MSCA DN / RAISE DN Evaluation Criteria Analysis

Definitive reference for anyone preparing a RAISE Doctoral Network proposal under HORIZON-RAISE-2026-01-03. Covers the scoring system, all three evaluation criteria with sub-aspects, panel statistics, tie-breaking rules, the RAISE dual evaluation process, evaluator expectations, and an interactive score calculator.

1. Scoring System Overview

Each proposal is scored by independent expert evaluators on a scale of 0 to 5, with one decimal place resolution (e.g., 4.3, 3.7). Evaluators first score individually, then reach a consensus score in a panel session.

Score Meanings (Evaluation Form V.2.2)

Thresholds

2. The Three Criteria with Weights

MSCA DN proposals are evaluated against three award criteria. The weights apply to the ranking calculation, not to the individual threshold determination.

3. How the Score Becomes a Ranking

The three criterion scores (each 0–5) are combined into a single ranking score out of 100 using the formula:

Ranking Score = (Excellence × 50% + Impact × 30% + Implementation × 20%) × 20

The factor of 20 converts the 0–5 weighted average into a 0–100 scale.

Worked Example

Overall Threshold Calculation

The overall threshold of 70/100 translates to: the sum of the three raw criterion scores must satisfy:

(Score1 + Score2 + Score3) × (100 / 15) ≥ 70

This means the sum of raw scores must be ≥ 10.5 (average ≥ 3.5 per criterion).

4. The 8 Evaluation Panels

Proposals are assigned to one of eight disciplinary panels. Budget is distributed proportionally to the number of eligible proposals in each panel. Cut-off scores vary substantially between panels.

5. Tie-Breaking Rules

When proposals have identical weighted total scores, they are ranked according to the following priority (MSCA Work Programme 2026–2027, p.76):

6. RAISE DN: The Dual Evaluation CRITICAL

RAISE DN proposals undergo a fundamentally different evaluation pathway from standard MSCA DN proposals. Understanding this three-step process is essential for strategy.

Step 1: Standard MSCA Evaluation

All proposals submitted to HORIZON-RAISE-2026-01-03 are first evaluated using the identical MSCA DN evaluation criteria (Excellence, Impact, Implementation) with the same weights, thresholds, and panel structure as standard DN proposals. There is no special treatment at this stage.

Step 2: Cross-Panel Reserve List

After the standard evaluation, proposals are ranked within their panel. Those funded from the main MSCA budget are removed. From the remaining reserve list (proposals that passed thresholds but were not funded due to budget limits), the highest-ranking proposals across all 8 panels are collected into a cross-panel list. The total requested EU contribution on this list is approximately 3 times the RAISE DN budget—roughly EUR 90M worth of proposals competing for EUR 30M.

Step 3: RAISE Topical Relevance Assessment

Dedicated RAISE experts evaluate each proposal on the cross-panel list for its relevance to AI in science. This is not a re-evaluation of quality (that was done in Step 1) but an assessment of whether the proposal genuinely fits the RAISE scope: is AI integral and indispensable to the research programme?

Final Ranking

The final RAISE DN ranking combines the MSCA evaluation score with the RAISE topical relevance assessment. The exact weighting between these two components is determined by the evaluation process and may not be publicly disclosed in advance.

7. What Evaluators Actually Look For

This section is based on the official Evaluation Form V.2.2, published evaluator guidance, and analysis of funded versus unfunded proposal patterns (accelopment research, MSCA-NET/RADIANCE Handbook).

Excellence: Deep Analysis

What the Evaluator Form Literally Asks

"To what extent are the research objectives ambitious, represent advancement beyond the state of the art, and are clearly defined?"

"How sound is the proposed methodology, including interdisciplinary approaches, open science practices, and the gender dimension in research content?"

"How credible and comprehensive is the training programme?"

"How appropriate are the supervision arrangements?"

Common Mistakes that Lose Points Pitfalls

• Broadly formulated objectives without measurable criteria — "We will advance the understanding of X" says nothing evaluable. Specify what will be measured, what constitutes success.
• Generic methodology without specifics — Mentioning "machine learning" or "simulation" without data sources, architectures, sample sizes, computational requirements, or validation protocols.
• Insufficient transferable skills in the training plan — Listing "communication skills" and "project management" without naming courses, providers, credit points, or assessment methods.
• Weak or unclear supervision arrangements — Not naming co-supervisors, not explaining how joint supervision works across institutions, no supervisory board structure.

Impact: Deep Analysis

What the Evaluator Form Literally Asks

"To what extent does the proposal contribute to structuring doctoral training at the European level, and to strengthening European innovation capacity?"

"How credible are the measures to enhance career perspectives and employability?"

"How suitable and effective are the dissemination, exploitation, and communication measures?"

"What is the magnitude and importance of the expected impacts?"

Common Mistakes that Lose Points Pitfalls

• Training that merely duplicates local PhD programmes — Evaluators want to see what the network provides that no single institution could. If your training calendar looks like a list of courses already available at each university, you have not demonstrated network-level added value.
• Vague sustainability plan — "We will seek to continue collaboration" is worth zero points. Name specific mechanisms: a joint degree agreement, a continuing seminar series with committed funding, shared lab infrastructure.
• Generic IP exploitation statements — Boilerplate IP text signals that you have not thought about what commercially or societally valuable outputs your research will generate.
• Missing concrete dissemination targets — "We will disseminate results widely" is not a plan. Specify: 3 journal articles per DC, 1 open-source software package per WP, 2 policy briefs for specific agencies.

Implementation: Deep Analysis

What the Evaluator Form Literally Asks

"How effective and coherent is the work plan, including appropriateness of effort allocation and risk management?"

"How suitable is each participating organisation, in terms of quality, capacity, and role in the project?"

Common Mistakes that Lose Points Pitfalls

• Deliverables at project end only, no intermediate milestones — This signals poor planning and makes it impossible for evaluators to assess progress checkpoints.
• Vague deliverables — "D1.1: Report on state of the art" is administrative busywork. "D1.1: Benchmark dataset of 10,000 annotated molecular structures with quality metrics" is a scientific deliverable.
• Generic risks without specific mitigations — "Risk: delays. Mitigation: adjust timeline." is worth nothing. "Risk: training data insufficient for convergence (probability: medium). Mitigation: pre-arranged access to partner X's proprietary dataset of 50K samples, plus synthetic data augmentation pipeline already developed in PI's previous project." is credible.
• Unclear partner collaboration mechanisms — The network must function as a network, not as parallel PhD projects. How do DCs co-author? How do WPs feed into each other? Where are the integration points?

8. RAISE-Specific Scope Requirements RAISE

The single most consequential decision for a RAISE DN proposal is whether your research genuinely fits the "AI integral and indispensable" requirement. The distinction between integral and instrumental use of AI determines whether your proposal qualifies.

The Integral vs. Instrumental Distinction

Key Requirements from Horizontal Activities WP Part 14

9. Historical Success Rates

Overall Trend

Success by Type (2025)

10. Score Calculator

Enter your estimated or actual scores for each criterion to calculate your weighted ranking score and see where you stand.

11. Self-Assessment Checklist

Use this checklist to review your proposal before submission. Each item maps to a specific evaluator concern. Unchecked items represent potential point losses.

Excellence

• Each DC project has a clearly stated, measurable research objective
• State of the art is reviewed specifically for each DC project, not just at network level
• Objectives explicitly go beyond state of the art with defined novelty
• Methodology specifies data sources, sample sizes, and computational requirements
• Validation strategy is described for each methodological approach
• Gender dimension in research content is addressed (where relevant)
• Open science practices are specified: repositories, licenses, FAIR data plan
• Training programme has a concrete calendar with named courses/workshops
• Transferable skills training is specific: courses, providers, credit points, assessment
• Inter-sectoral exposure is mandatory for every DC (secondments to non-academic sector)
• Both supervisors named per DC with complementarity explained
• Supervisory board structure and conflict resolution mechanism described

Impact

• Non-academic partners have substantive, defined roles (not just hosting secondments)
• Sustainability plan names concrete post-project mechanisms (joint programme, continuing partnership)
• Career development plan includes labour market analysis for target sectors
• Specific placement/employment targets are stated
• Dissemination plan has quantified targets (N publications, N conferences, N datasets)
• Exploitation plan addresses IP management specific to the research outputs
• Communication plan targets non-expert audiences (public engagement, policy briefs, media)
• Scientific impact is connected to documented needs, not generic claims
• Societal impact is substantiated with specific beneficiary groups or policy contexts
• Economic impact identifies potential for commercialisation, industry adoption, or standardisation

Implementation

• Gantt chart shows intermediate milestones every 6–12 months
• Deliverables are scientific (datasets, software, publications), not just administrative reports
• Risk register identifies 5+ specific risks with probability and concrete mitigation
• Each partner's contribution is mapped to work packages with person-months justified
• Recruitment strategy addresses diversity and how to attract high-quality candidates
• DC collaboration mechanisms described (joint workshops, co-authored papers, shared infrastructure)
• WP interdependencies are explicit (which WP feeds into which)
• Management structure is described: coordinator, WP leaders, advisory board
• Financial management and reporting procedures are outlined
• Hosting arrangements per DC specify institutional resources and integration plans

RAISE-Specific RAISE

• AI development is integral and indispensable (not just tool use)
• Scientific outcomes are directly dependent on novel AI capabilities
• Every DC receives dedicated doctoral-level AI training with assessed learning outcomes
• Publications target domain science venues (not only CS/AI conferences)
• AI development is visible in the WP structure (dedicated WP or cross-cutting tasks)
• Feedback loop between AI advancement and scientific discovery is explicit
• Proposal would not function as a research programme without the AI components
• Consortium includes AI methodology expertise (not only domain science expertise)

12. Sources