The Partners You Keep: What a Graph of Past EU Projects Reveals About Repeat Consortia

4 August 2026

Repeat partnerships are visible everywhere in the record of funded EU projects, and they are not an accident. Map past Erasmus+ and Horizon Europe consortia as a network and funded projects stop looking like lists of partners: they look like structures, most often a stable core of organisations that have worked together before, surrounded by a periphery of partners who appear once and move on. Open data cannot prove that repeat ties cause success, because rejected consortia leave no public trace. What it does show, clearly and repeatedly, is that experienced consortia are built around repeat ties, and that coordinators who think in networks assemble their partnerships differently from coordinators who think in lists.

Why "find a partner" is the wrong question

Most partner searches run as a shopping exercise. The coordinator opens a directory, filters by country and organisation type, collects expressions of interest, and fills the empty rows in the application form. Each partner is evaluated on its own merits: track record, capacity, thematic fit.

That evaluation misses the thing that determines how the project will actually run. A consortium is a temporary organisation. Budgets, deliverables and blame flow through relationships between partners, not through their individual profiles. The questions that decide whether month nine goes smoothly are relational: who has worked with whom before, who trusts whose financial reporting, who will pick up the phone when a partner goes silent.

So the unit of analysis should be the tie, not the organisation. Two mid-sized partners who have delivered two projects together are, for coordination purposes, a different kind of asset than two impressive strangers. A partner list hides this. A graph shows it.

What a node, a tie, and a repeat tie mean in the data

We maintain a consortium graph built from CORDIS and Erasmus+ open data, and it is worth being precise about what it encodes, because network language turns vague quickly.

  • A node is an organisation that has participated in at least one funded project. Even identifying "the same" organisation across years is real work: names change, departments apply separately, identifiers get recorded inconsistently. Any honest graph carries some of this noise.
  • A tie between two organisations means only that they appeared in the same funded consortium. They were listed on the same grant agreement, and that is all the data records. It does not mean they collaborated closely or would choose each other again.
  • A repeat tie means the same pair appears together in more than one funded project. This is the interesting signal. One shared project can happen by accident of a call. Choosing each other again, in a different call and sometimes a different programme, is a deliberate and costly act. It usually means the first collaboration was at least tolerable, and often that it was good.

Everything the graph reveals about consortium formation rests on that distinction between a tie and a repeat tie.

The shapes that recur

Look at enough consortia through this lens and a few shapes come back again and again.

The most common is the stable core with a rotating periphery. A small set of organisations forms the core: they appear together across projects, calls and years, and they typically hold coordination, quality assurance and financial management. Around them, each new project adds peripheral partners chosen for what the call requires: a country to complete the geographic spread, a target group, a pilot site, a sector voice. Peripheral partners often appear once and are not seen with that core again. That is not a flaw in the model; it is the model. The core carries the machinery, the periphery carries the specificity.

The second shape is the broker: an organisation sitting between clusters that otherwise never touch. A network of universities and a network of municipalities may share no direct ties at all, while one intermediary, often an NGO or an association, holds repeat ties into both. Brokers are how newcomers actually enter EU projects, far more often than through cold outreach: a broker vouches for you on one side because it is trusted on the other.

The third shape is the closed clique: a group that partners almost exclusively with itself. It coordinates cheaply and delivers predictably, and over time it can start to read as a circle re-funding itself, with little new blood and shrinking reach.

What this implies for a coordinator

The graph does not tell you whom to invite. It tells you where to place people.

  • Put repeat ties where failure would stall the whole project. Coordination, reporting, budget management and quality assurance should sit on pairs with shared history. This is where repeat ties genuinely reduce risk: the expensive negotiations already happened in a previous project.
  • Put newcomers where their contribution is content, not coordination. A newcomer with a bounded, well-specified role (one pilot, one country, one deliverable) adds reach without adding much risk. A newcomer holding a cross-cutting work package multiplies it.
  • Pair every newcomer with an experienced neighbour. If a first-time partner's closest collaborator in the consortium is an organisation your core has worked with, onboarding has an owner. If the newcomer connects to nobody, onboarding lands silently on the coordinator.
  • Watch both extremes. A consortium with no repeat ties at all will spend its first year negotiating basics. A consortium that is one closed clique should ask what a fresh partner would add, and the answer is usually exactly the reach and credibility the proposal needs.

Reading your own draft partner list as a network

Before submission, spend an hour turning your list into a picture. It requires no tooling beyond patience.

Step 1. Draw your partners as points and connect every pair that has shared a funded project before. CORDIS and the Erasmus+ project results platform both let you check an organisation's project history.

Step 2. Find your core. If the connected pairs do not include the partners holding coordination-critical work packages, you are running critical machinery on untested relationships, and your risk section should say how you will manage that.

Step 3. Find your isolates: partners connected to nothing. Each one is a real coordination cost. Decide, by name, who onboards each isolate.

Step 4. Check for a broker. If your consortium spans two communities, someone must be credible in both. If nobody is, the seam between the two halves is where the project will strain.

Step 5. Reread the work plan against the picture: strong ties carrying critical work, newcomers in bounded roles, no partner floating free. That shape is what experienced consortia converge on.

What the graph cannot see, and how to fill the gap

Open data records participation, not performance. The graph cannot see delivery quality: an organisation can appear in many projects and have been carried through every one of them. It cannot see internal conflict, near-misses, or the partner quietly dropped from the next proposal. It cannot see staff turnover, and ties really live between people: when the project manager who built a relationship leaves, the tie on paper outlives the tie in reality. And it only ever shows funded consortia, so every pattern in it is a pattern among survivors.

This is why the network lens complements, and never replaces, partner due diligence: reference conversations with previous coordinators, checking that claimed deliverables are actually public, asking which named people will staff the work. The graph tells you where to look and which questions to ask; the answers still come from humans. Structural partner analysis of this kind is part of the grant intelligence work we do at StrandsUnited for coordinators preparing applications.

Where this comes from

StrandsUnited maintains a consortium graph covering 114,731 organisations and 83,490 projects, built from CORDIS and Erasmus+ open data for the 2014-2027 programming periods. The shapes described here are the ones we see recurring in that graph, tempered by what we learn operating community platforms such as Impactful, where the coordination problems behind these network patterns play out in daily practice.