HealthCare.gov: How a Mandatory Big-Bang Launch Let Only Six People Enroll on Day One

HealthCare.gov shipped as a single, mandatory, all-at-once launch — built by dozens of loosely coordinated contractors with no one owning end-to-end integration, tested for days instead of months, and pointed at the entire country simultaneously with no way to phase in demand. When the federal marketplace opened on October 1, 2013, the site buckled almost immediately: by day's end, per internal documents later released to Congress, only six people nationwide had completed an enrollment, despite years of work and hundreds of millions of dollars in contracts.

The Affordable Care Act set October 1, 2013 as the statutory start of open enrollment. HealthCare.gov — the Federally Facilitated Marketplace — let residents of the roughly three dozen states that declined to run their own exchange compare plans, verify identity and income, and check subsidy eligibility against IRS, Social Security, and state Medicaid data, all behind a mandatory account-creation step. The build spanned reportedly around 60 contracts to about 33 vendors, per post-launch case-study analyses, with CGI Federal holding the largest contract — originally about $93.7 million in December 2011, growing to roughly $292 million by launch, per Wikipedia's sourced account.

The system was reportedly designed for 50,000–60,000 concurrent users. One day before launch, an internal test found the site struggling at around 1,100 simultaneous users, according to reporting on the launch; a separate identity-verification component was apparently load-tested to only about 2,000 concurrent users beforehand, per case-study accounts. Henry Chao, the CMS deputy CIO running the technical build, had told colleagues months earlier — per contemporaneous reporting on his House Oversight testimony — that he hoped the launch wouldn't be “a third-world experience.” On launch morning, an estimated 250,000 concurrent users tried to get in — roughly five times the design target.

The result: HealthCare.gov reportedly drew more than four million unique visitors on October 1 and became largely unusable within about two hours. By the close of that first day, according to internal documents later released by the House Oversight and Government Reform Committee and reported by NBC News, just six people nationwide had successfully selected a plan. Some contemporaneous estimates put the completion rate at around 1% of the roughly eight million visitors that first week. Officials withheld official enrollment numbers for weeks; by November 13 fewer than 27,000 people had enrolled through the federal marketplace, and by November 30 that had grown to more than 137,000, per figures reported at the time.

The White House brought in Jeffrey Zients to lead a fix and assembled a “tech surge” of outside engineers, including Google site reliability engineer Mikey Dickerson, from mid-October. Officials said the site could handle around 35,000 concurrent users by December 1, 2013, and enrollment recovered enough that roughly 1.2 million people had reportedly signed up nationwide by December 28, and more than 8 million had selected a plan by the end of the first open-enrollment period on March 31, 2014, per HHS figures. CGI Federal was replaced as lead contractor by Accenture in January 2014. Total cost estimates vary by accounting cutoff: roughly $834 million per Sylvia Mathews Burwell in early 2014, about $1.7 billion per an HHS Office of Inspector General report that August, and $2.1 billion per a Bloomberg News estimate that September. A July 2014 GAO report, summarized by the Senate Homeland Security and Governmental Affairs Committee, found CMS staff had improperly authorized contractors to spend more than $30 million outside established oversight procedures. HHS Secretary Kathleen Sebelius resigned in April 2014; the White House created the U.S. Digital Service that August, with Dickerson as its first administrator.

No one owned the whole system end to end. CMS acted as its own systems integrator despite lacking the technical depth to do it well, while CGI Federal — the largest contractor — reportedly saw its role as building the front end, not integrating everyone else's modules. With work split across dozens of contracts to dozens of vendors, nobody was accountable for proving identity verification, the plan finder, the subsidy calculator, and the federal Data Services Hub all worked together under real load, rather than each passing its own isolated unit tests.

Scope kept expanding against a fixed date. A late decision to require every visitor to create and verify an account before browsing plans — reportedly pushed so the administration would have concrete user metrics — added a hard, stateful bottleneck at the front door of the system. Meanwhile a CMS letter sent in August 2013 flagging CGI's code quality and testing gaps was, per documents later obtained by the House Oversight Committee, retracted a day later in favor of continuing to collaborate rather than escalate. Warnings existed; the incentive to keep moving toward the statutory date kept winning.

Testing was compressed into the final weeks and confined to individual modules, not the full path a real user would take. A true end-to-end load test — the kind that would have surfaced the gap between the 50,000–60,000-user design target and the roughly 250,000 who showed up — reportedly never happened before launch; the closest thing was a single stress test the day before, which already buckled at a fraction of that traffic. Because the date was fixed by statute, not readiness, there was no staged rollout, no waitlist, no way to gate demand once it arrived.

This pattern doesn't require anyone to be careless. A launch date picked for political, marketing, or fundraising reasons — not engineering readiness — becomes immovable once it's announced, and every later decision bends to preserve it instead of questioning it. Multi-vendor integration debt is invisible on a status report where each contractor's module shows green; it only surfaces once real, uncoordinated traffic hits every part of the system at once. And raising a hand to say “we are not ready” gets harder, not easier, the closer the date gets and the more has already been spent — exactly when the warning is most needed and least welcome.

A launch date is a scope constraint to negotiate against, not a finish line to build blindly toward. If a system can't be tested end to end, under realistic load, well before the real deadline, then the deadline — not the testing — has to move, or the launch has to be staged so failure is contained instead of total. Every irreversible, all-at-once release bets that nothing in the untested integration breaks on the one day you can't afford it to; the fix is rarely “try harder to hit the date” — it's building an organization that can say no to a big-bang launch.

Most of what turned HealthCare.gov's launch into a national story was avoidable with ordinary engineering discipline applied earlier and honestly. The checklist below reflects the specific gaps the record shows.

Several factors compounded each other: demand roughly five times reported design capacity, a mandatory account-creation step added late that bottlenecked every visitor, components from dozens of contractors never proven to work together under full load, and a statutory date that left no room to delay or phase in traffic once problems appeared. Any one alone might have been survivable; together they produced a near-total failure on day one.

The White House assembled a “tech surge” of outside engineers, including Google site reliability engineer Mikey Dickerson, from mid-October 2013 to rebuild monitoring and fix the worst bottlenecks. Officials said capacity grew to around 35,000 concurrent users by December 1, 2013, and enrollment recovered enough that more than 8 million people had selected a plan by the end of the first open-enrollment period on March 31, 2014, per HHS figures.

In August 2014 the White House created the U.S. Digital Service, appointing Mikey Dickerson as its first administrator, to bring practices like incremental delivery, real end-to-end load testing, and single accountable technical ownership to future federal projects — an institutional response traced directly back to the HealthCare.gov launch.

This case study draws on: Wikipedia's sourced entry “HealthCare.gov,” for contracting history, capacity figures, cost estimates, and the CGI-to-Accenture transition (en.wikipedia.org/wiki/HealthCare.gov); NBC News, “Only 6 able to sign up on healthcare.gov's first day, documents show,” for the day-one enrollment figure sourced to House Oversight and Government Reform Committee documents (nbcnews.com); the Senate Homeland Security and Governmental Affairs Committee's summary of the July 2014 GAO report on CMS contract mismanagement (hsgac.senate.gov); CBS News, “Healthcare.gov plagued by crashes on 1st day,” for launch-day conditions (cbsnews.com); and Henrico Dolfing's case-study analysis “Case Study 17: The Disastrous Launch of Healthcare.gov,” for contractor structure and pre-launch testing scope (henricodolfing.ch). Figures that vary across sources are presented as ranges or hedged with “reportedly.”