There is a particular kind of institutional stubbornness that only a government department can produce. I’ve spent years reading about British technology history, and few organisations illustrate that stubbornness, and the strange, lurching progress it eventually yields, quite like HMRC. The story of HMRC digital history and the long road to Making Tax Digital is not a clean narrative of visionary leadership. It is, if anything, the opposite: a two-decade accumulation of false starts, abandoned systems, eye-watering contractor invoices, and occasional genuine breakthroughs that almost nobody noticed at the time.
Self-assessment online: the late 1990s experiment
The Inland Revenue, as HMRC was then known before its merger with HM Customs and Excise in 2005, launched online self-assessment in 2000. The ambition was modest by today’s standards: let individual taxpayers file their returns over the internet rather than posting paper forms. The reality was considerably more fraught. Early adopters encountered browser compatibility nightmares, connection timeouts on dial-up lines, and a system that required users to download and install software before they could even begin. For a public still getting to grips with services like Demon Internet and the culture of Britain’s early web, this was asking quite a lot.
Take-up in those first years was low. HMRC’s own figures showed fewer than 10 per cent of self-assessment filers used the online route in 2001. The department responded by extending deadlines for online filers, a tactic that slowly worked. By 2010, around 6 million returns were being filed online. That number sounds impressive until you remember it took an entire decade to get there, and the underlying technology had barely changed in that time.
The PAYE online disaster nobody remembers
Parallel to self-assessment, HMRC was attempting something far more complicated: digitising PAYE. This is the system through which employers deduct income tax and National Insurance before wages ever reach workers, and it had run on paper-based processes for decades. The National Insurance Recording System, known as NIRS2, is still cited in government IT circles as a cautionary tale. Contracted to Andersen Consulting in the 1990s, the project overran substantially and arrived with significant defects. Staff at what was then the Department of Social Security found themselves manually correcting millions of records.
HMRC’s own internal systems were a patchwork of legacy COBOL code, some of it dating to the 1970s. I’ve read the National Audit Office reports from this period and they make uncomfortable reading: repeated references to data quality problems, manual workarounds, and a department that knew it had a problem but lacked the organisational coherence to fix it. The NAO’s published work on HMRC’s performance during the 2000s catalogues a department under genuine strain.
The 2006 data disaster and what it revealed
In November 2007, HMRC lost two discs containing the personal data of 25 million people, essentially the entire child benefit database. They were sent by internal post to the National Audit Office, and they never arrived. Junior civil servant, unencrypted discs, no tracking. The incident was catastrophic for public trust and forced a reckoning about how the department handled data. It also exposed something more structural: HMRC was sitting on enormous quantities of sensitive information inside systems that had grown organically over decades, with insufficient controls and inconsistent processes.
The data loss was not strictly a digital transformation failure in the conventional sense, but it hardened the argument inside government that HMRC’s technology infrastructure needed root-and-branch reform rather than incremental patches. The department commissioned reviews, appointed new chief information officers, and began making the case for substantial investment in modernisation. It took years before that investment materialised in any coherent form.
Real-time information: the reform that actually worked
The most underappreciated success in HMRC’s digital history is Real Time Information, or RTI, which launched in April 2013. Before RTI, employers submitted PAYE data annually. Under RTI, they submit it every time they run payroll, so HMRC receives information on what every employee earns, in near-real time, throughout the tax year. This sounds bureaucratic. Its implications were substantial. RTI underpinned the rollout of Universal Credit by making earnings data available to the DWP on an ongoing basis rather than in arrears. It also reduced tax credit overpayments, which had cost the Treasury billions.
RTI was not without difficulties during its pilot phases. Some smaller employers and payroll software providers struggled to adapt. But by comparison with almost everything else HMRC had attempted digitally, it worked. It came in broadly on schedule and delivered real operational benefit. For a department that had accumulated a reputation for failed projects, this mattered.
Making Tax Digital: the ambition and the delays
Making Tax Digital was announced by George Osborne in his 2015 Budget with considerable fanfare. The idea was that businesses and the self-employed would maintain digital records and submit quarterly updates to HMRC via approved software, rather than filing annual returns. HMRC framed it as reducing errors, which are disproportionately caused by the one-shot-per-year nature of traditional filing. The argument had merit. Research commissioned by HMRC estimated that avoidable mistakes cost the Exchequer around £9 billion annually.
What followed was a sequence of retreats. The original timetable had all businesses above the VAT threshold on the system by April 2019. After substantial lobbying from accountancy bodies and small business groups, many of whom pointed out that the software ecosystem was nowhere near ready, the scope was narrowed. Making Tax Digital for VAT launched in 2019 for businesses above the £85,000 threshold, and even that rollout produced teething problems. Making Tax Digital for Income Tax Self Assessment has been delayed multiple times; the current confirmed date for sole traders and landlords with income above £50,000 is April 2026, with further tranches following in subsequent years.
The history of HMRC digital history is, in a sense, a history of optimistic timetables meeting the reality of an enormously complex tax code and a diverse business population ranging from one-person freelance operations to multinational corporations. The department is not uniquely incompetent; it is dealing with a genuinely hard problem. When you consider how the UK government built GOV.UK, a relatively clean, user-focused service built on modern principles, and compare it with HMRC’s digital estate, the contrast tells you something about what happens when legacy systems accumulate over decades versus what’s possible when you start fresh. I’d recommend reading about how the GOV.UK project approached digital government from first principles for that contrast in sharp relief.
Contractor costs and the question of who built all this
No account of HMRC’s digital journey is complete without acknowledging the contractor economy that sustained it. Over the years, the department has spent hundreds of millions of pounds on external suppliers: Capgemini, Accenture, IBM, and various others held significant contracts at different points. The government’s own spending data, published through Cabinet Office transparency records, shows HMRC consistently among the largest departmental spenders on technology. Critics argue the reliance on contractors created a knowledge gap, that institutional understanding of the systems lived outside the civil service, in consulting firms, making it harder and more expensive to change anything.
This dynamic is not unique to HMRC. It echoes the NHS’s troubled history with its National Programme for IT, and the general pattern of large public sector technology projects in Britain across the 1990s and 2000s. But HMRC’s scale made it particularly visible. The department processes hundreds of millions of transactions annually. Its systems cannot go down at year-end. The risk aversion that produces over-reliance on established suppliers is understandable, even if its long-term costs are significant.
Where things stand now
HMRC’s digital estate in 2026 is genuinely more capable than it was twenty years ago. The online personal tax account, accessible through Government Gateway, lets millions of people manage their tax affairs without ever picking up a phone. RTI functions reliably at enormous scale. The VAT Making Tax Digital infrastructure is largely stable. The department has moved some workloads to cloud hosting and has been working to decommission the oldest legacy systems, though that process is measured in years rather than months.
The story of HMRC is, I think, an honest mirror of how large British institutions interact with digital change: slowly, expensively, with occasional genuine innovation, and always complicated by the weight of what came before. For anyone interested in the history of UK government technology, HMRC is as instructive as any other case. The preservation of that institutional memory matters too, much as the efforts described in the fight to preserve UK web history remind us that the digital record of these systems and decisions is itself fragile and worth keeping.
Leave a Reply