At 18F, The U.S. Looks to Fail Fast on Government IT Projects Instead of Failing Big
BY Alex Howard | Thursday, April 3 2014
Two weeks ago, a front page story in the Sunday edition of the Washington Post reminded the nation of the continued existence of a sinkhole of bureaucracy in federal government where retirement filings were still being processed on paper. David Fahrenthold painted a vivid picture of an underground facility where workers at the U.S. Office of Personnel Management disappeared into the dark bowels of the Earth to work.
When "The Cave" still exists and efforts to address veterans benefits struggle against beast of bureaucracy, it's easy to wonder if federal government IT will ever get better, despite greenshoots of innovation and the success of the Consumer Financial Protection Bureau in leveraging technology to accomplish its mission.
That's particularly true today, when the Obama administration's various attempts to modernize government through automation and information technology have been swallowed up in the public imagination by the troubled launch of Healthcare.gov. Even after a big fix by a "trauma team got Healthcare.gov working for the remainder of the open enrollment period for health insurance, the administration's record on government IT is far different today than it was a year ago.
After five years of President Barack Obama being hailed as the "first Internet president," after a 2008 election campaign that tapped into the Web, text and email in unprecedented ways and riding to re-election in 2012 with a dream team of data nerds to give its field teams an organizing edge, his administration now faces historic lows in trust in government, including its ability to do "big things."
Over a long, long winter in DC, many people have been wondering if the federal government can avoid another Healthcare.gov debacle, given the abysmal track record of big IT projects in the public sector. After following the pace for years, it would be easy to conclude that nothing will make a dent in a broken procurement system, tangled human resources process, heavy regulatory burden, dearth of project management expertise and a culture focused on avoiding risk instead of managing it.
To check in on what approaches are being taken to address them, I interviewed the United States chief information officer, Steven VanRoekel.
"You have to look at the way that we build things traditionally," he told me last week, in his office in Washington. "Faced with a problem, government's normal motion is to go do a monolithic internal effort, to go get a single procurement professional to write a single monolithic procurement that goes out into the street to a single monolithic vendor, who then produces a single monolithic solution that is probably destined to fail."
That's how federal technology has been bought, built and maintained for decades. To be blunt, there is no silver bullet for fixing what ails government information technology. There are practical barriers to better government Web products that anyone seeing better outcomes must acknowledge and take on directly, from culture to procurement to hiring to standards.
"Those in government who don’t understand just how much implementation matters to the ultimate success of their policies are doomed to create terrible user experiences for citizens," wrote Tim O'Reilly, in a recent essay on government IT. "Government has long gotten away with this because there is no alternative, and people are forced to get through whatever awful application or service is provided. It is all too easy to forget that there is an alternative: it’s called apathy, distrust of government, and a failure of the goals of the programs that policy makers set out to achieve."
That said, there are reasons for hope that the information technology reforms that the Obama administration has quietly advanced over the winter will lay a foundation for future success. As I reported earlier this month, 18F, a new development unit within the U.S. General Services Administration (GSA), named for its location, is primed to tap into the success of the United Kingdom's Government Digital Services.
"This was a long time coming, as are a lot good ideas," said Andrew McMahon, senior advisor to GSA Administrator Dan Tangherlini, in an interview. "We are attacking, first and foremost, the problem of core capacity in government from a human capital perspective. We want to bring the most talented people in government, the private sector, and foundations together on a team and have them really help agencies be successful."
If 18F can prove that government can buy, build and deploy technology like a 21st Century technology startup instead of a lumbering 20th century institution, the model could not only offer a roadmap for other national and state governments but upend a federal contracting world that former U.S. chief information officer Vivek Kundra memorably described as an IT cartel.
Checking into 18F
"There are two goals," said McMahon. "First, build great user-centric technology for government, specifically for things out facing the public. The second order, just as important, is to prove an agile approach and lean startup can work in government, at scale."
That's a mighty big "if," but it's a hypothetical worth considering in light of the United Kingdom's success in outfoxing a broken procurement system after its own mammoth 12.7 billion pound government IT debacle at the National Health Service. While IT at the NHS remains broken, the team behind Gov.uk has been steadily rolling out new, beautiful IT services for years now, with more to come.
"When you commission at the point of need, there is a massive change in emphasis," said Mike Bracken, digital director for the United Kingdom, in an interview. "It seems like a tweak to the fundamental system to say 'let’s procure better.' What we say is don’t procure at all if you can’t change the service and iterate daily."
The United Kingdom's Government Digital Services (GDS) team's approach has been grounded in a simple strategy: hire well, deliver quickly, take on hard problems, focus on end-users, move to the cloud and keep costs low. Specifically, the GDS hired senior technologists at high pay grades, enabling the unit and its suppliers to buy, build and fix quickly. They've also adopted agile development and an open source approach to coding and collaboration. While these are all familiar approaches in modern startups, in combination they're still fairly radical in government technology, particularly in-house.
The 18F team is following a markedly similar playbook, with shout outs to the GDS team's approach, mixed with an American sense of humor and style. (Their staff titles include API Agent, Counter Agent, Free Agent, Service Agent, Change Agent, User Agent, Agent Schmagent, Reagent, and Agent onGover().) The initial team of 15 is a mix of Presidential Innovation Fellows and technologists on detail from other federal agencies. These folks are, by and large, people who make things, instead of just proposing them. They're hiring more such people, with a total headcount that could increase by sixfold by the end of 2014.
Federal officials that talk about using agile development or applying the "lean startup" might be called out for mouthing buzzwords, as opposed to offering solutions. For instance, despite the efficacy of that approach in developing the front end of Healthcare.gov in the spring, the failed launch of the health insurance marketplace in the fall led dozens of commenters on a feature article I wrote that documented that approach to cry foul.
That criticism, however, should be balanced with recent outcomes. In April 2014, using ">agile development in government is far from a new idea, going back at a decade to A.T. Kearney. In the interim, using agile development to build technology has been
"If you're going to do agile inside the federal government, you have to have partnerships with different functions," said Greg Godbout, lead developer at 18F. "They also have to become agile in the group we share. You can't have finance based upon requirement documents when we're delivering pilots or prototypes in weeks. What happens in a release? People in government are used to 'finished' but we say never say software is finished. We're educating people about a new way. They're shocked at rapid delivery but want to get hands on it, once we start iterating."
Introducing lean government
Government taking a lean startup approach, by contrast, is a much newer idea.
"It is a different way of building and delivering tech," said the GSA's McMahon. "Lean startup and the agile methodology are not widely adopted in government. I don't think anyone would disagree. There will be disagreements about what we're doing and working on. We know we're not the only voice in the room here, but we're committed to building technology that delivers for the American public. They're the ones paying for it. The adoption of those methods into government, from a management sense, is a good thing. If it delivers successful outputs, I can go home as a public servant and sleep well. "
As with many experiments in the nation's labs of democracy, it began bubbling up at the local level, in cities, where Code for America fellows have been applying the lean startup to building tech in Boston and beyond since 2011, when the inaugural class went "through the wall."
Over the past two years, the Obama administration has been experimenting with applying lean government at the federal level, from fellows at the Consumer Financial Protection Bureau and Health and Human Services to Presidential Innovation Fellows to U.S. chief technology officer Todd Park.
I first heard "lean startup" used explicitly by a federal official in 2012, in a presentation on "lean government by Park and former White House director of digital Macon Phillips at the SXSW Interactive Conference.
While the White House presentation in Texas was a bit low on specifics, the basic kernel of the lean idea filtered through: move quickly to get a minimal product or service in front of consumers and citizens, put analytics in place to test their experience, and use that feedback to improve it. (Putting such analytics in place was one of the first things "Obama's trauma team" did when they were brought in to rescue Healthcare.gov.)
"Lean startup is a new way of effecting change management that, frankly, a lot of people in government find a lot more intuitive than waterfall and which is producing better results," Park told me, in 2012 interview. "You don’t think of strategy, policy, ops and tech as different phases of the project, you think of them as different facets of the same exercise."
Park broke down the basics of what might be called "lean government" into three components: building an interdisciplinary team, developing and deploying a minimum viable service, and then iterating and redeploying rapidly in response to user feedback.
"I don’t think folks in government who are the really talented change-makers need to be reprogrammed to think lean," he said. "I think that’s how they already think and how they already act. We just have to actually give them the opportunity to do that."
Lean government isn't about re-imagining huge agencies as startups, from this perspective, but trying a different way of approaching building services, something Park emphasized:
"Government, of course, is not a startup, but change projects in the government are startups," said Park.
Officials at the GSA further emphasized this perspective.
"At 18F we focus on user needs, develop in the open, and we use lean principles and agile methodologies," said Lena Trudeau, associate commissioner for strategic innovations at the GSA, in an interview. "These tools and others are used in government, but they’re not yet the standard. The work of 18F is intended to show what’s possible when we apply known best practices consistently, and share what we’ve built (and what we’ve learned) for the benefit of all. Currently, GSA’s DigitalGov unit provides no-cost platforms and communities of practice to help promote all things digital. 18F builds on this great foundation, and as a delivery unit, also partners with agencies on a reimbursable basis to build and deploy user-centric digital services."
That's in-line with the mission of 18F. The results desired there are quite specific: according delivering digital services in partnership with other federal agencies and creating tools and services that "result in governmentwide reuse and savings, allowing agencies to reinvest in their core missions,"
The words that should jump out in that last sentence are "reuse and savings." Uncle Sam is making a small but considered bet that technology built this way will be cheaper to make and can be redeployed across government: By default, all projects produced by innovation fellows are public domain and released on Github. The active 18F Github account is following the same model.
"18F is about building permanent infrastructure for responsive design," said Godbout, one of 18F's developers. "We will be able to take good projects and continue them after PIFs leave. We share excitement for open source. From our team's perspective, it's one of our core values when we take on a project, heavily weighted about whether it will be open source."
The US CIO was more measured about its use.
"It's got to be about the right tool for the right job," said VanRoekel, regarding sharing open source code. "There's a natural fit for approaches that say let's do something once and then scale that more broadly."
When pushed for more examples, he pointed to security and identity efforts by the administration.
"I think the work we've done on identity management, NSTIC work, where you things once and start to build on that foundation. The work we've done in cybersecurity, I think covers that, to protect a federal network we do once, use often. There's other places where this modular approach can lead to incredible outcomes. It's really about taking something on Github and allowing someone to grab it and fork it into their own solution, or, better yet, just use the core solution in some way that's reusable."
While it's still not clear if all of the "lean startup" approach can be brought into a government context -- despite 18F seems to be committed to applying as much as they can, from citizen-centric design, rapid prototyping using agile development, development of minimal viable services, and testing.
"18F is a new initiative, housed in an existing agency, so it has the benefit of both a start-up culture and a firm foundation," said Trudeau. "Like 'lean startup,' our plan is to quickly deploy working prototypes and iterate rapidly based on customer feedback and analytics. By leveraging low-cost experimentation, we believe we can achieve better results."
VanRoekel maintains that the approach 18F is taking will be a significant departure from the norm.
"This is different because it helps us on our path, our journey, to 'fail fast' versus 'fail big,'" he said. "Failing fast is all about getting things out quickly, testing, iterating, building upon that foundation, and taking the next step, and the next step, and the next step. So many times, when you see these big waterfalls, that monolithic approach in government, where they define these 3 year things and they run out and they actually go 5 years, 6 years, or 7 years, and finally someone smart comes along and shuts the thing down, because some decision made in year 1 had severe impact in year 6."
The big picture
While 18F is small today, it could become something much larger, affecting how government technology is bought, built and maintained. It's going to take years for that impact to ripple through the federal agencies but if you start seeing negative comments and quotes from the marketing and development executives at big IT vendors regarding the quality of the work of "government technologists" and agitation from politicians, you'll know that they've made a difference.
After all, if their mission is really to build better, faster and cheaper, this team will be directly competing with the huge contractors that typically big and build government IT and, if projects like RFPEZ succeed, partnering with smaller tech firms, like the GDS. That could make waves, as it has in the UK.
"There are small number of companies for whom this is their bailiwick that are less than thrilled," said Mike Bracken, with characteristic dry delivery. He says the approach is "vastly cheaper," an outcome that 18F and the White House is clearly focused upon.
"We've seen using open source and open source tools reduce costs," said McMahon. "In the end, this is another conversation where we see tech not as a cost center but as a driver of greater productivity and effectiveness. If we're building effective solutions and technology, the value of investment is worthwhile. There will still be large expenditures on federal IT. if we can spend more efficiently, and get the same or better outcome, good. We are following what we see as the new normal in private sector and expect to apply in the government."
One detail that may not be clear from the outside is that 18F has adopted more than lean and agile from the private sector: they're also explicitly standing up a "devops team." For those unfamiliar, devops stands for "development operations," a response to the increased interdependence of the two functions within mammoth tech companies like Amazon, Google and Facebook.
While this "secret sauce" has been well-known in the private sector for years, it's now entering government.
"It's a great way to get products out the door and deliver," said Godbout. "In the nerdiest, wonkiest way, this devops team is the thing that brings me greatest excitement. If you think about it, we're delivering software at federal government, not just building it. We need this capability, but really delivering in federal government is often the clunkiest piece. We're responding to success in the commercial world. Our devops team is highly educated people, focused on security, infrastructure, and technical bandwidth. We're working on hiring a full time manager for the devops process."
According to Godbout, more infrastructure is coming, including support for application programming interfaces mandated by the digital government plan,like common tools, services and throttling and traffic.
"We plan to reuse a lot of code and share it with the rest of the federal government," he said. "The importance of this infrastructure partnership is creating a continuous deployment environment. When they see something fixed the next day or changed, in a responsive way, people will react. Part of our job to show that responsiveness is our commitment."
Whether all of this leads to better, cheaper digital services remains to be seen, although the UK's model is promising and 18F's FBOpen pilot is a promising start. Every staffer I spoke with about 18F was clear-eyed about the current public perception of government technology, flaws and the issues that underpin them. They're also deeply optimistic about their mission and ability to change minds through developing faster, better technology that looks and works differently.
"We’ll be successful in a year if we’ve (1) scaled the team from start-up to fully operational; (2) partnered with several additional agencies and are shipping great products early and often; and (3) provided a measurable increase in our agency partners’ ability to deliver on their missions," listed Trudeau. "In five years, we hope our efforts to serve as a successful model for procuring, building and delivering digital services are the norm in government IT."
The only way to succeed will be through a relentless focus on delivery, as in the the UK. If they don't, trust in the Uncle Sam's ability to create 21st century government, as opposed to talk about it, will be further eroded.