Score a grant proposal by applying agreed criteria and anchored ratings to the evidence supplied for the round. Keep eligibility separate, record the passage behind each judgment and calculate the weighted total under a documented rule. Review disagreements, missing information and conflicts before the committee decides. A score organizes judgment; it does not verify every applicant claim or remove bias by itself.
This lesson is for funders and selection teams applying an existing rubric to proposals. You will create a scored review record with source references and an uncertainty note. If the round’s criteria are not yet agreed, complete that design before scoring actual applicants.
- Confirm eligibility, criteria, weights and the missing-evidence policy.
- Define rating anchors and acceptable evidence.
- Review each criterion independently against its sources.
- Calculate the total and check the arithmetic.
- Calibrate disagreements and handle permitted clarifications consistently.
- Record the committee decision and its reasons.
What belongs in the rubric?
Include criteria that support the round’s purpose. Separate eligibility rules from scored judgments and administrative information. An application identifier should not influence merit. A mandatory eligibility condition should not be quietly compensated for by points elsewhere unless the approved policy explicitly permits that treatment.
Criteria and evidence expectations should fit the applicants and intended work. Do not automatically reward organizational size, polished prose or a long track record when those are not relevant requirements. PEAK Grantmaking’s guidance on bias encourages examining the application and decision process itself, not only individual reviewer intentions.
How do rating anchors help?
An anchor describes what a rating means for a particular criterion. “Excellent” alone leaves too much interpretation. For feasibility, a higher rating might require a coherent plan, relevant responsibilities and a credible schedule within the proposal’s scope. The exact standard belongs to the round, not to a generic AI prompt.
| Illustrative feasibility anchor | Evidence judgment |
|---|---|
| 1 — major weaknesses | The supplied plan has substantial unresolved gaps that undermine delivery under this round’s criteria. |
| 3 — adequate basis with questions | The plan is plausible, with specific uncertainties the review should record. |
| 5 — strong supported basis | The relevant plan, roles and timing are clearly supported in the supplied material. |
Define intermediate ratings if you use them. Keep “not yet assessable” separate from a low rating when the round’s policy allows clarification. Missing evidence is not automatically a failure, but neither is it automatically an invitation to submit new material after the deadline. Apply the published process consistently.
What should the review record contain?
For each criterion, record the rating, source location, reason, uncertainty, reviewer and rubric version. An attachment may corroborate a statement, contradict it or simply repeat the applicant’s claim. A citation establishes where the information came from; it does not make it independently verified.
In a fictional proposal, the narrative says that the delivery team is ready. A staffing appendix lists responsibilities but leaves one role unfilled. The reviewer can note a specific staffing uncertainty under feasibility. They should not convert “ready” into a verified fact or penalize the same missing role repeatedly under every criterion without a justified rule.
How do you calculate a weighted score?
State the formula before review. The following fictional example uses ratings from 1 to 5 and weights totaling 100%. Each contribution is rating divided by 5, multiplied by the criterion’s weight.
| Criterion | Weight | Rating | Contribution |
|---|---|---|---|
| Fit with round purpose | 30% | 4 | 24 points |
| Feasibility | 30% | 3 | 18 points |
| Community relevance | 25% | 4 | 20 points |
| Resource justification | 15% | 2 | 6 points |
| Total | 100% | — | 68 of 100 possible points |
Under this formula, all ratings of 1 produce 20 points, not zero. Do not describe 68 as a 68% probability of success or a validated measure of impact. It is the result of this rubric. Other transformations are possible, but the rule must be explicit and stable for the round.
If a required criterion is unassessed, do not silently substitute zero or average only the available criteria. Show a provisional status and apply the agreed policy. A complete-looking total can conceal a missing judgment.
How should reviewers handle disagreement?
Have reviewers first explain which anchor and evidence led to their rating. A difference may reveal ambiguous criteria, overlooked material or a legitimate judgment difference. Record the resolution rather than averaging away the reason for disagreement.
Test the rubric on varied sample proposals before full review. Include different languages and presentation formats where relevant to the round. Document conflicts of interest and apply the assignment process. A fixed rubric can still encode unsuitable expectations, so review the criteria as well as reviewer consistency.
What happens after clarification?
Keep the request, response and review decision with the application. Reassess affected criteria and check whether the new information also changes another part of the review. Do not assume only one score can change. Retain the earlier ratings and the reason for revision.
A supplied budget table may resolve a missing attachment while raising a new feasibility question. Receipt is not the same as approval. The committee remains responsible for the final decision, including any permitted considerations beyond the numerical total.
How can Sopact assist?
You can run this method with a rubric and source register. In a configured Sopact workflow, applications and attachments can stay connected while AI suggests ratings and source passages against the agreed rubric. That can reduce repeated reading and make differences easier to review.
Test extracted passages, missing-evidence handling and rating suggestions across representative applications. Require human review and documented overrides. Do not claim that citations eliminate bias, that generative scoring is deterministic or that the system validates every claim. Keep reviewer access and shared outputs appropriate for the application material.
Practice: review a provisional score
Use the fictional weighted table. Write a reason and source reference for feasibility, then mark resource justification unassessed. Explain why the total should now be provisional. Add a clarification response and record which criteria require another look before a committee decision.
Frequently asked questions
Should eligibility receive a weight?
Usually it serves a different role from comparative scoring. Define its treatment in the round’s rules and keep it visible. Do not let a weighted total silently override a mandatory condition.
Does a source citation verify a claim?
No. It identifies the source of the statement. Assess whether that source provides independent evidence, supporting detail or only a repeated assertion. Keep the evidence status explicit.
Can missing information be clarified?
Apply the round’s stated process consistently. Some gaps may be eligible for clarification and others may not. Record the decision and avoid creating a different opportunity for applicants based on an informal reviewer preference.
Does the highest score automatically win?
Only if the approved selection process says so. Scores are decision support within that process. The responsible committee records the final outcome and any authorized reasons beyond the total.