Devolution needs data to succeed

By Theo Blackwell MBE, Chief Digital Officer for London

---

Rewiring the State moves powers and funding to mayors, strategic authorities and councils. That is what devolution has always meant: the ability to decide, and the money to act. But today this now depends on a third aspect: data. Data decides what a place can identify, target, join up and serve. In the digital age it has become as much a condition of real agency as new powers or money, and a government that does not grasp the data challenge limits its own impact.

NEETs is the example at hand

Take the findings of Alan Milburn’s interim report on young people and work. The data that could tell us who is disengaged, where they are and how long they have been out of the system sits fragmented across the NHS, DWP, DfE, schools, colleges and local authorities. None of it is connected so we can’t get an accurate aggregate picture of the problem, let alone the localised insight needed to intervene early.

Running through every failure Milburn describes as “the broken system architecture” is the fundamental problem of data. Without joined-up data and knowledge, the system cannot see the problem it was built to solve. The same holds for homelessness, long-term health conditions, violence against women and girls, early years support and so on. You cannot prevent a crisis you can only see once it has happened, and you cannot see it early if the information is split across a council, or between councils, NHS trusts and departments that do not share data well.

Three directions of data-flow

The problem runs in three directions at once.

Downwards from government. Central government holds data local government needs and cannot get. Universal Credit is the clearest case (there are others) : councils lost visibility of most benefits claimants in their area after its introduction, having held that information themselves just a decade ago. A coalition of councils has repeatedly told ministers how much could be done at little cost - from Pension Credit take-up to free school meals to homelessness prevention - but although progress inches forwards, it remains hard miles.

Across a place. A household at risk may be partly known to housing, partly to an NHS trust and partly to the police, and fully known to none. Services often cannot see the problem until it reaches crisis point. This is a challenge local government, the NHS and Test Learn and Grow programmes encounter on the ground.

Across a region. Individual authorities see their own caseload and little beyond it. Rare and complex cases are the clearest instance: an authority meets too few to reveal prevention patterns and demand forecasting - though those cases carry the highest cost and the greatest risk of harm.

Different blockers

These flows are obstructed for different reasons, and the remedies differ.

The obstacle to downward flow is Whitehall incentives. Departments are built around budgets, mandates and accountability that run through silos, not across them. There is no career reward for a leader who opens their department’s data to a shared platform, and no measure of whether the data reached the people who could have used it. Practitioners have made this case for years. What has not happened is senior leadership treating data sharing as a delivery imperative.

Horizontal and regional data flow is obstructed by similar issues locally and the way the local government technology market has evolved. The 2025 State of Digital Government Review found councils struggle to make systems interoperable because of legacy suppliers and the cost of APIs, which made systems ‘talk’ to each other. Work on children’s social care shows what that means in one service: a market held by very few products, most safeguarding information still moving as email attachments rather than automated exchange, and incompatible systems and inconsistent identifiers as the barriers councils hit most. Long contracts, individual buying and the lowest proportion of digital professionals in the public sector, leave councils with little leverage. In practice councils work in an environment where they commonly have to pay a commercial supplier extra for permission to move data it owns, about residents for whom it holds statutory responsibility. This is the shape of pretty much the entire local government market for all services and the legacy of separate decisions taken entirely reasonably but in isolation, and why no single council or mayor can undo it.

It’s also why the AI conversation is really a data conversation. Unlocking vast public sector IP across systems containing structured and unstructured data - transactional data, programme evaluations, years of lived experience - and putting it in the hands of experienced practitioners creates huge potential public service value, but will remain illusory unless data can be connected.

If the demand runs this way, why does data rarely top what mayors ask of government for? It’s because data does not naturally scan as power. It presents as an IT problem rather than a devolution one, and is handled several layers down from decision-makers. The centre should not mistake a quiet or latent ask for a critical need

What works, and where it stops

Nonetheless there are examples of change. London’s Data for London programme connects 20,000 urban datasets and enables impactful data services like the Rough Sleeping Strategic Insights Tool with all 32 boroughs, the Greater London Authority, faculty and frontline providers. For the first time the city can see how individuals move through the rough sleeping system as a whole rather than as a series of separate encounters. We can now see the ‘big picture’, assumptions can be questioned and further actions taken.

The Scalable Approach to Vulnerability via Interoperability (SAVVI) programme in the north-west shows the same method applied to standards rather than a single tool. Led by Tameside and Sedgemoor, worked through with MHCLG, DWP, GDS, NHS and the voluntary sector, it produced a reusable framework for identifying people at risk that other councils could adopt without renegotiating onerous information governance from scratch. Building confidence in this way allows staff to think about how the system as a whole can help people in new ways, rather than staying in-the-lane because it wouldn’t be possible.

At the centre, Digital Planning shows what happens when there is joined-up action. MHCLG combined data standards, legal levers and funding for alternative products, and moved a service area that had been stuck on proprietary systems and paper-era formats. It did not require councils to negotiate individually with incumbents, because the terms are set once, centrally. That is the method, and it is transferable to other core service areas.

These approaches work, and are transferable but not yet the norm. Test, Learn and Grow is where the lesson is being tested at national scale, locally. Its accelerators are putting central and local teams on the same problem in the same place, including Barnsley and Essex on AI and data support to the frontline, Wakefield on economic inactivity, and Northumberland and Manchester on integrated family support. Importantly, the programme’s second strand, ‘Grow’, names poor data infrastructure alongside fragmented funding as one of them.

The centre already has the tooling

GDS holds the Digital Commercial Centre of Excellence, the National Data Library and GDS Local. Beyond these, GDS has spent fifteen years building shareable service transformation method and capability, and has a bold new service transformation offer with CustomerFirst.

MHCLG holds Local Digital, Local AI and the Mayoral Data Council. In other local government-facing departments there are projects to reform care, licensing, infrastructure and transport systems.

The centre’s job here is to consolidate these coherently and build the delivery economy that local agency runs on: the standards and market oversight that only work if done once, and the data flows that let a place see what it is dealing with.

What to do

Five decisions for central government and Number 10 North:

Number 10 North should make departments more accountable for data sharing. A published obligation on named departments to release specific datasets to strategic authorities and councils, starting with Universal Credit, reported against. The point is not a new legal power, which largely exists, but a measure: opening data up should be something a Whitehall leader is judged on.

Extend the Digital Planning method to other markets in local government-facing departments. Children’s social care first, combining legal powers, standards and credible alternative products.

Act on commercial practice. Direct the Digital Commercial Centre of Excellence at market shaping - and specifically API and exit charging, so interoperability is default for everyone and reorganisation does not become a windfall for incumbents and the whole sector benefits.

Build more local government into the National Data Library. Fund more kickstarters co-created with councils through the Mayoral Data Council and regions, and produce common data-sharing agreements the sector can adopt rather than renegotiate.

Open more GDS capability outwards. Make components, design and assessment practice and capability programmes available locally, and measure uptake as a published commitment.

Most of this is prioritisation, standards, powers and commercial leverage and less expensive against what the current arrangements cost. Much of this was called for last year by The Future Governance Forum in their report Impactful Devolution 02: Local Government for the Digital Era.

Regions and strategic authorities are closer to the right scale to see across boundaries and work collectively, and large enough to back a smaller supplier no single council would choose alone. They need the arrangements - some of which already exist - that make routine sharing possible.

In the broadest sense, each capability should sit at the level of government with the authority, scale and proximity to make it work, rather than just at the level politically accountable for the problem. In old money, this was called subsidiarity.

Conclusion

None of this is newly discovered. Practitioners in local government and programme deliverers in Whitehall have described or encountered these constraints for the best part of a decade, and the evidence is settled. What’s more, much of the tech sector is open for change - but being settled markets, they lack the incentive to move together.

It is difficult for councils to build the connections between organisations on their own. Rules and standards, and the terms on which systems exchange data, should be set, for everyone, by the centre, working with the best in local government. Leaving them unset is over-deference to a legacy form of localism at best - and delivery absenteeism at worst.

So practical devolution needs a centre that is more, not less, active in a rewired state: changing market conditions, setting standards, opening its capability outwards and making data move. That is the new delivery economy the new settlement depends on. The alternative devolves the responsibility and leaves the constraint exactly where it was.