number and the bankers came in and automated the process. Each check or handful of coins was salve on a broken city’s wounds. But for all the gifts that came in (from individuals, from corporations), the ones that struck me most were proceeds from lemonade stands and local running races and other initiatives people had taken upon themselves to organize as complements to ours. Mostly by accident, and maybe with a sliver of prescience in a week full of quick decisions, we too had created an innovation platform.

In fact, there was a crush of requests to build other side efforts in order to support ours. We heard from local groups and national retailers. We fielded plans to put the One Fund Boston logo on garbage cans and to use the fund’s name for a concert full of headliners. There were two possible reactions to these requests. The first was to lean toward control and ownership. That is, it’s our logo, our name. We want to protect the integrity of both. We want to ward off fraud. Don’t say your work is part of ours. The second approach was to cede that control. It’s a scarier approach. It says, Go ahead. Plug in. It says that even though you don’t work for us, we are working in the same direction. It says that even though we don’t know you, we think you’ll have something to add. In the early days of the marathon response—and probably due more to the limits on our own time and capacity for screening all the potential partners than to any special wisdom—we chose the second reaction. We chose a platform approach, and a rather open one at that. In all the hundreds of activities that sprang up to support One Fund Boston, there was hardly a contract or memorandum of understanding to be found.

One Fund Boston was a platform in another way too, a way that was more subtle and that touched the heart. The fund was, in a pretty straightforward manner, a transaction platform; straightforward in that it was an intermediary that made it easier for companies and individuals to donate and for survivors and the families of the victims to receive those donations. At One Fund Boston, we weren’t fundraisers. We didn’t ask for money. We just made sure it went from where it came from to where it should go. And it’s easy to see in that way the cross-side network effects: the more contributors there were on the one side, the more donations to the One Fund Boston community on the other side. But there were same-side network effects, too. The members of that One Fund community helped each other. Of course, we all hated that anyone at all had been killed or hurt in the attacks, but it was true that the many survivors provided value to each other. This was a lesson imparted to us from those who had helped after the 9/11 attacks: that for all the ways the money helps, the bringing together of the community through the process is really what heals. And so there were small things and important things we did to support those on that side of the platform, to help make sure they could help each other. On the one-year anniversary of the attacks, and around the same time that those two survivors had told me “you have to show people that government can do new things,” one of them, Patrick Downes, told a gathered audience and the world: “We chose to love, and that has made all the difference.”

He continued: “To our fellow survivor community, what would we do without each other? We should have never met this way, but we are so grateful for each other. We have shared our despair, sense of loss and challenges, as well as our hope, gratitude, and triumphs. We have been there for each other, and we will continue to be there to pick each other up and celebrate milestones for years to come. Most of all we will cherish the bonds of mutual admiration. And to those who continue to struggle through despair, ongoing medical care, and the prospect of heart-wrenching surgical decisions, don’t forget for a second that we will be there for you at a moment’s notice. . . . I am so proud to be a Bostonian, because I am so proud to be connected to all of you.” Downes closed his address, choking back the tears he had for the “fallen angels” who didn’t survive that day, while expressing his love for the bonds of his city.34 Boston Strong was the post-attack rallying cry. Boston-as-a-platform would have been a lot less eloquent, but no less accurate.

Platforms Fail, Too

“The idea of government as a platform applies to every aspect of the government’s role in society,” Tim O’Reilly said, and we should look for places to apply it.35 On balance, there is evidence that platform businesses grow faster, more profitably, and more productively; they do more with less.³⁶ Which is what we’ll need to do by necessity, if not also philosophy, to address the long list of public challenges we face.³⁷

But the same evidence shows that most platforms fail. Three scholars compiled data on platform efforts dating back to 1995, and analyzed not only the 43 that survived but also the 209 (or 83 percent) that didn’t.³⁸ Platforms are exercises in possibility, too.

Worse yet, platforms can succeed in ways that hurt the public. Even the ones that have survived can lay only contingent claim to success. Think of broken laws, violated trust, concentrated power, double-edged swords. Platforms may leave out ecosystem members, drive up resentment, and antagonize government agencies if they “don’t keep their power or ambitions in check.”³⁹ Governments as platforms can—and will, and have—too.

I witnessed a platform that failed in many of these senses. In the summer of 2018, I was in Beijing to try to understand how ofo had gone from being a school-based startup to a bike-share behemoth in a matter of months, topping an all-out market-share battle with its rival Mobike. The company had provided billions of bicycle rides in pursuit of its own “what cities won’t we be in” mission: “Anytime, anywhere, a bike to ride.” It had put 6.5 million bikes into the streets of China in its first year and a half—and since, millions more.

Bike sharing wasn’t new when ofo launched on the campus of Peking University in 2015, but its dockless model mostly was. That model gave riders the convenience of leaving bicycles practically anywhere they wanted in the city. And they did. Soon the bikes were everywhere. Blocking sidewalks. Showing up in rivers. Stopping up alleyways. The city of Hangzhou confiscated twenty-three thousand of them over a short period, and the staggering pictures were only previews of more to come.⁴⁰ Thousands of bikes piled up in Shanghai, thousands more outside a factory in Xiamen. There shouldn’t be such a thing as a “bike graveyard,” and the photos of them were haunting.41

Looking at them, you can’t help wondering how something like this could ever happen. How could it? On the list of reasons: The network effects were never as strong as the company thought they were, especially given the presence of so many other bike-share companies. Sure, it helps me, if I’m a rider, that some other user has ridden a bike to the building where my meeting is letting out, allowing me to pick it up from there. But given the presence of so many other companies, and given how easy it is to switch to any other of many competitors, it’s not that helpful. So on the same side of this platform, there wasn’t considerable power at play. Moreover, the company and others like it operating at the same time failed to cultivate a second side of the platform early enough. The bike maintainers, the reallocators . . . there could have been a whole ecosystem of people engaged to keep the bikes functioning and where they were supposed to be. But the efforts at this, while eventually substantial, were too little, too late.

There were initiatives to use financial inducements to get riders to leave bikes in better places. And there was even talk of integrating ofo’s data with the government’s, so they could both weed out problem riders, repeat offenders, and the like—shared data and application programming interfaces (APIs) being the toolkits of platform players. And what if the company had pursued such integration? If it had, I suspect it would have ended up as part of China’s social-credit system—a government platform, really—for monitoring and ranking China’s citizens.

Six months after my visit, ofo was essentially over. Now the pictures were of users lining up outside headquarters trying to claim deposits they had provided at sign-up. The company was headed for bankruptcy.⁴²

Platforms as Government?

It may seem relatively clear that platforms for monitoring and quashing dissent are bad and platforms for channeling money to attack survivors are good, but one of the things that makes government as a platform so challenging is that many of the cases feel more gray. Aadhaar is not China’s social-credit system, but it has raised significant concerns about big-brother government, about privacy, and about disenfranchising people without the ID numbers. Efforts to make an Aadhaar number mandatory for accessing certain services generated substantial pushback that reached all the way to the Indian Supreme Court.43

Even Waze’s CCP raises deeper questions, questions Paige Fitzgerald tackled the first time I met her. Ostensibly for the purposes of better urban planning, some cities had requested that CCP share data involving speeding trends. Waze declined to share the data with governments, and Fitzgerald explained why: “Even my father teases me that he will soon be getting speeding tickets in the mail because tech companies work with police departments. He won’t because of us; we won’t share it.” She added that her father seemed slow to believe her. Waze providing CCP as a platform with cities as one side, and Wazers and editors making up another side, raises a related squeamishness. One student of mine put these words to it: “Do you want city hall to be one side of the platform, or do you want government to be at the center? I think that’s where the tension is. Are we ceding governments to now just be part of a platform that other people run? Or do we want government to be the platform?”

Indeed, these questions won’t always come in headline form. Occasionally they will be on the order of the Facebook–Cambridge Analytica scandal accompanied by congressional and parliamentary hearings and promises of reform. But more often they will be in the form of looming policy questions. How will we manage the e-scooters that have arrived in town? Should we, government, create an API to help gather data on their status and location? Should we, tech companies, participate? These were the questions in the air in Los Angeles in the spring of 2019.44 Should we, government, partner with Uber on public transit, as Denver’s authorities and those in twenty other cities had by the summer of 2019, so that the company might sell public transit tickets through its app or even provide “public” rides where busses didn’t reach?⁴⁵ When Uber’s CEO says he wants the company to become the Amazon of transportation, a platform for much of the transit sector, we should take him seriously, and literally, and have the tools for deciding.⁴⁶ Which brings up a final case in point: Amazon’s overlap with government as a platform gets headlines when it’s about a $10 billion cloud contract for the US Department of Defense, or when it’s about the company’s Rekognition facial-recognition tools, and especially when it’s in connection with surveilling neighborhoods or identifying immigrants.⁴⁷ But Amazon has more than five thousand government agencies using Amazon Web Services (AWS) for everything from police-records management to identity solutions to election tools, and that number will grow, as will the number of underlying functions.⁴⁸ Competition from Microsoft’s Azure and others will make that growth not completely unchecked, but it’s possible that within a decade AWS will be a sort of de facto operating system of government. How many cities—and small towns and large states and federal government agencies—won’t they be in? Amazon as government as a platform will offer tremendous efficiency, reliability, and security, and yet it will continue to reraise the questions of who, when it comes down to how government works and what it does, is really in charge. It is time we relearn the techniques of government as a platform and past time we raise the scrutiny on platforms as government.

How Many Kids Won’t We Save?

“How many cities won’t we be in?” is a rallying cry. It says, “no more” to programs that start and stagnate, that dwindle, or that grow only in line with the resources they are given. It’s the answer to the question in a spring budget meeting about growing that new program for the homeless or the one that’s cutting down on school absenteeism. “How many corners won’t we reach?” “How many kids won’t we save?” It’s money. It’s roads. It’s identity. It’s reach.

And that kind of ambition for scale comes with a set of tools. Waze’s Gai Berkovich told me, “I think it’s kind of witchcraft to manage a community.”49 He meant—I think—that it’s part science and part art; what you do as the platform provider and what you leave to the ecosystem. But government as a platform comes with a set of earthly tools, too: hardware, software, rules, and processes that can accentuate positive network effects and mitigate the negative ones. And it comes with, if we hone them, a set of strategies for governments’ getting there, involving building platforms or, on occasion, borrowing them from private partners or lending them out.