Every rejected Horizon Europe proposal we review has, somewhere in it, genuinely good science. That is rarely the reason it fails. After enough proposals, a pattern becomes obvious: the projects that miss threshold usually lose points on structure, not substance — on how the case is built and framed, not on whether the underlying research is sound.

This matters because it changes what "fixing" a proposal actually means. Applicants who get a low score tend to respond by strengthening the science further — adding more preliminary data, tightening the methodology, expanding the literature review. That effort is rarely wasted, but it is rarely what moves the needle either. Evaluators are reading against a rubric, under time pressure, across a stack of proposals in a single sitting. What pushes a proposal below threshold is almost always one of a small number of structural patterns. Here are the five we see most often, and what to do about each.

1. Topic-scope mismatch

The proposal answers a real and interesting research question — just not quite the one the call text asked. This happens most often when a research group has an existing line of work and reshapes it to fit a call, rather than starting from the call's actual scope and building toward it. The gap can be subtle: a call asking for a validated solution at TRL 5-6 receiving a proposal that reads as fundamental research, or a call scoped around a specific application domain receiving a proposal whose demonstrator sits one step removed from it.

Reviewers are instructed to check alignment with the topic description line by line, and a proposal that requires the reader to infer the connection has already lost ground before Excellence is scored on its own merits.

Fix: Before writing a word of the technical case, map every sentence of the call's scope paragraph to a specific section of your proposal. Anything in the call text you cannot point to directly in your draft is a gap an evaluator will find first.

2. The impact pathway doesn't connect to anything concrete

Excellence sections are usually the strongest part of a proposal — they are what the applicant's expertise is built on. Impact sections are frequently the weakest, because they ask for a different skill: translating technical outcomes into pathways, uptake, and measurable change for named stakeholders. A common failure mode is an Impact section that restates the project's ambition in more enthusiastic language without ever naming who adopts the results, on what timeline, or what specifically changes for them.

Evaluators are trained to look for a credible, specific pathway from result to outcome to impact — not a description of how important the field is in general.

Fix: For every claimed impact, name the actor who realizes it, the mechanism by which they do, and a KPI or milestone that would let a reviewer — or you, at mid-term review — check whether it happened.

3. Consortium complementarity is asserted, not demonstrated

Partner sections often list qualifications and past projects without making the underlying logic explicit: why this partner, for this specific work package, and what happens to the project if they are removed. Two failure patterns recur — visible overlap between partners with no explanation of the division of labor, and a critical capability (regulatory affairs, a specific fabrication process, a clinical site) that the consortium needs but does not visibly have, with no subcontracting or justification offered.

Fix: Draw the dependency, not just the role. For each partner, state the one thing the project cannot do without them, and confirm every capability the work plan requires has a named owner.

4. Risk and mitigation reads as a template, not a plan

Risk tables are frequently the most generic section of a proposal — "risk: partner delay; mitigation: close communication" — copied across work packages with minor edits. Evaluators read this as a signal that the applicant has not seriously stress-tested their own plan, which undermines confidence in Quality of Implementation even where the plan itself is reasonable.

Fix: Write the risk section last, after the work plan is finished, and make every mitigation an action with an owner and a trigger condition — not a general intention.

5. Compliance sections are treated as a formality

Gender dimension, open science, and ethics self-assessment are frequently the last sections written and the first ones evaluators recognize as boilerplate. A gender dimension paragraph that is not integrated into the research design itself, or an open science plan that restates the mandatory minimum without specifics on data type, format and repository, both read as compliance theater rather than genuine planning — and reviewers increasingly weigh this as a signal about the applicant's overall rigor, not just a checkbox.

Fix: Integrate these sections into the actual research design rather than appending them. A gender dimension that references your specific study population, or an open science plan that names your actual datasets, reads as substantive because it is.

Why these patterns are hard to catch on your own

None of this is a knowledge problem — most experienced applicants know all five patterns exist. The difficulty is structural: you wrote the proposal, so you already know what you meant. A reviewer does not have that context, and will not go looking for it. Catching your own topic-scope drift, or noticing that your Impact section restates ambition instead of naming a pathway, requires reading your own draft the way someone with no prior context and thirty minutes would — which is close to impossible to do for text you wrote yourself.

That gap between what you know you meant and what a first-time, time-pressured reader actually takes from the page is where most rejections live.