A policy speech tells you what an institution wants to be seen wanting. A procurement document tells you what it has agreed to pay for. The two are not the same, and the gap between them is usually the more interesting reading. A press release costs nothing to write and binds no one. A call for tender names a budget, a deadline, a set of deliverables and a contract that someone will sign and then have to deliver against. Intentions are cheap. Commitments leave a paper trail, because somebody has to do the work and somebody has to account for the money.

That is the whole premise. If you want to know where a system is actually heading, read the documents that commit resources, not the documents that announce ambitions. The European Union publishes both, in the open, and the procurement layer is more honest than the rhetoric layer for the simple reason that it is operational. Below is a way to read it, with European health data as the worked example, because that is where the method pays off most clearly right now.

How to read between the layers

The thing to understand first is that EU funding is not one document. It is a stack, and each layer is more concrete and less deniable than the one above it. Reading well means reading the transitions, not the individual documents.

At the top sits the work programme. For health this is an annual Commission Implementing Decision, a financing decision that sets the themes and allocates the budget for the year. The 2026 EU4Health call runs under the 2025 Work Programme, adopted by Commission decision in July 2025. This is the layer closest to rhetoric: it speaks in objectives and added value. But it already commits money to named themes, so it is the first place the intentions acquire a price.

Below that sit the calls for proposals, the grant invitations. Here the direction starts to show, because the Commission is offering to co-fund projects in specific directions and the directions are listed, costed, and given deadlines. This is where you read what the institution wants others to build.

And below that, the calls for tender: the actual procurement. Not “we will co-fund your project in this direction” but “we are buying this service, here are the specifications, here is the ceiling”. A tender is the least deniable layer. Someone wrote a statement of work, someone will bid, someone will sign, and at the end there will be a contract notice naming the contractor and the value. By the time you reach a signed contract, intention has become obligation.

The reading happens between the layers. A theme that appears in the work programme but never descends into a tender is talk. A tender that appears without a visible policy parent is worth a second look, because the money is moving ahead of the stated reason. And a theme that runs cleanly down all three layers, work programme to call to signed contract, is a direction the institution is actually committed to, whatever the surrounding language says.

What the documents say, before any interpretation

The codes themselves are readable once you know the grammar. Take EU4H-2026-SANTE-PJ-04. It parses as programme (EU4Health), year, directorate (SANTE, health), instrument (PJ, a project grant), and a sequence number. Behind it sits a second internal code, CR/CV&NCD-g-25-16, which encodes the policy cluster: cardiovascular and non-communicable disease. The taxonomy is not hidden. It is sitting in the reference number, and it tells you which bureaucratic drawer the money came out of.

The call fiche for the 2026 health grants lists nine topics. Three are cancer screening pilots, for gastric, lung and prostate cancer. Several are clinical: lifelong cardiovascular prevention, orphan medical devices for children, a medicine pricing tracker. But two of the nine are not about a disease at all. They are about data. Topic 04 is a flagship initiative to leverage AI and health data for cardiovascular care, framed as laying the foundations of a trusted European data ecosystem for health innovation. Topic 08, budget €14.4 million, is health data for biotech innovation leveraging the European Health Data Space, including the deployment of what the call calls AI factories.

The fiche also rewards close reading in a way summaries do not. Its history of changes records that for topic 04, the expected number of grant agreements was revised from one to “one or more”. A single line, but it tells you the flagship was reconsidered: from a single large consortium to possibly several. The revisions are where the uncertainty shows.

Then drop a layer, to a tender. HADEA/2024/OP/0034, “support to stakeholders on secondary use of data within the European Health Data Space”, a ceiling of €1.5 million. This is not a grant for someone’s project. It is the Commission buying a service. Read the specification and the actual purchase is narrower and more telling than the title suggests: not data infrastructure, but coordination. The contractor is to build and nurture a community of practice for the secondary use of health data, support the governance procedures around it, and run the whole thing under a certified project management method. It is, in plain terms, a secretariat-and-facilitation contract for the human layer around HealthData@EU, the cross-border infrastructure for reusing health data.

The specification also rewards the method directly, because it names its own neighbours. It lists the adjacent initiatives the contractor must work alongside: a 2023 capacity-building call, a joint action in which member states prepare for the reuse of data, the self-established community of practice of Health Data Access Bodies, the QUANTUM project labelling dataset quality, the HealthData@EU central services. One tender, and the whole surrounding constellation is legible from its reference list. You do not need to find the other documents; the document tells you they exist and how it sits among them. And the keystone: the contract was signed in November 2025. Intention became obligation. Someone is now doing this work.

So the chain is complete and visible, no rhetoric required: a work programme that prioritises a European Health Data Space, calls for proposals that fund AI on health data and biotech use of it, and a signed contract for the human scaffolding of secondary use. Three layers, one direction, money moving through all of them.

I am the one interpreting

Everything above is on the record. What follows is not. The moment you move from “the documents say this” to “this means that”, you have crossed from reading into interpretation, and interpretation has an author with a position. Significance is not a property of the documents. It is something a reader assigns, and the reader is always somewhere, looking from a particular place with particular concerns. I am not exempt from that and neither is anyone who tells you what a tender “really means”. So take the next part as mine, not as the data’s.

What the procurement layer reads like to me is sequencing. The European Health Data Space runs as a continuous thread through every layer, and the operational support was bought ahead of the law that governs it. When this tender was let, in late 2024, the EHDS Regulation was still a provisional agreement, and the specification says so in a footnote: pending finalisation and publication in the Official Journal. The regulation has since been adopted, in February 2025, and entered into force that March. So on its own the timing is unremarkable: a regulation in provisional agreement is close to final, and procuring against it is a reasonable bet.

The sharper point is downstream of that. The regulation is in force, but the part this tender serves, the secondary use of health data, does not become applicable until March 2029. The community of practice, the governance procedures, the stakeholder coordination, all of it is being built now for a legal regime that does not bite for years. That is not a bet on a draft law. It is scaffolding erected well ahead of the thing it scaffolds, and the gap is measured in years, not months. The direction is being committed to, and staffed, long before the obligations it prepares for come into effect.

That is not a scandal and I do not want to inflate it into one. Building the human infrastructure ahead of the legal deadline is defensible, even sensible: a community of practice takes years to grow, and standing it up the week the obligations bite would be too late. But the order is legible, and the order is a choice. The money and the coordination are moving on a timeline of their own, set years ahead of the law’s, and that tells you the direction is settled well before the public deadline that is supposed to mark its arrival. The documents do not tell you the order was wrong. They tell you what the order was, and that is already more than the press releases do.

There is a second thing the layering shows, quieter than the first. The disease topics in the call, the cancer screening, the cardiovascular prevention, are the visible health policy, the part that reads as care. The data topics are framed in the same language of health and innovation, but what is being built underneath is infrastructure: a durable, cross-border, reusable substrate for health data. Infrastructure outlasts the policy that justified it. Whatever the screening pilots achieve or fail to achieve, HealthData@EU will still be there, and the question of who governs it and on what terms is settled now, in tenders, while attention is on the diseases.

What the method cannot tell you

The honest part. A procurement document is a strong source precisely because it is narrow, and the narrowness cuts both ways. It tells you what was bought. It does not tell you whether the thing will work, whether it will be delivered on time or at all, or whether the contractor will quietly underperform against a specification nobody later checks. A signed contract is a commitment, not an outcome.

It also does not tell you who wins. A call names a ceiling and a scope; the award notice names a contractor, but the deeper question of which firms have built the capacity to bid on European health data infrastructure, and how few of them there are, is not answerable from any single document. You can read the direction of the money. You cannot read, from the procurement alone, the concentration of the market that receives it. That requires reading across many awards over time, and even then you are inferring.

And the method is silent on the thing that matters most and is the least visible: whether any of this improves anyone’s health. Procurement measures expenditure and deliverables, because those are what an institution can hold a contractor to. Health outcomes sit too far downstream, mediated by too many other systems, to be a line in a contract. So the documents that are most honest about what is being built are also structurally incapable of telling you whether it was worth building. That is not a flaw in the reading. It is the boundary of it, and pretending otherwise would be the same move the press releases make, dressed in evidence instead of ambition.

The method gives you direction, commitment, and sequence, read from documents that had to be honest because someone had to act on them. It gives you those things and stops there. The step from “this is what they bought” to “this is what it means” is one you take yourself, from wherever you happen to be standing, and the most useful thing the method does is make that step visible as a step, rather than letting it pass as a reading of the facts.