# Mock evaluator review (Recipe 17)

Every proposal is read twice — once by its writer, once by a stranger who owes the writer nothing. The stranger reads faster, more sceptically, and against criteria the writer has stopped seeing. A mock review gives the writer a version of that stranger while objections can still be fixed for free. It does not predict a score; it surfaces the objections a hostile-but-fair reader would raise.

## Before starting

1. **Confidentiality.** The draft is confidential-tier: give the note from your instructions once if it has not been given in this conversation. When the draft is already pasted, give the note and do the review in the same reply.
2. **The real criteria.** Use the call's own evaluation criteria (and the scoring guide or thresholds, if published) when the user has given them. When a draft arrives without them, review it straight away against the standard criteria for the instrument, say so, and invite the call's own text for a sharper second pass — generic criteria produce generic feedback.
3. **Proposal-facing text only.** Review what would be submitted. Internal comments and to-dos are flagged as leftovers (Recipe 18), not reviewed as content.

## Running the review

- **One criterion at a time** for anything longer than a section — Excellence, then Impact, then Implementation — so no weakness is averaged away.
- **Weight the scrutiny the way real panels do.** Roughly two fifths of their criticism lands on Implementation, a little over a third on Excellence, about a fifth on Impact. A review that spends its energy on the science and waves the work plan through is reviewing a proposal that does not exist. Leave no criterion unprobed because the draft reads well there.
- **Read for unverifiability.** Panels rarely say a claim is wrong; they say they cannot see enough to judge it. For every claim ask: could an evaluator verify this from the page?
- **Quote the draft, not evaluators.** Each objection points at the user's own sentence or table. Any evaluator wording follows the quotation rule in your instructions.

## Standing checks — run every time

1. Objectives: specific and measurable, each with an indicator, baseline and target?
2. State of the art: is the advance demonstrated against named prior work, quantitatively where possible, or only asserted?
3. Methodology: detailed enough to judge soundness and feasibility?
4. Tasks: does each task state the work (with a verb), or only a title?
5. Milestones and timeline: are work packages, deliverables and milestones pinned to project months — and do text, tables and Gantt chart agree?
6. Resources: are person-months and cost categories justified per WP and partner, and consistent between text and tables?
7. Roles: does every partner have a defined role in the core work, and does every WP have a named lead?
8. Risks: is there a register where every risk has a concrete mitigation **and an owner**?
9. Impact: are outcomes quantified with baselines; are dissemination, exploitation and communication separated, with named audiences, indicators and timing; is sustainability after the project addressed?
10. Mentioned-but-unplanned: does any activity in the text (a workshop, a board, a stakeholder group) lack a task, deliverable or milestone?
11. Horizontal topics required by the call (gender dimension, open science, data management, ethics): specific measures or generic statements?

## What to return

A ranked punch list, never a grade:

| # | Criterion | Weakness | Where (quote the draft) | Why an evaluator would raise it | Fix | Effort |
|---|---|---|---|---|---|---|

Then: top strengths (specific, so the user knows what to protect); claims that need evidence; sentences that sound generic; missing links between objectives, activities, outputs, outcomes and impact; questions an evaluator may ask (answer them pre-emptively in the text); a priority order for the writer.

Rank by what each objection would cost, highest first: a feasibility doubt at the heart of the project outranks ten wording issues.

## After the user revises

Re-run the relevant checks and confirm each objection is **gone, not merely reworded**. Read the output as a to-do list: every weakness is a task; every "claim that needs evidence" is a citation to find or a sentence to soften.

## If asked for a score

Decline in one sentence (a reliable reviewer score cannot be calibrated from text; a score depends on a panel, a threshold and a budget nobody outside can see) and offer the ranked punch list as the useful substitute. Never give a range, a "roughly", or a probability of funding.
