The one thing about FDP that no break clause can fix

The Canonical Data Model

The Canonical Data Model (CDM) is the most valuable asset in the NHS Federated Data Platform. It is more important than the supplier, more important than the platform technology, and more important than any individual product. If FDP succeeds in transforming NHS data infrastructure, it will be because the CDM gave 220 Trusts a shared language for the first time. If it falls short, it will be because the CDM was not given the governance, resources, and collaborative development it needs.

But this is not just about FDP. The CDM underpins the entire vision for NHS data modernisation, including the Transforming and Connecting Care programme, which will drive the move towards a Single Patient Record and event-based data flows. These ambitions sit within the £10 billion commitment to NHS technology announced in the spending review. If the CDM is not right, these programmes cannot deliver on their promise.

I am writing this because I want FDP to succeed. Critics of the programme should not read this as ammunition. It is an argument for doubling down on the investment and getting the foundations right, not walking away from them.

What the CDM is and why it matters

The CDM defines how NHS data is structured, labelled, and connected across every Trust instance. It is the main reason a product built in Dorset can be adopted in Newcastle without rewriting code. It is the reason AI can traverse the data and understand what a referral means, what a waiting list is, and how they relate to each other. It is the reason staff can move between Trusts and be productive on day one. Without the CDM, FDP is just another data warehouse. With it, FDP is a national operating standard for NHS data.

The CDM also extends far beyond FDP. It is the foundation for the Single Patient Record. EPRs should align to it. All applications live in the NHS should map to it. Without that alignment, data cannot be easily shared between systems regardless of whether those systems sit inside FDP or outside it.

Tito Castillo, an Enterprise Architect and Data Management Consultant, has written compellingly about sovereignty as a property of the data definition layer, not the infrastructure layer. His argument is that an organisation which does not own the definitions its data is built on does not control its own systems, regardless of who supplies them. The CDM is that definition layer for the NHS. Its long-term health determines whether the NHS truly controls its own data estate.

A governance success that needs to scale

Getting data governance right costs money. It requires dedicated resources, formal skills, and sustained investment. It is worth recognising that the data governance function at NHS England was a rare success. The Tech and Data Integration function within the Data and Analytics directorate, with Tomas Sanchez Lopez leading the architecture, represented one of the first times the NHS created a governance function based on proper data management principles, using DAMA concepts, that actually got funded and staffed.

The challenge is that governance has not yet scaled to match the pace of delivery. That is a success problem, not a failure. The programme was onboarding Trusts faster than expected. The governance needs to catch up, and that requires investment, not criticism.

Where the gap opened

The original CDM work was slowed by reorganisation, recruitment freezes, and headcount reduction following the NHS Digital and NHS England merger. When the government announced a further reorganisation in March 2025, the recruitment freeze was not lifted. None of this was the team’s fault. They did an excellent job in challenging circumstances.

At the same time, the FDP contract was signed and the delivery team was under pressure to get Trusts live. The implementation moved faster than the governance, and the delivery team used external resource via the Palantir contract to build the operational CDM within deployed Trust instances. This was pragmatic. You cannot wait for perfect governance when you have 220 organisations to onboard.

But it means that two parallel workstreams developed: the original CDM governance function building a carefully designed data standard, and the delivery team building the operational ontology that actually powers the live platform. These two workstreams now need to be brought together.

Both versions of the CDM are publicly available. The Tech and Data Integration team published their CDM (currently at v0.14 DRAFT, dated September 2025) through a public NHS Futures page, while the FDP delivery team’s operational data model is published in an OGL GitHub repository. The existence of two public versions of the canonical data model, developed by different teams within the same organisation, illustrates the coordination challenge more clearly than any commentary could. Open scrutiny of both versions, and a transparent process for converging them, would be a significant step forward.

What the community is finding

Tom Smith, Head of Activity Analysis and Forecasting at Nottingham University Hospitals NHS Trust, recently built a CDM explorer app from the publicly available data model on GitHub. Within hours he had surfaced real questions: an admissions entity with no discharge entity, a birth entity with no delivery entity, no link from the labour entity to the theatre session entity despite around 40% of labours involving a visit to theatre for a caesarean section.

Smith’s work illustrates two things. First, the CDM is published openly, which is a strength of the programme that deserves recognition. Second, there are completeness gaps that would benefit from wider community engagement. His findings are different from the governance question I am raising here, but they are connected. Completeness gaps arise when the development process does not draw on enough expertise from enough sources. A more collaborative, open approach to CDM development would surface these issues earlier and fix them faster.

What Trusts are experiencing

I have spoken to Trusts that are pioneering the integration between the CDM and their local data models. They are pressing ahead, surfacing the areas where the CDM needs to grow. The centre needs to match their ambition with proper governance, published standards, and responsive support. When a Trust wants to bring a new data source into FDP, it needs to map that source to the CDM. If the CDM does not cover the data elements the Trust needs, there should be a clear, fast route to request an extension.

There is also a hidden cost. Many Trusts have relied entirely on EPR suppliers to define their data models. As Trusts develop their own requirements into a standardised data model, more resource and expertise are required, and these are not currently in place.

As FDP moves towards replacing local Trust data warehouses, the CDM becomes the glue that connects the entire platform. If every Trust can build on a common, well-governed data model, the NHS achieves something it has never had: a truly interoperable data estate. If the CDM fragments because governance did not keep pace, Trusts will build their own local models and the standardisation that makes FDP valuable will be lost.

The case for a collaborative approach

The CDM should not be developed in isolation by a central team, however talented. It should draw on every available source of expertise, including wholesale adoption of data models that are already proven to work.

There are big gaps in areas like workforce data, where the CDM is being developed slowly from scratch while mature staff rostering tools already deployed across the NHS contain sophisticated, proven data models describing shifts, rotas, staff availability, skills, and capacity. These models exist. They work. They should be adopted into the CDM rather than reinvented. Where a proven, operational data model exists in a deployed NHS system, the default approach should be to adopt it, not build a parallel version from scratch.

This collaborative approach must extend to Trusts themselves. Data meaning must be owned locally. There needs to be a mechanism for all 220 Trusts to have their say in how the CDM develops. Trusts must also be able to extend the CDM quickly on their own terms and then standardise. It cannot take 18 months to change the standard. An agile governance process is essential.

What needs to happen

This is a governance gap that opened because delivery outpaced governance. It can be closed. But it requires structural investment, not just goodwill. 

Castillo has argued persuasively for a separation of functions between delivery, governance, and assurance. Governance dissolves under delivery pressure when it sits within or alongside the delivery function. The delivery team has the budget, the resources, and the mandate. The governance function needs equivalent institutional weight if it is to create the healthy tension that gets the data model right, not just fast.

Invest in separated governance. CDM governance needs its own mandate, resources, and authority. The FDP delivery team should be a consumer of the CDM standard, not its owner.

Protect governance from restructuring. Given the abolition of NHS England and the track record of governance functions being cut in reorganisations, CDM governance needs institutional protection. It serves a broader constituency than the NHS alone, including the research community through the Health Data Research Service. An independent or arms-length governance function would be more sustainable.

Open the CDM to community contribution. Tom Smith’s explorer app shows what happens when the data model is publicly accessible: the community finds gaps, asks questions, and contributes insight. The programme should actively encourage this, with published documentation, clear contribution pathways, and responsive engagement.

Make governance agile and enforceable. There should be enforceable timelines and named accountability. This must not drift into the slow cadence that characterises much of the current information governance landscape.

Ring-fence resources. Dedicated resources for CDM governance must be protected from headcount reduction. The pattern of governance functions being absorbed, merged, and hollowed out through successive NHS restructures is the single biggest threat to the CDM’s long-term viability. Trusts need protected capacity too.

The CDM is worth fighting for

Pritesh Mistry at the King’s Fund recently asked publicly whether the CDM had been scrapped or failed. It has not. It is alive, it is in use across all Trust FDP instances, and it is backing every national FDP product. It needs more investment, more collaboration, and proper governance to fulfil its potential.

The political debate about whether Palantir should be replaced is a distraction from this. Even if the supplier changed tomorrow, the CDM would need to be right. It is the one thing that no break clause can fix and no new supplier can replace. As Castillo argues, sovereignty in digital systems is a property of the data definition layer, not the infrastructure layer. The CDM is that layer. We must not allow multiple American companies to define, independently of each other, what our NHS data means.

This issue must be addressed visibly, thoroughly, and with continued scrutiny. Not just because FDP depends on it, but because the £10bn spending review allocation depends on it, Transforming and Connecting Care depends on it, the Single Patient Record depends on it, and the entire vision for NHS data modernisation depends on it. The CDM must be governed independently of any single delivery programme or supplier relationship, but its development must be informed by and accountable to the NHS organisations, technology companies, and research communities that depend on it.

The foundations are sound. The vision is right. The CDM just needs the governance, the leadership, and the investment to match the ambition. That is an argument for more support, not less.

Tom Bartlett

Founder, Bartlett Data Ltd

Former Deputy Director of Data Engineering, NHS England

bartlettdata.co.uk