EU project applications get rejected for four repeating reasons, and only one of them concerns the idea: relevance failures (answering the call you wished for instead of the one that was published), design failures (an intervention logic that does not connect needs to activities to outcomes), credibility failures (budgets, timelines and partner roles that contradict the narrative), and formal or eligibility errors that mean the proposal is never scored at all. These families repeat across Erasmus+, Horizon Europe, CERV and Interreg because every evaluation grid, whatever its wording, tests the same questions: does this matter to the programme, will this design produce the change it promises, and can this consortium deliver it. Rejection is rarely about the idea; it is structural, checkable, and preventable before submission.
Before any evaluator reads a word of the narrative, programme staff run a formal check. Proposals that fail it are not scored low, they are never scored. The rejection arrives without comments, so there is nothing to learn from and a full funding cycle is lost.
The patterns we see repeatedly in drafts we review are mundane. A partner whose legal entity registration was never validated. A coordinator applying under a strand its organisation type cannot lead. Missing annexes, declarations or mandate letters. Page limits exceeded, so the excess pages go unread and the methodology ends mid-sentence. A requested grant above the ceiling. A last-minute submission on a portal under peak load.
The cause is organisational: the checklist gets delegated to the least senior person and checked in the final week, while the senior people polish prose. Reverse that order. Verify every partner against the call conditions at the start, on the official call page of the Funding and Tenders portal or in the Erasmus+ programme guide, and again before submission, because consortia change along the way.
The most common scored failure is an intervention logic with broken links. The needs analysis cites continent-level context but never evidences the problem for the specific target group and territory. Objectives turn out to be activities restated: "to deliver a training programme" describes work, not change. Indicators measure effort (events held, participants reached, toolkits produced) instead of the difference that effort is supposed to make.
The cause is usually simple: the logframe was assembled after the narrative, by a different person, to satisfy a template. The two diverge, and an evaluator tracing any objective to its indicators finds the chain broken.
Run the three-link test. Every stated need must be answered by an activity. Every activity must serve a stated need, otherwise it is cost. Every outcome must have activities that could plausibly produce it, otherwise it is a promise. We cover the full logframe coherence check in a separate article in this series.
Evaluators cross-read. A total in the narrative that differs from the budget table, staff effort concentrated in the coordinator while the narrative claims co-creation, a partner with a large described role and near-zero budget: each mismatch reads as carelessness at best, a decorative consortium at worst, and both readings cost points.
Timelines fail the same way. Pilots that start before the development work package ends. Evaluation compressed into the final phase, with no time to feed findings back into delivery. None of this is a writing problem; it is a planning problem that the writing exposed.
The repair is a cross-check pass, not more prose: every figure in the narrative reconciled against the budget tables, every described role against allocated effort, every timeline dependency walked forward. The budget cross-check has its own article in this series; the short version is that narrative and tables must come from one source of truth.
A recycled proposal is easy to recognise. It uses the vocabulary of a previous call, addresses a neighbouring priority, and connects itself to the published call text with pasted priority language. Verbatim pasting is not relevance; it is quotation. What the grid rewards is mechanism: which priority, addressed by which activity, producing which contribution the programme can count.
The pattern we see in strong applications is unglamorous: a working document that lists every priority, expected outcome and horizontal issue from the call text and maps each one to a section of the proposal. Where the map has a blank cell, either the proposal changes or this call is not for you this round.
Evaluators can tell when the final work package was written on the final weekend: a website, social media, a final conference, a newsletter. They discount it because they have read the same list in proposal after proposal, and it describes activity that stops the day the funding does. The website expires with the hosting contract; the outputs end up in a PDF nobody can find.
What earns belief is infrastructure with an operator and a reason to exist after the grant: a catalogue a community actually consults, a knowledge base a named organisation commits to keeping current. Plan the operating model inside the proposal, with names against who hosts, who edits and who pays. Building platforms that survive their funding period is a design decision made at proposal stage, not in the project's final months.
Run this sequence with a reviewer who did not write the proposal, armed with the call text and the published evaluation grid, and timeboxed the way a real evaluator is.
This is the same sequence we run when consortia ask us to check a proposal before submission; the order exists because it matches how proposals actually die.
Structurally, no. There is no blacklist, and evaluation panels assess the proposal in front of them, not your history. The real cost is the lost cycle. What does hurt is resubmitting the same text: the weaknesses named in the evaluation summary will be found again, sometimes by the same expert. Check the call conditions for any resubmission rules, revise against the comments, and treat the previous evaluation as the cheapest expert review you will ever receive.
Map every comment to the section that triggered it, then classify it. "Not sufficiently demonstrated" and "insufficient detail" are evidence gaps: the idea survived, the proof was missing, so add data, named partners and worked examples. "Not convincing" or "unclear added value" aim at the concept itself and need rethinking, not rewording. A criterion scored near or below threshold means that section gets rebuilt, not polished. Comments are terse; read them as a repair map, not a verdict.
StrandsUnited runs structural checks on real proposals: logframe coherence, budget reconciliation, work-package consistency. We also operate platforms like Impactful that carry project results past their funding period. Behind that work sits our consortium graph of 114,731 organisations and 83,490 projects, built from CORDIS and Erasmus+ open data covering 2014-2027, which is where our sense of how consortia form, win and fail comes from. The failure families in this article are the ones that keep resurfacing in that work.