Government digital services fail mostly for organizational reasons — fragmented ownership, procurement that buys compliance rather than outcomes, and no team accountable for the user's task — and the documented fix is the same everywhere it worked: a named, empowered digital team measured on task completion. When the US rebuilt Healthcare.gov's failed 2013 launch, the follow-on surge effort and the US Digital Service it spawned became the template; the UK's Government Digital Service reported cutting its government web estate from roughly 2,000 sites to under 200 after consolidating onto GOV.UK in 2012-2014, per GDS's own published reporting. The technology was never the hard part.
Why does a simple government form take months online?
Because the form is not the system. A benefits application touches an eligibility engine, legacy case-management software sometimes decades old, privacy statutes that dictate data handling, and procurement rules that make changing any vendor a re-competition. Each constraint lives in a different office, and the website is the last layer on top. The Government Accountability Office has kept federal legacy modernization on its high-risk list since 2019, documenting systems where agencies lacked complete inventories of what they even run, per GAO's published high-risk series. When nobody owns the whole path, the path breaks in the middle where no one is measured.
What does procurement have to do with bad websites?
Everything. Traditional government IT buying specifies requirements in advance, awards to the lowest bidder who meets them, and pays on delivery of documents rather than working software — a structure that punishes learning during the project. The federal alternatives developed over the last decade — modular contracting, agile delivery pilots under the Technology Transformation Services, the USDS-created 18F's engagement model — exist precisely because monolithic contracts failed repeatedly, per TTS's own published materials on modular procurement. States vary widely: California's Office of Digital Innovation and New Jersey's Office of Innovation, both created in 2019-2020, were explicit attempts to build in-house capacity rather than rent it, per each office's establishing documents.
Which turnarounds are actually documented?
| Jurisdiction | What was built | Documented result |
|---|---|---|
| United Kingdom | GOV.UK consolidation, 2012-2014 | ~2,000 sites reduced to under 200, per GDS reporting |
| United States | Healthcare.gov surge, 2013; USDS founded 2014 | Site stabilized; team model institutionalized by executive order |
| Estonia | X-Road data-exchange layer, from 2001 | Base for near-whole-of-government e-services, per e-Estonia documentation |
Each case names a jurisdiction and its own numbers; what transfers is the pattern — a stable team with authority — not any country's legal specifics. Estonia's X-Road, for instance, depends on a legal identity framework most countries do not have.
What separates the projects that work?
Four traits recur in the documented successes. A single accountable owner with budget authority, not a steering committee. Measurement on task completion — the percentage of users who finish the actual job — rather than uptime or tickets. In-house technical capability, even a small one, so the government can judge what vendors propose. And iteration: GOV.UK launched as a beta and improved in public for years, per GDS's own history of the project. None of these are technology decisions, which is why buying a new platform rarely fixes a broken service.
What should reformers watch instead of launches?
Press announcements celebrate launches; the record lives in the queue after launch. Watch the service's own published performance dashboards — US federal agencies must report digital-service metrics under the 21st Century Integrated Digital Experience Act of 2018 — and the procurements: a service that goes quiet after launch and re-appears in a three-year re-competition is telling you the model, not the site, is the problem. What the evidence establishes: ownership and measurement move outcomes. What it does not: that any particular country's structure can be transplanted without its legal and institutional soil.
For more context, read FedRAMP: How a Cloud Service Gets Federal Authorization.