The 2026 call for 3Rs Support Grants closed with 112 applications, around ten times the number received by any comparable earlier round. That was far more than the budget could fund or the expert panel could review in depth, so we changed the assessment procedure after the call had closed. This page explains what we did, what was funded, and what might change for the next call in 2028.
| Applications received | 112 |
| Passed the minimum-requirements screen | 112 |
| Drawn for full expert review | 12 |
| Funded | 7 |
| Total awarded | CHF 137,000 (of which CHF 45,000 restricted to Refinement) |
| Decision confirmed | 3 September 2026 |
Why the procedure changed
The Support Grant scheme is new in 2026. It succeeds the Refinement Grants, widened to all three Rs, and the two predecessor rounds received 14 and 12 applications. The call was planned on that basis: every application reviewed in depth by a small expert panel, and written feedback for every applicant.
Reviewing 112 properly would have meant recruiting and contracting a panel several times larger before any review could begin. Reviewing them with the panel we had would have meant a shallow assessment for every applicant and a proper one for none.
Underneath the capacity problem was a harder one. Support Grant applications are short by design, and ranking 112 of them from first to last to draw a funding line at seven would claim a precision that does not exist: at that level the differences between strong proposals reflect how an application is written at least as much as the project behind it. Research on grant peer review consistently finds that fine-ranking near-equivalent proposals is unreliable and tends to favour applicants and institutions with greater name recognition. A pre-screen would not have solved this. The difficulty was not an excess of weak applications but an excess of good ones that could not be reliably separated.
So we drew at random among the applications meeting the call’s requirements, and reviewed those drawn in full. We would rather state the trade-off plainly than not: this procedure does not claim to have funded the best seven applications of 112. It claims that every eligible application had an equal chance of being drawn, and that every funded project was properly reviewed. Draws of this kind are established practice in research funding. The Swiss National Science Foundation has used one since 2021 to decide between applications it cannot separate on quality, and the Austrian Science Fund, Wellcome, the British Academy and the Health Research Council of New Zealand all run draws among applications judged fundable.
How the round was assessed
Screening. Every application was checked by the 3RCC Directorate against the call’s basic requirements: eligibility, budget and relevance to the 3Rs. This was not a scientific assessment and not a ranking. All 112 passed, and none was excluded from the draw.
The draw. On 16 July 2026, a structured random draw selected 12 applications from the 112, using a published seed and the selection code set out at the foot of this page. It was constrained so that the 12 would be spread across the field: at most 3 applications from any one institution, 5 from any one regional hub and 4 in any one research field; at least one each addressing Replacement, Reduction and Refinement; and three Refinement applications drawn into a separate, restricted funding stream. These are anti-concentration limits rather than quotas, and the reason for them is what the scheme is for. Support Grants are seed funding, and seed does more good spread across the landscape than concentrated in one patch.
Review. Each of the 12 was reviewed independently by three members of a panel of seven international 3Rs experts, three of them recruited for this round so that every application could be matched to reviewers in its research field and its 3Rs focus. This was the full assessment the call promised: scientific quality, impact on the 3Rs, feasibility, communication and openness. The panel did not rank. A project was funded only if all three of its reviewers recommended funding, and a single “no” was a veto, to be used only for a disqualifying flaw and never to choose between fundable projects. Each dissenting reviewer was asked to reconsider in the light of the other two reviews before the outcome was fixed.
Decision. The panel’s recommendations were confirmed on 3 September 2026 by the 3RCC Management Team, on which both the Executive Board and the Strategic Board is represented.
What was funded
Seven projects, CHF 137,000 in total. We will announce them individually once the awards have been accepted.
The draw was built to spread the round across the 3Rs, across institutions and across research fields, and it did:
- Three of the seven projects address Replacement, five Reduction and three Refinement; several address more than one.
- The seven awards go to seven different institutions, one each, in five regional hubs. The 112 applications came from 14 institutions.
- The seven projects span eight of the sixteen research fields represented in the applications.
Three things those numbers do not show, which we would rather say than be asked.
Five of the twelve drawn projects were not funded, where the design expected eight to ten awards. Four were blocked by a single dissenting reviewer with the other two in favour; the fifth was declined by two of its three reviewers. In three of the six “no” votes cast across the round, the objection was that the project fell outside the scheme’s 3Rs remit rather than that its science was weak. That is a finding about the call, and it is addressed below.
Oncology and immunology were not drawn at all, though together they account for a third of the applications. Re-running the same constrained draw 50,000 times over the same pool, the chance of drawing neither is about 1 in 110. It was an unlucky draw rather than a fault in the procedure, but it shows that caps limit concentration without guaranteeing coverage.
The funded set leans towards Refinement: three of seven awards, from 16 of the 112 applications. CHF 45,000 of the round came from funds held on behalf of the former 3Rs Foundation of Switzerland, restricted to refinement research and unusable for Replacement or Reduction work. Three Refinement applications were therefore drawn into a separate stream, of which at least two had to be funded; two cleared review. Refinement applications consequently had a higher probability of being drawn. Had we not run a separate stream, the CHF 45,000 could not have been used at all. The restriction is now fully discharged, and no equivalent earmark applies in 2028.
Checking the draw
We cannot publish the data the draw ran on, because it describes 112 confidential applications. We can publish the seed and the selection code itself, which is where the balancing described above is actually implemented, and it is reproduced below as it ran.
The draw makes exactly one random decision. The seed is the codeword "2026", the call year, chosen so that it is evidently not picked to produce a particular result. It shuffles the 112 applications into a random order. Everything after that is deterministic: the draw walks the shuffled order three times, first seating three Refinement applications into the restricted stream, then seating any of the three Rs not yet represented, then filling the shortlist to twelve, and at every step skipping an application that would breach a cap.
This is the function that does it, from the published script. Only the bookkeeping that records each pick’s stage for the draw record has been left out:
# Parameters, fixed before the draw (SG26_lottery_design.md §B4–B5)
K = 12 # applications drawn
FOUND_REF = 3 # Refinement applications for the restricted stream
MAX_ORG = 3 # cap: applications per institution
MAX_HUB = 5 # cap: applications per regional hub
MAX_FIELD = 4 # cap: applications per research field
FIELD_EXCLUDE = {"Other", "Multisystemic Studies"} # catch-alls, deliberately uncapped
def official_draw(pool):
order = pool[:]
random.shuffle(order) # the ONLY RNG call
oc, hc, fc = Counter(), Counter(), Counter()
sel, found, sids = [], [], set()
def fits(a):
if oc[a["org"]] >= MAX_ORG or hc[a["hub"]] >= MAX_HUB:
return False
return all(fc[x] < MAX_FIELD for x in cap_fields(a))
def take(a):
sel.append(a); sids.add(a["id"])
oc[a["org"]] += 1; hc[a["hub"]] += 1
for x in cap_fields(a):
fc[x] += 1
# 1. 3Rs Foundation stream: draw 3 Refinement
for a in order:
if len(found) >= FOUND_REF:
break
if a["r"] == "Refinement" and a["id"] not in sids and fits(a):
take(a); found.append(a)
# 2. Seat >=1 of each R
for R in ("Replacement", "Reduction", "Refinement"):
if any(s["r"] == R for s in sel):
continue
for a in order:
if a["r"] == R and a["id"] not in sids and fits(a):
take(a); break
# 3. Fill to K
for a in order:
if len(sel) >= K:
break
if a["id"] not in sids and fits(a):
take(a)
return sel, found
cap_fields returns the research fields an application lists, minus the two catch-all categories above. Note what fits can and cannot do: a cap is only ever a test that skips an application, so the balancing can never promote one application over another on merit. It can only decline to seat one that would over-concentrate the shortlist. In this draw, no cap was reached.
The walk needs each application’s institution, regional hub, research fields and 3Rs category, which are confidential, so it can be read and checked line by line but not re-executed from outside. The full script also re-runs the draw through a separately written implementation, the one used for the advance feasibility testing, and stops if the two disagree.
What this means for applicants
Because the procedure was designed after the call had closed, applicants were not told of it before the outcome. That should not have happened, and the changes below are meant to ensure it does not happen again.
Applications that were not drawn were not assessed on merit: no external reviewer read them, and there is no written feedback for this round. That is the part of this procedure we are least comfortable with, and a direct consequence of the volume. Not being drawn is not a judgement on a project. Applicants whose applications were drawn have received all three written reviews in full, and where funding was refused, the specific objection that decided it.
The next Support Grant call is planned for 2028. Applicants are encouraged to contact their local 3Rs Node Coordinator for advice on alternative sources of funding in the meantime.
What changes might come for Support Grant 2028
The draw was a response to an exceptional situation, not a change of policy. Our aim for 2028 is a procedure that can assess every application on its merits at whatever volume the call attracts, and that is set out in full in the call itself. The measures under consideration will be settled before the call opens:
- announcing the call in January rather than April, leaving room to adapt if the response again exceeds expectations
- a two-stage process, with a short letter of intent first and full applications only from those invited to submit them, so that if a draw were ever needed again it would happen before anyone had written a full proposal
- a more precise definition of the scheme’s 3Rs remit, settled at screening rather than left to reviewers
- stating the expected number of awards in the call, with the budget sized on the maximum request rather than an average
A cap on applications per institution and a narrower, more targeted call were both considered and set aside for now: the first works against the collaboration the scheme exists to encourage, the second against its researcher-driven character.
Contact
Questions about this procedure can be sent to funding@swiss3rcc.org.