- A review round has to produce a ranking somebody can explain months later, not just a ranking. That means the score, the rubric version, the reviewer and one sentence of reasoning are all part of the output.
- Calibration is the step that decides whether scores compare at all. Have every reviewer score the same two or three applications before the round opens, then compare. What it usually exposes is a badly written criterion, not a bad reviewer.
- Conflicts of interest belong at panel formation, not per application. A recusal nobody recorded looks identical to an absence when the round is questioned.
- Never average scores across a shortlist boundary. Averaging hides disagreement, and the place it hides it is exactly where the decision is being made.
A grant review process is the sequence a funder runs between a set of eligible applications and a funding decision: assembling a panel, calibrating it, scoring against a published rubric, resolving disagreement, and recording the basis of each decision. What it has to produce is not a ranking. It is a ranking somebody can explain six months later.
The rubric gets most of the attention in this stage. In practice, rounds fail on the steps around it.
What a review round has to produce
Three outputs, and only the first one is obvious.
A ranking, so a funding line can be drawn somewhere in it. A documented basis per decision, so any single outcome can be explained without reconstructing it from memory. And a defensible account of the declines, which is the output most rounds neglect and the one most likely to be tested.
A round that produces only the ranking has done the work and lost the evidence.
Who sits on the panel
Two reviewers per application is the working minimum, and three is worth the cost for applications sitting near the funding line, where a single divergent score changes the outcome.
Composition matters more than size. A panel needs programme knowledge, so someone recognises what the call was for; subject knowledge, so someone can judge whether the proposed work is credible; and where the funding affects a specific community, someone with direct experience of it. Panels built only from programme staff reward proposals that are well written rather than well designed.
External reviewers raise the credibility of a decision and cost calendar time. Budget for the delay rather than discovering it.
One role should not be a reviewer at all. The chair runs the round, holds the rubric, and arbitrates disagreement. A chair who also scores loses the standing to resolve a dispute between two people who did.
Calibration, the step that decides whether scores compare
Scores from different reviewers are only comparable if the reviewers were assessing the same thing. Calibration is how you find out whether they are, and it happens before the round opens, not after.
The practice is simple. Every reviewer scores the same two or three applications independently, then the panel compares results side by side. Where scores diverge, the panel talks through why.
What this exposes is usually not a bad reviewer. It is a criterion the scoring rubric describes badly. Two competent people reading impact as reach and as depth will score the same application four points apart and both be right, and no amount of averaging afterwards fixes that. Rewriting the anchor does.
Skipping calibration does not remove the disagreement. It moves it out of sight and into the totals.
The blind first pass, and when to lift it
Blind review on the first pass, with identity revealed for the capacity assessment, is a common and sensible pattern. Reviewers judge the proposal on what it proposes, then judge the organisation on whether it can deliver it.
Be honest about what blinding can achieve. Removing an organisation's name does not remove its identity from a proposal that names its project sites, its partners, its founder, or a budget scale only a handful of organisations could carry. A reviewer who knows the field will often recognise the applicant anyway.
That does not make the practice pointless. It removes the automatic advantage of a familiar logo on the first read, which is a real effect, while making no claim to anonymity it cannot deliver.
Conflicts of interest, settled before the round opens
Conflicts get declared at the wrong moment in most rounds. Asking a reviewer to flag a conflict when an application appears in their queue relies on them recognising a name in a list of forty, at the point when they are trying to get through the list.
Declare at panel formation instead. Each reviewer lists the organisations they sit with, advise, have funded, or have a personal connection to, and the assignment step excludes those applications before anyone sees them. The exclusion should be a property of the system, not of the reviewer's memory.
Then record it. A recusal nobody wrote down is indistinguishable from a reviewer who simply did not respond, and the difference matters when someone asks how a decision was reached. Our note on keeping an audit trail covers what the record needs to hold.
When two reviewers disagree
Divergence is information, not noise. How a panel handles it decides whether the round is credible.
One rule overrides the table. Never average across a shortlist boundary. Averaging a 4 and an 8 produces a 6, and a 6 may sit exactly where the funding line falls, which means the method has quietly made the decision that the panel was supposed to make.
The decision meeting
The panel's job in the room is narrower than it usually assumes. The ranking already exists, produced by the scoring. What the meeting decides is where the line falls, and what to do with the applications immediately either side of it.
Time-box the discussion per application and hold to it. Rounds run long because panels re-litigate applications that were never near the line, and the cost lands on the ones that were.
Near-misses deserve a named outcome rather than a silence: declined, declined with an invitation to reapply, or held for a later round if the programme works that way. Application scoring gets an application to the line. Only the panel can decide what happens to the ones sitting on it.
Contact us or request a demo to stop wasting time on spreadsheets— and start managing grants with speed and clarity.
Documenting a round so a decision survives being questioned
Six months after a round closes, someone will ask why a particular application was declined. The question comes from a board member, a repeat applicant, an auditor, or a colleague who has inherited the programme.
Five things answer it, and all five have to be captured while the round is live: the score per criterion, the version of the rubric in force, which reviewer produced it, the date, and one sentence of reasoning. The sentence is the part teams skip and the only part a person can actually read.
Reconstructing this afterwards from a spreadsheet and an inbox is the situation to avoid, because the reconstruction is exactly what cannot be trusted.
Where review sits in the funding cycle
Review belongs to pre-award, after the filters and before the award. What reaches it should already be eligible and complete, which is the job of the stages before it: our guide to automated grant application screening covers those, and the letter of intent stage is where the pool gets narrowed before any reviewer reads a full proposal.
If you are assessing platforms rather than process, the comparison of grant management software covers scope and pricing, and the guide to grant management systems starts from the process instead.
How Optimy handles the review round
Optimy routes shortlisted applications to internal or external reviewers with conflict-of-interest controls applied at assignment, and keeps scores hidden between reviewers until the review period closes, so one early score cannot anchor the rest of the panel.
Criteria and weights are held as a structured rubric rather than a spreadsheet, so every reviewer sees the same anchors and the totals stay comparable across a whole round. Where two reviewers diverge sharply on one criterion, the gap is surfaced for the reviewer panel to discuss rather than averaged away silently. Every score, decision and recusal stays on the application record.
See the grant management software page, or book a walkthrough to run your own rubric through it.
