An independant checklist for HORIZON Europe proposals

For the past eight years, I have participated in the evaluation of Horizon (2020) project proposals as independent expert for the European Commission. Those assessment have not only been a great opportunity to challenge my technical knowledge and meet a rich variety of experts in my field, but also have help me refine the key steps in defining a project for maximising its chances to 1-be funded and 2-succeed. This article presents a condensed check-list targeting proposal writers for screening their document in order to avoid the most common mistakes I encountered, sadly prevented very appealing projects to rank up the list.

Disclaimers:

• This article has been written independently from any institution, public or private, and only represents my personal learnings;

• It does not aim at covering all aspects of a proposal of the three criteria - Excellence, Impact, Implementation - but only my Data Product Architect point of view on those.

1. Pertinent impact assessment

Clearly distinguishing objectives, milestones, implementation KPIs, and impact KPIs ensures a structured approach to project planning, enabling effective tracking of progress, resource allocation, and long-term impact assessment. Those elements are the foundation for a credible project proposal. All elements in the list are absolute must to present a comprehensive picture of the project's impact potential.

✔ Objectives: The overarching goals that define the purpose and ambition of the project. Those objectives relate to the Excellence of the solution proposed, and should reflect innovation beyond the state-of-the-art.

Example: "Develop an open-source API that enables SaaS companies to accurately measure and report their environmental impact."

✔ Milestones: Key checkpoints that indicate progress toward achieving the objectives. The delivery of deliverables is not a milestone, it's a russian doll.

Example: "Release the beta version of the API with carbon footprint calculation capabilities by Month 18."

✔ Implementation KPIs: Metrics that track execution efficiency and project tasks and activities. The number of components developed is NOT an interesting KPI as it does not prove anything towards the global objective.

Example: "Number of companies integrating the API during pilot testing."

✔ Impact KPIs: Indicators measuring the long-term effects and real-world outcomes of the project. This section should go beyond the technical development and reflect on the "why" the solution requires public ressource funding.

Example: "Percentage reduction in carbon footprint among SaaS companies using the API within the first two years."

⭐ Note that for each KPI presented, a baseline with strong literature references is expected, as well as the number expected by the project consortium. Also note that a percentage with no volume value attached would be considered lacking precision.

Complete impact example: By month 36, 80% of the companies involved in the pilots present a 15% carbon emission reduction (~2,250 metric tons of CO₂e), placing them in a strong position to be carbon-neutral by 2050.

👉 Refer to this article how to write SMART KPIs

2. Describe your data

In the context of big data processing, data description became more crucial to the project's description, in order to assess budgets allowances for technical equipment, but also to assess if the solution proposed is adequate in terms of engineering efforts.

✔ the 4 V

Big data is defined by Volume, Velocity, Variety, and Veracity. Volume impacts storage and computing needs, while Velocity dictates processing speed (real-time or batch). Variety covers data formats and sources, requiring flexible pipelines. Veracity ensures data quality for reliable analysis. Assessing these four dimensions helps optimize budgets and engineering efforts.

✔ data availability

Data availability depends on timing (real-time, batch), ownership (internal teams, partners, public sources), and access methods (APIs, databases, data lakes). Clear mapping of these factors ensures seamless integration, security, and alignment with project needs.

3. Make pilots user focused

Project methodology, in particular in data-driven environment, is the backbone of a successful project development. By frequently aligning end-users' objectives with the projects technological innovation, we maximise the impact of the efforts invested in the solution. We also guarantee to allow the solution to evolve with the market and emerging new concepts or tools throughout the lifetime of the project.

✔ User requirements gathering happens before the architecture design phase

✔ Frequent feedback loops with end-users are planned

✔ Meeting the broader users beyond pilots and their feedback integration is planned

✔ Appropriate communication methods, channels, vocabulary is planned with end-users

✔ Plan to measure the uptake of the proposed solution

Additionally, pulling the end-user at the center of your development strategy often ensure serious uptake, greater involvement and results, facilitates the go-to-market phase and the commercialization of the solution beyond the scope of the project. Should we add that it increases the credibility of each member of the consortium in their area of expertise?

4. Build towards fair data science and AI

Many new aspects of information processing are being discussed and normalized within the European space. In particular, the Data Act, the AI Act, and the Do No Significant Harm" principle are ambitious regulatory frames in our world towards a fairer use of our resources. Those should not only be treated as industry-standards (they are not new anymore!) but ideological references to build for a better future. Find below an applied-checklist to consider.

1. Data Governance & Compliance

✔ Clearly define data ownership, access rights, and responsibilities in contracts (EU Data Act)

✔ Classify AI system risk level (low, high, or unacceptable) and document mitigation measures (AI Act)

✔ Provide data-sharing mechanisms (APIs, dashboards) with clear usage policies

2. Privacy & Security

✔ Anonymize or pseudonymize personal data before processing (GDPR)

✔ Encrypt sensitive data in transit and at rest, using industry standards (AES-256)

✔ Implement user consent flows with opt-in/opt-out options for data collection

3. Sustainability & Ethical AI

✔ Measure and report energy consumption of data processing (CO₂ equivalent per query)

✔ Run bias audits on training datasets before model deployment

✔ Justify model complexity to balance performance with energy efficiency

4. Transparency & Explainability

✔ Maintain a data lineage log detailing sources, transformations, and retention policies

✔ Provide a model explainability report (e.g., SHAP values for AI decisions)

✔ Publish a user-friendly FAQ on how data is used and decisions are made

5. Presentation, readability

Last but not least, experts reading your proposals are humans. Humans with human eyes to grasp your document, human hands for flipping the pages, and human feelings when they get the sense of being fooled.

✔ overall the document is pleasant to read: font size is readable, colors are meaningful (and colorblind-friendly), document outline is clear, spaces between paragraphs and page margins give us time to breath

✔ page number are not just nice-to-have: they are key for internal references and discussions between experts.

✔ architecture diagrams are easy to understand, clear connections between components appear and are labeled ; graphs present state-of-the-art features (title, axis title, number scale and digits significance, appropriate choice of data visualisation)

✔ with no offense to non-native english speakers intended, the documents reads well and follows a natural logical flow - also style is homogeneous between paragraphs is appreciated. Grammar and punctuation should not be neglected

✔ copy/paste from other parts of the documents is a no-go, just as copy-paste from external sources, previous or competing proposals

✔ references to other paragraphs within the document are limited, and if present they are double-checked to avoid any contradictions or mismatch.

---

Writing a project proposal and interested in having it screened by experts familiar with the review process? Please do get in touch. I would be more than happy to offer pre-submission advise on the aspects mentioned above, in the sole name of impact-measurement quality, free of conflict of interest.