Official evaluation mechanics
How Horizon Europe proposals are actually evaluated
Worth understanding before trusting any pre-submission score, including ours. The rules below are the ones that decide outcomes, and most of them are invisible in the call text itself.
The process
Three phases, and only the last one reaches you
Individual evaluation
Several independent experts read the proposal separately and each records scores and comments without seeing the others. Independence is the point: it is what makes the later disagreement informative rather than an artefact of who spoke first.
Consensus discussion
The experts convene, compare their individual assessments and negotiate a single agreed position per criterion. Scores commonly move here, and a comment that survives this stage is one at least two readers were willing to defend.
Panel review and the ESR
A panel checks consistency across the whole set of proposals, resolves ties and ranks. The applicant eventually receives the Evaluation Summary Report — the consensus comments and scores, and nothing about how they were reached.
Bidshark reproduces this shape deliberately — independent readings, then a reconciliation step where disagreement is tested, then one senior decision — because a single pass produces one opinion with nothing to check it against.
The scale
Three criteria, five points each, half marks throughout
Each criterion is scored 0 to 5 in steps of 0.5. For most RIA and IA calls a criterion must reach 3.0 and the total must reach 10.0 of 15 — but clearing the threshold and being funded are different questions.
Excellence
0–5.0Clarity and pertinence of the objectives, the extent to which the work is ambitious and beyond the state of the art, and the soundness of the methodology.
Impact
0–5.0The credibility of the pathway from results to the outcomes the call expects, and the quality of the measures to maximise it — dissemination, exploitation and communication.
Implementation
0–5.0The quality and effectiveness of the work plan, the appropriateness of effort and resources, and the capacity of the participants to deliver it.
Threshold is the floor, not the target
In oversubscribed calls, proposals well above threshold are routinely unfunded because the ranked list runs out of budget first. A total in the low tens clears the formal bar and still loses. This is why our report separates threshold status from competitive position rather than reporting a single verdict.
The rules that decide outcomes
What the call text does not tell you
Sub-criteria differ by type of action
RIA and IA, CSA, COFUND, MSCA and ERC each carry distinct sub-aspect lists. Applying one fixed Excellence/Impact/Implementation checklist to every call is structurally wrong for a meaningful share of proposals — and is the most common way a generic AI review goes astray.
Three severities, not a free-form guess
Evaluators classify what they find as a minor shortcoming, a shortcoming, or a significant weakness — the last meaning treatment so limited or ineffective that it pushes the criterion below threshold. The number follows from that classification, not the other way round.
Judged as submitted
The proposal is assessed as it stands. An evaluator may not credit it for a fix the applicant could obviously make, and the ESR carries no recommendations.
No double penalisation
One weakness may not be charged against two criteria. A thin work plan is an Implementation problem; it does not also reduce Excellence.
Hidden checks that lower Excellence on their own
Open science practices, the gender dimension in research content, SSH integration where the topic flags it, and the technical robustness of any AI involved. Each is easy to omit and each can cost points by itself.