How to write a winning Horizon Europe proposal

Winning a Horizon Europe proposal is not about writing well. It is about writing to be scored, and building the substance before you write a word.

That distinction sounds small. It is the whole game. Most proposals are rejected, and the ones that fail are rarely badly written: they are well-written documents built on a weak backbone. Since the shift from Horizon 2020, a collaborative project must show it is not research and innovation without impact. Strong science with a weak or generic pathway to impact is precisely the profile that fails today.

This guide covers collaborative Research and Innovation Actions (RIA) and Innovation Actions (IA) under Pillar 2 of Horizon Europe. It walks the whole journey at orientation level: what evaluators actually do, how an application is put together, the four-phase method that puts substance before prose, and the three criteria your proposal will be scored against. Each section links to a deeper guide when you are ready to go further.

Writing well is not the goal, scoring well is

Before you type anything, understand who you are writing for. Horizon Europe evaluators do not read your proposal the way a colleague would. They read it with the evaluation form open beside them, hunting for the elements the form asks them to score.

The European Commission's own briefing for evaluators is explicit. Evaluators must assess a proposal "as submitted and not on its potential if certain changes were to be made". And when they find problems, they "explain shortcomings, but do not make recommendations".

Read those two sentences again, because together they define your job. There is no benefit of the doubt, no reading between the lines, no credit for what you meant to say. If it is not on the page, it does not score. A brilliant methodology that lives in your head, an impact your consortium takes for granted, a risk you managed but never wrote down: to the evaluator, none of it exists.

This is why the strongest teams write against the scoring grid, not against their own enthusiasm. To see exactly what that grid looks like, how scores and thresholds work, and where fractions of a point are won and lost, read how Horizon Europe proposals are scored.

How a Horizon Europe application is put together

A Horizon Europe application has three parts, and only one of them wins or loses the grant.

Part A is the administrative form, completed directly in the Funding & Tenders Portal: participants, budget tables, declarations. It has to be right, but it is not evaluated for quality.

Part B is the technical description. This is the document evaluators read and score: a single PDF on a fixed template, with a page limit of 40 pages for RIA and IA (45 pages for lump-sum funding) in the current 2026 application form. Every argument you want scored must live inside those pages.

Part C and the annexes carry the supporting material: the ethics self-assessment, letters of support and other required documents.

The Part B template is not a suggestion. Its structure mirrors the evaluation criteria, and every section exists because something in the scoring grid depends on it. For a section-by-section walk through what goes where, and what evaluators look for in each, see the Horizon Europe Part B template, section by section. And to see how those sections read once they are filled in, with an evaluator's notes in the margin, walk through an annotated proposal example.

RIA, IA or CSA: choose the right vehicle first

One decision comes before any writing: the action type. A RIA, an IA and a CSA are evaluated with different expectations, funded at different rates, and demand different consortia. Pitching an innovation-heavy project as a RIA, or research-heavy work as an IA, sets you up to disappoint the evaluators whatever you write.

One consequence worth knowing now: for an Innovation Action, the Impact criterion carries a ranking weight of 1.5 at the panel stage. On close calls, and most funded calls are close, Impact decides which IA gets the money. Choose your vehicle with your eyes open. The full comparison, including how the evaluation emphasis shifts between action types, is in RIA vs IA vs CSA, which action fits you.

The method: build the backbone before you write

Here is the mistake almost every first-time consortium makes: they open the Part B template and start writing. Weeks later they have forty pages of prose and no way to check whether the underlying project holds together.

A winning proposal is built in four phases, and the writing is the third. The first two build the backbone: the structured substance of the project. The writing then fills the template from that backbone, and the final phase evaluates the result. Teams that start with prose end up with a beautiful document and a broken backbone. Teams that build the backbone first, collaboratively with their partners, write faster and score higher.

Phase 1: strategy, "what am I proposing?"

Strategy answers one question: what exactly are you proposing, to solve what, for whom?

Start by validating and quantifying the problems. Evaluators want figures, not generalities: a problem you cannot quantify is a problem you have not yet understood. Then define specific target groups. Not "stakeholders", but, for example, "regional transport authorities in eight EU countries": who is affected, and who will adopt the results.

From there, build objectives that are specific and measurable, each covering at least one of the call's expected outcomes. This mapping is not optional decoration; the call topic defines the outcomes the Commission is paying for, and an objective that covers none of them is dead weight. Identify the key results, meaning what the project will actually produce: a prototype, a dataset, a method, a policy recommendation. Finally, assemble a consortium that covers the call's requirements, not one assembled from whoever happened to be available. Pulling that consortium's contributions into one coherent proposal is a job in itself, covered in coordinating a Horizon Europe consortium.

Phase 2: design, "how will I do it, and why will it matter?"

Design turns the strategy into a project that evaluators can believe.

First, set the scientific backbone: the innovation claim and the methodology overview. What is genuinely new, and how will you do it?

Second, build the impact pathway, and make it traceable end to end: key results, then dissemination, exploitation and communication, then outcomes, then impacts. Write the outcomes as actor-centric statements: a named target group adopting a specific result, quantified, with a timeframe. This is where most proposals lose points. "Improved awareness of the problem" is an aspiration; a specific group of authorities using your method within a defined period is an outcome. The pathway logic, and how to make it credible rather than generic, is unpacked in the Impact section, demystified.

Third, plan the implementation: work packages, tasks, deliverables, milestones, risks, the Gantt. And list the barriers to impact, which are external obstacles to adoption (regulation, market structure, user behaviour), not internal project risks. Confusing the two is a classic tell of a rushed proposal.

The measures that carry your results to their adopters deserve their own design pass; a generic bullet list of "conferences and a website" convinces nobody. How to build measures evaluators actually credit is covered in the dissemination, exploitation and communication plan.

Phase 3: write, "I draft the Part B"

Now, with the backbone in place, the writing becomes assembly rather than invention. The three sections of Part B fill from what you have already built: Excellence draws on your strategy and scientific backbone, Impact draws on your design work and impact pathway, Implementation draws on the work plan. You refine the text, not the substance.

This is the phase where good habits pay off page by page. Every claim gets its evidence. Every objective keeps its indicator. Every outcome stays actor-centric and quantified, because prose has a way of softening precise statements back into aspirations if nobody is watching.

This is also the phase where teams reach for a general chatbot, so it is worth knowing early what AI can and cannot do here, and what you must never paste into it: see whether AI can write your Horizon Europe proposal.

Phase 4: submit, "is it ready?"

Only now do you switch from building to evaluating. Score the proposal the way an evaluator would, criterion by criterion, with the briefing rules in mind: as submitted, no benefit of intent. Fix the weak sections. Then check every formal requirement, because an admissibility slip can waste a year of work, and export.

The final verification pass has its own discipline: page limits, mandatory sections, annexes, portal formalities, internal consistency between Part A and Part B. Work through the pre-submission checklist before you press submit, not after.

The three criteria, and where proposals fail

Every RIA and IA is scored on the same three criteria. Here is what each one really tests.

Excellence tests whether the project is worth doing and whether you can do it. That means clear, measurable objectives; ambition that is genuinely beyond the state of the art, demonstrated rather than asserted; and a methodology that is operational, not just descriptive. It also covers the cross-cutting content that teams routinely underweight: the gender dimension in the research, open science practices, data management, and the integration of social sciences and humanities where the call expects it.

Impact tests whether anything will change because your project existed. Evaluators want a credible pathway from your results to the call's expected outcomes and impacts, plus dissemination, exploitation and communication measures strong enough to carry results to the people who will use them. Since Horizon 2020, this criterion is where the profile of the failing proposal has shifted: excellent science, generic pathway, no funding.

Implementation tests whether this consortium can deliver this plan. A coherent work plan, honest risks with real mitigations, and an effort table that matches the ambition: a work package promising the moon on a sliver of person-months undermines everything written before it.

How the scores on these criteria translate into thresholds, rankings and funding decisions, including where the IA Impact weighting bites, is detailed in how Horizon Europe proposals are scored. And because the fastest way to improve is to study how others fail, the recurring failure patterns, criterion by criterion, are collected in typical mistakes in Horizon Europe proposals.

The margin is everything

Most Horizon Europe proposals are rejected. That is not a reason for pessimism; it is the reason method matters. The proposals that win are not the ones with the best ideas. They are the ones where a good idea was built into a traceable backbone, written to be scored, and checked against the evaluator's rules before submission.

The difference between funded and rejected is usually not a threshold-breaking weakness. It is an accumulation of fraction-of-a-point shortcomings: an outcome written as an aspiration here, an unquantified problem there, a methodology that describes instead of operating. Each one is small. Together they are the margin.

You now have the map. Build the backbone before you write. Write for the scoring grid, not the reader you wish you had. Evaluate before you submit, with no benefit of the doubt.

→ Before you submit, let GrantForge's pre-submission evaluation score your proposal the way real evaluators will, and separate the fraction-of-a-point shortcomings from the threshold-breaking weaknesses, while you can still fix both.