Questions Explored
- How can public trust in official data be rebuilt after a period of strain and scrutiny?
- What does “quality” mean in the context of national statistics in the age of AI?
- How should data institutions balance ambition with capability?
- What can other sectors learn from the ONS’s approach to transparency, accountability, and recovery?
“The truth still needs a reference point, and that is the role of national statistics.” – James Benford, Director General, Economic, Social and Environmental Statistics at the Office for National Statistics (ONS)
From Data Expansion to Trust Erosion
The presentation began with a striking reminder of data’s accelerating influence on the UK economy: Nearly a quarter of businesses now report using some form of AI, and it was less than a tenth only two years ago. Yet, he cautioned, this transformation depends entirely on trust as AI is only as good as the data it learns from, and that data is only as good as the trust behind it.
He positioned the Office for National Statistics (ONS) as the reference point for truth in the age of algorithms. But acknowledged the institution had faced a significant test of credibility since the pandemic. The intense demand for real-time data during COVID-19 had stretched the ONS’s systems to breaking point. “By 2022, we were doing too much, too quickly, and on too many fronts,” he admitted. “The quality of our regular output suffered, and the programmes we launched failed to achieve our mission.”
Independent reviews found both failures within the ONS’ culture while recognising the expertise and dedication of ONS staff and the institution’s capacity for renewal.
Resetting the Mission
In response, the ONS has fundamentally reframed its purpose with a new mission statement: “To deliver trustworthy, independent, high-quality statistics that underpin the UK’s most critical economic and societal decisions.”
This new focus means deliberately “stopping or scaling back analytical work that is not required to improve the quality of our statistics.” Instead, the ONS is concentrating on core outputs such as refining economic indicators, improving survey methods, and rebuilding the reliability that underpins fiscal and policy decisions.
The shift has already created tangible capacity. “We have reallocated resources to create 150 new roles – 100 of which will be filled this year – in support of our economic statistics recovery plan,” Benford confirmed. The emphasis is on stabilising before scaling: focusing on labour market statistics as the test case for improvement and investing in the technology and people needed to sustain that progress.
Governance as the Architecture of Trust
Benford was explicit that governance and not bureaucracy is the mechanism for restoring confidence. “Good governance is not red tape… It is the architecture of effective delivery and of trust.”
To achieve this, ONS has restructured around a single operational line of sight: “We’ve brought together all aspects of statistical production, from surveys through to publication, under common management.”
Externally, governance has been reinforced through a new steering group including the Bank of England, Treasury, Office for Budget Responsibility, and Cabinet Office. These partnerships ensure alignment between the statistics the ONS produces and the policy decisions they inform.
The next frontier for ONS is technology rationalisation. Since taking on the role, Benford has “been struck by the sheer number of different solutions for very similar problems. We will agree the strategic solutions and migrate to them quickly.”
Openness and Accountability
For Benford, recovery cannot rest on structures alone as it depends on culture. The ONS has therefore embedded three principles to restore trust:
- Creating space for excellence
- Working in the open
- Supporting people
Working in the open is central to this cultural shift. Benford announced that the ONS will publish quarterly public reports showing progress against recovery goals, including where things go wrong. “Alongside correcting the record, we use our blog to explain what happened and what we’re doing to prevent a recurrence,” he said.
This transparency is already changing perceptions: the reinstatement of producer price data (suspended earlier after a serious error) was presented as a proof point. “The reinstated data are better quality than before the error occurred, and our processes are stronger so there will not be a repeat,” Benford noted.
Leadership That Listens and Learns
Much of Benford’s message centred on leadership behaviour. “Each leader in the organisation is now signed up to a commitment to build trust with our teams, to encourage challenges, to listen, to take decisive action on our priorities.”
This shift includes embedding 360-degree feedback into performance assessments and reframing accountability. “When things go wrong, my starting point is that there’s likely to be a systems problem, which means responsibility rests with leadership, not the team or individual involved.”
To strengthen its learning culture, the ONS is looking outward to sectors such as aviation, nuclear energy, and healthcare as these are industries that have mastered learning from error. These analogies underscore a deeper ambition: to make statistical production as robust and safety-critical as any regulated industry.
Innovation with Purpose
While the recovery plan focuses on fundamentals, Benford outlined a pragmatic innovation vision. Automation and AI are being deployed internally to improve efficiency and accuracy: “A bespoke AI tool is already freeing up hundreds of hours per year by automatically classifying products, occupations and sectors based on free text.”
The ONS is reducing its reliance on traditional surveys by integrating new data sources, from VAT records to supermarket scanner data. “We’re expanding our use of other data beyond surveys, and using it to replace surveys where possible,” he said, describing these advances as the foundation of a more adaptive, data-rich statistical system.
Rebuilding Trust as a Collective
Benford closed with a call to partnership across the data community. “Many of your companies feed us data to power national statistics. Many of you are users of ONS data. You will, I am sure, have ideas and contributions to make.”
Rebuilding trust, he stressed, is not a solo effort but “a collective endeavour, a team sport.” His message to the audience was clear: “Your National Statistics Institute needs you. Get in touch, get involved, and we will recover faster.”
Practical Learnings
- Rebalance ambition with capability. Pursue innovation only when operational foundations (quality, governance, and capability) are secure.
- Build transparency into delivery. Publish progress, learnings, and even errors; trust grows through visible accountability.
- Design governance as enablement. Treat governance as the structure that supports clarity and alignment, not as compliance overhead.
- Invest in leadership culture. Encourage challenge, feedback, and psychological safety as trust is modelled from the top.
- Collaborate across the ecosystem. Data as public good depends on shared ownership between institutions, businesses, and citizens.
