Latest / The Fiscal Policy Podcast with Fexingo: Government Budgets, Stimulus, and Public Spending / Why Government Software Projects Fail 70 Percent of the Time
Transcript
- Lucas: Seventy percent. That is the failure rate for large government IT projects globally, according to a 2023 Oxford study. That is not a typo. Seven out of ten major public-sector software initiatives either bust their budget, miss their deadline by years, or get abandoned entirely. Luna: And we are not talking about small-town website redesigns. We are talking about systems that cost hundreds of millions, sometimes billions, of taxpayer dollars. Lucas: Right. The poster child for this is the UK's National Health Service — the NHS — and its attempt to build a centralized electronic patient records system. The project launched in 2002 with a budget of roughly 6 billion pounds. By the time it was officially abandoned in 2011, the government had spent around 10 billion pounds, and not a single region had a fully functioning system. Luna: Ten billion pounds for nothing. That is roughly 13 billion dollars at the time. How does something like that even happen? Lucas: It is a perfect storm of structural problems. First, the procurement process. The NHS put out a single massive tender — a monolithic contract worth billions — and awarded it to a handful of big IT contractors like Accenture, CSC, and BT. The system was supposed to do everything: store patient records, schedule appointments, manage prescriptions, link across hospitals. That level of ambition, combined with a fixed-price contract, meant the contractors had zero incentive to admit early on that parts of the design were impossible. Luna: So they just kept building, knowing it wouldn't work, because backing out would cost them the contract. Lucas: Exactly. And the government's project management office was not staffed with experienced software buyers. They were career civil servants who had never run a technology program of this scale. So when the contractors said 'we need more time, more money,' there was no one to push back with credible technical questions. Luna: That sounds like a recipe for disaster. But is this just a UK problem, or is it universal? Lucas: Universal. The US Department of Veterans Affairs spent over 16 billion dollars trying to modernize its electronic health records system using a single vendor, Cerner. That project is still not fully operational and was paused in 2023. Australia's My Health Record system, Canada's Phoenix pay system for civil servants — the pattern repeats. Huge contracts, rigid requirements written years before any code is written, and no room for iteration. Luna: It is interesting that you mention that the project management office lacked technical expertise, because that is actually something we see a lot in this show — the government often does not have the in-house talent to oversee complex contractors. And if walking through the economy with us has made something click, or if you just enjoy having these stories told without ads, listener support is what keeps us ad-free. You can find us at buy me a coffee dot com slash fexingo. No pressure, just a quiet way to keep the show going. Lucas: Appreciate that. So back to the IT problem — there is a better way, and some governments are starting to use it. It is called modular contracting. Instead of one giant contract for a monolithic system, you break the project into small pieces — a scheduling module, a billing module, a patient portal — and you contract each piece separately, often using agile development methods with short delivery cycles. Luna: And you test each piece before moving on to the next. So if one module fails, you have not bet the whole farm. Lucas: Right. The US federal government's 18F and the UK's Government Digital Service have pushed this approach. For example, the state of California recently used modular contracting to build a new child welfare system. They awarded small contracts to multiple vendors, used open-source components, and delivered working software in increments. It is not perfect, but the failure rate has dropped significantly. Luna: Are there any notable successes we can point to? Lucas: One that stands out is Estonia. They built their entire digital government infrastructure — e-residency, digital ID, tax filing, health records — using a modular, open-source approach from the start. Their total spend over a decade was less than what the NHS wasted in a single year. And it works. Citizens file taxes in three minutes. Prescriptions are digital. Voting is online. Luna: So the lesson seems to be: go small, go modular, and build in-house technical capability. But why is that so hard for big governments to adopt? Lucas: Incentives. Procurement officers are rewarded for following rules, not for delivering outcomes. A modular contract requires more oversight, more contract management, more risk acceptance. A single monolithic contract is easier to write, easier to award, and if it fails, the blame is spread across a giant vendor. It is the safe bureaucratic choice — even if it wastes billions. Luna: So the real fix is not just technical. It is cultural and structural. Governments need to change how they buy software, not just what software they buy. Lucas: Exactly. And that is starting to happen. The US passed the MISSION Act in 2023, which requires federal agencies to consider modular approaches for large IT projects. The UK's Government Digital Service now mandates a 'digital assurance' review before any project over 10 million pounds can proceed. Small steps, but they point in the right direction. Luna: It feels like we are slowly learning that you cannot build a skyscraper with the same methods you use to build a shed. Government IT is a skyscraper, but they keep treating it like a shed. Lucas: That is a great analogy. And the cost of that mistake is not just wasted money. It is delayed healthcare, broken services, lost trust. When a government IT system fails, real people suffer. So getting this right matters far beyond the budget line. Luna: Well, let us hope more governments take the modular path. And thank you for walking through that with us. Lucas: Thanks for listening.