RIA vs IA vs CSA: Which Horizon Europe Action Type Fits Your Project

Updated 2026-07-21 · 10 min read

Action type sets your reimbursement rate, your page limit, and how your proposal is ranked. Picking the wrong call is expensive in a way no amount of good writing recovers.

Every Horizon Europe topic specifies a type of action. It is easy to skim past, but it determines how much of your costs are reimbursed, how long your proposal may be, and — for one action type — how your score is weighted against competitors. Three types cover most collaborative calls.

The three action types

• Research and Innovation Action (RIA) — activities that primarily establish new knowledge or explore the feasibility of a new or improved technology, product, process, service or solution. This may include basic and applied research, technology development and integration, testing, demonstration, and validation of a small-scale prototype in a laboratory or simulated environment.
• Innovation Action (IA) — activities aimed directly at producing plans, arrangements or designs for new, altered or improved products, processes or services. These may include prototyping, testing, demonstrating, piloting, large-scale product validation and market replication.
• Coordination and Support Action (CSA) — activities contributing to the programme's objectives that are not themselves research and innovation: networking, coordination, studies, conferences, policy support. Research activity is excluded, with a narrow exception under the widening-participation component.

Funding rates — the part that decides your budget

The reimbursement rate attaches to the action type, not to you:

• RIA — 100% of eligible direct costs, for every type of participant.
• IA — 70%, except for non-profit legal entities, where a rate of up to 100% applies.
• CSA — 100%.
• Programme co-fund — between 30% and 70%. Public procurement of innovative solutions — 50%. The EIC Accelerator grant component — 70%, capped at €2,499,999.

Indirect costs are reimbursed at a flat 25% of eligible direct costs, with defined exclusions including subcontracting and financial support to third parties. The 25% sits in the Horizon Europe Regulation itself, so no work programme can vary it.

A consequence that surprises mixed consortia: a single Innovation Action can carry two different funding rates inside one grant agreement, because the non-profit exception applies per beneficiary. A university partner at up to 100% and an SME partner at 70% in the same project is normal, not an error.

Roughly half the 2026-2027 call budget uses lump sum funding. In a lump sum grant the agreement states that the funding rate is not applicable and there is no indirect-cost article at all — the rate and the 25% become budgeting inputs you use when building the proposal, not terms you are later reimbursed against. Topic-level page limits also differ for lump sum calls.

What about TRLs?

The convention that RIA covers roughly TRL 2-5 and IA roughly TRL 5-8 is useful, widely repeated, and not an official rule. Horizon Europe's general annexes do not map action types to Technology Readiness Levels. Where TRLs matter, they are stated in the individual topic's call conditions — and the topic text overrides any general heuristic, including this one. Treat the mapping as a way to sanity-check your fit, then read the topic.

How proposals are evaluated

RIA and IA proposals are scored against three criteria: Excellence; Impact; and Quality and efficiency of the implementation. Each is scored 0 to 5, half-marks permitted. The threshold for each individual criterion is 3, and the overall threshold, applying to the sum of the three scores, is 10. CSA proposals use the same 3-and-10 thresholds with differently worded aspects. ERC evaluates on Excellence alone.

Here is the rule most secondary guides state incorrectly, usually by asserting the three criteria are equally weighted: for Innovation Actions, the Impact score is given a weight of 1.5 when determining the ranking. Research and Innovation Actions are unweighted. Crucially, the weighting affects ranking only — the 10-point overall threshold is computed on the unweighted sum even for an IA.

Where proposals tie, a defined cascade decides priority, and it does not start with the scores:

1. Proposals addressing aspects of the call not covered by more highly ranked proposals take priority — portfolio coverage outranks any score comparison.
2. Then the Excellence score, then Impact. For Innovation Actions this reverses: Impact first, then Excellence.
3. Then gender balance among the researchers in leading roles.
4. Then geographical diversity, measured by Member States and Associated Countries not already represented higher in the ranking.
5. Then factors at the panel's discretion, recorded in the panel report.

"Quality and efficiency of the implementation" never enters the tie-break. It can only sink you by falling below its threshold.

Page limits

Page limits attach to the work programme, not to the calendar year, so a call published under an earlier work programme keeps its old limits even if its deadline falls in 2026. Under the 2026-2027 work programme:

• RIA and IA full applications — 40 pages, or 45 for lump sum topics.
• CSA — 25 pages, or 28 for lump sum topics.
• Programme co-fund — 65 pages.
• First-stage applications in two-stage calls — 10 pages.

These are reductions from the previous work programme, where RIA and IA sat at 45 pages and CSA at 30. A guide quoting 45 flatly for a 2026-2027 RIA is working from the older cycle. Evaluators are instructed to disregard anything past the limit.

Two-stage calls

Some topics run in two stages. Stage 1 is a 10-page outline evaluated against only two criteria, Excellence and Impact, and within those only the aspects printed in bold. Successful applicants are invited to submit a full application at stage 2.

Two features of stage 1 run against intuition. The per-criterion threshold at stage 1 is 4, higher than stage 2's 3. And the overall threshold is dynamic rather than fixed: it is set so that the total requested budget of proposals admitted to stage 2 is as close as possible to three times the available budget, and not less than twice it — typically landing at 8 or 8.5. A score of 8 may or may not carry, depending on the competition.

That multiplier loosened for 2026-2027, from a previous two-times target. More proposals now reach stage 2, which makes stage 2 more competitive than older guidance implies. Your full application must also stay consistent with the outline — substantial changes between stages are not permitted.

Submitting: the practical traps

• Everything goes through the EU Funding & Tenders Portal. Paper submission does not exist.
• Use the forms inside the electronic submission system. The templates published on the topic page are for information only, and building your proposal on them is a recurring, avoidable mistake.
• Every participant needs a 9-digit Participant Identification Code (PIC), but a validated PIC is not a prerequisite for submitting. Validation is run after evaluation, and only for beneficiaries and affiliated entities — not associated partners or subcontractors.
• Deadlines are stated per call and are, in practice universally, 17:00:00 Brussels local time — meaning CET or CEST depending on the date. Late is disqualified.

One further condition catches organisations rather than proposals: a Gender Equality Plan is required of public bodies, research organisations and higher education establishments established in Member States and Associated Countries. It does not apply to private for-profit companies including SMEs, or to NGOs. It is checked at grant signature rather than at submission, so it will not block you from applying — but it can hold up a grant you have already won.

Choosing, in one line

If you are generating knowledge or validating something in the lab, you want a RIA at 100%. If you are demonstrating, piloting or replicating something close to market, you want an IA at 70% — and you should write to Impact, because Impact is what ranks you. If you are coordinating, studying or networking rather than researching, you want a CSA. Then read the topic text, which overrides every generalisation above.

Frequently asked questions