How EIC Accelerator evaluators score your proposal, explained by one of them

I am a certified EU Expert Evaluator, and I have sat on the reviewing side of the EIC Accelerator, Horizon Europe and EIT calls. Applicants almost never see that side of the process. You submit, you wait, and weeks later a letter arrives with a verdict and, if you are rejected, a short report that reads like it was written by someone who has already moved on to the next file. In a sense, it was.

What follows is not advice on how to write a better pitch deck. It is a description of the machinery: how many people actually score your proposal at each step, what scoring system they use, what has to happen for a proposal to get a GO, and what is and is not public about the people making that call. All of it is drawn from the European Commission's own Guide for Applicants (Version 6.0, current for the 2026 Work Programme) and the official EIC FAQ, not from second-hand summaries. Where the process is genuinely ambiguous or undisclosed, I say so rather than filling the gap with a guess.

Three steps, three different scoring systems

The single biggest misunderstanding I see is applicants assuming the same scoring logic applies at every stage. It does not. Each step uses a different panel of people and a different scale, and conflating them is how founders misread their own results.

Notice what changes between rows. Step 1 is a binary judgement call made independently by four people who never discuss it with each other. Step 2 is a numeric score built by a much smaller group working from a detailed rubric. Step 3 drops the numbers entirely and returns to a yes/no, but this time as a group decision reached in the room. A founder who treats all three as "the evaluators" making "a decision" is missing that the decision-making mechanism itself is different every time.

Step 1: the short proposal is a fast, independent yes/no

Your short proposal is read remotely by four independent experts, each working alone, each producing a single GO or NO-GO for the whole document, not per section. There is no numeric score at this stage. The rule for passing is specific: if at least three of the four give GO, you advance to the full proposal. If two or more give NO-GO, you are rejected.

What this means in practice is that Step 1 rewards clarity and speed of comprehension more than depth. Four people, working independently, are each going to form a yes/no impression of your innovation, your market and your team inside the space of a twelve-page document. There is no scoring cell to partially satisfy. An evaluator either believes, on a first read, that this is worth a closer look, or they do not. The most common way a strong project fails here is not a bad idea badly explained, it is a good idea that reads as generic on a fast pass: no sentence a reader can point to and think "that is a specific, defensible claim." Genuinely differentiated language, up front, is what a Step 1 reviewer is scanning for.

Step 2: this is where the real scoring happens

The full proposal is where the machinery gets more serious, and where the 2026 process differs meaningfully from the older model still described in a lot of consultancy blog posts. Your proposal is now assessed in parallel by two different kinds of expert: a technology expert who conducts a roughly one-hour online interview with up to three of your team members (no external consultants allowed) to fact-check your technical, commercial and financial claims and produce a due-diligence report, and an evaluation panel of at least three experts who take that report into account and set the final score.

The scoring itself is numeric: each of the three award criteria, Excellence, Impact, and Implementation (which folds in team, risk and the case for EU support), is scored from 0 to 5. Two thresholds have to be cleared simultaneously. Every single criterion needs at least 4 out of 5: one weak criterion, even if the other two are excellent, is enough on its own to sink the proposal. And the sum across all three needs to reach at least 13 out of 15 overall.

Even clearing 13/15 does not guarantee an interview. The panel ranks all proposals that pass, and only the top-ranked group, roughly 2.5 times the number that the available grant budget can fund, is invited to Step 3. A proposal can score above threshold on every criterion and still not get an interview slot in a particularly competitive cut-off, which is one of the reasons the raw success rate numbers move the way they do between cut-offs. If your proposal clears 13/15 but is not invited, it does not disappear silently: it is automatically considered for a Seal of Excellence, discussed further down.

What actually produces a NO-GO or a low score on each criterion

I am bound by confidentiality about any specific proposal I have reviewed, so what follows are patterns, not case studies. Evaluators are not looking for perfection, they are looking for a defensible claim they can stand behind if challenged.

• Excellence drops when the innovation reads as an incremental improvement dressed up as a breakthrough, or when the TRL evidence is asserted rather than demonstrated: a claim of "validated in a relevant environment" with nothing an evaluator can point to as the validation.
• Impact drops when the market sizing is top-down (a global TAM divided by an aspirational percentage) rather than built from named customers, real pricing and a credible go-to-market sequence. This is the criterion evaluators scrutinise hardest, because it is the one applicants most often pad.
• Implementation drops when the team's track record does not match the ambition of the plan, when identified skill gaps have no credible hiring plan attached, or when the budget and milestones do not obviously map to each other.

I go into these failure patterns in far more depth, with the data behind them, in why EU grant applications fail. This article is about the scoring mechanics; that one is about the writing mistakes the mechanics expose.

Step 3: the interview, briefly, because it deserves its own piece

The jury interview has enough of its own mechanics, panel composition, timing, what the jury can and cannot see, that I have covered it separately in how the jury scores you at Step 3. Two points worth stating here because they connect directly to the scoring logic above: the jury reaches its GO or NO-GO by consensus, not by counting individual votes the way Step 1 does, and the jury has access to your Step 2 remote evaluation result but explicitly does not see your short proposal, its evaluation, or the record of any earlier resubmission attempt. Every interview is judged on the full proposal and the pitch in the room, not on your history with the programme.

Reading the Evaluation Summary Report like it was written for you, not about you

If you are rejected at Step 2, you receive an Evaluation Summary Report, the ESR, built from the panel's comments and recommendations. If you are rejected after interview, the jury collectively prepares an equivalent report with the outcome and its comments. Either way, this document is the single most under-used asset in the entire process, and I say that as someone who has watched the same mistake happen to strong teams repeatedly: they read the ESR once, feel the sting, and file it away instead of treating it as what it actually is, evaluator-written raw material for the next attempt.

An ESR is written by people working through a batch under time pressure. It will not be generous with prose, and it will not always explain the full reasoning behind a number. What it will do reliably is name the criterion that fell short and gesture at why. Read it the way you would read a peer reviewer's comments on a paper: assume the reader was competent and time-constrained, not wrong. If the same criterion comes back weak across a resubmission without a substantially different answer to it, the panel evaluating the second attempt will notice, because in a meaningful share of cut-offs the applications in front of them are not first attempts.

The resubmission rule: what it actually limits

The rule that matters most for planning is what the Commission itself calls, informally, the three-strikes limit. After three unsuccessful submissions to the EIC Accelerator, at any stage, short proposal or full proposal, and under any form of support (Challenge, Open, Grant Only, Blended Finance or Equity Only), an applicant may not apply again to the EIC Accelerator under Horizon Europe. It is a hard ceiling, and it applies to the legal entity, not to the specific proposal text.

Two details soften this in practice. First, withdrawing a proposal before you receive the evaluation result does not count toward the limit, so a genuine administrative mistake caught early is not a wasted attempt. Second, a rejection at Step 1 and a rejection at Step 2 both count as one of your three, they are not tracked separately, so a founder weighing whether to push a borderline short proposal through or hold it back for another cut-off is making a decision with real consequences for how many further tries remain.

What is actually public about the people scoring you

This is the part most applicants get wrong in both directions. Some assume total anonymity, others assume they can look up exactly who reviewed their file. Neither is accurate.

All EIC experts, at every stage, are bound by confidentiality agreements and are required to be free of any conflict of interest, financial, political, family or otherwise, with the projects they assess. Applicants are given the chance, at submission, to name up to three specific reviewers they would like excluded to avoid a conflict of interest, and the identity of the specific person or people scoring your specific proposal is not disclosed to you in advance or afterward.

What is genuinely public is the pool, not the assignment. The lists of Technology Experts and of Evaluation Panel members are published on the EIC website and updated regularly as the pool rotates, with at least a quarter of each pool renewed every year specifically to limit the same reviewers seeing the same applicant repeatedly. If you are invited to the Step 3 interview, you are also told, in advance of the session, the names of the specific jury members allocated to your panel, drawn from a jury list that is likewise published and continually updated. So the pool of people who might review you is a matter of public record; which of them actually did is not, and the distinction is deliberate: it is what lets an evaluator score your work honestly without becoming a target for lobbying.

Read your own proposal like an evaluator would

Before you submit, it is worth going through your own draft the way a panel actually will: fast, looking for gates, not looking to be persuaded by tone.

• Before you reread anything, look at who has already been funded in your space. Filter the EIC Accelerator winners database by technology and country (for example, every funded medtech company since 2021). Evaluators judge novelty against what they already know exists: if companies close to yours have won, your Excellence section needs one sentence on what you do that they do not.
• Read the Excellence section alone. Is there a sentence a domain expert could not dismiss as "everyone claims this"? If not, that is a 4/5 gate at risk.
• Read the Impact section alone, and count how many market numbers are backed by a named customer, a signed pilot, or a specific price point, versus how many come from a market-research percentage.
• Read the Implementation section alone. Does the team's stated track record actually match the size of the ask? Are the gaps named, or glossed over?
• Now put all three sections back together and ask: does this read as an average of three good scores, or as three sections that each independently clear a floor? An evaluator is checking the second thing.
• If you have a prior ESR, reread it before writing a word of a resubmission. Does your new draft answer the specific criterion it flagged, in language a time-pressed reader will actually notice, or has the section only been reworded?

This is exactly the exercise the free EIC readiness assessment is built to force you through in a structured way, criterion by criterion, before you commit the fifty-plus hours a serious full proposal costs. It scores your project the way a panel would, including the same floors, and it takes a few minutes.

For a full walk-through of the application itself, Step 1 through the grant agreement, see the complete EIC Accelerator application guide. For what actually changed in the programme this year, see EIC Accelerator 2026: 4 key changes, and if you are still deciding between the Open call and a thematic Challenge, this breakdown of the five 2026 Challenges covers the fit question. You can also see who has actually won, company by company, in the database of every EIC Accelerator winner since 2021.

Questions