The narrative you intend to submit. It need not be the final version: a mostly complete draft is enough. The panel reviews it as the proposal you would submit, so you hear what a real panel would raise while there is still time to fix it.
Two documents.
Ten minutes of preparation, at most.
The panel needs less than you think, but what it gets shapes what it can tell you. This page covers the required uploads, the optional ones that measurably sharpen the review, the formats that parse cleanly, and when in your drafting cycle a run is worth spending.
What a run cannot start without.
The programme announcement, FOA, RFA, BAA, or call you are answering. This is what the panel judges you against — without it, the review has no rulebook and no criteria.
Four additions that change what the panel can say.
None are required. The budget, biosketches and letters go in the proposal PDF; prior reviews go in as a supporting PDF. Each closes a specific blind spot, named beside it.
The summary statement or panel summary from a previous submission of this work. The reviewers read it as background, so they know what an earlier panel raised and can see how this draft answers it.
The panel checks whether the money matches the science: effort that supports the aims, equipment the methods actually require, and years that line up with the timeline. A budget check also adds up the lines it can read and holds the budget to the call's ceiling, cost-sharing and indirect-cost rules, showing the figures behind each numerical flag.
Reviewers assess whether the team can do what the proposal promises. Without biosketches the panel can only judge the claim, not the evidence behind it.
Where an aim depends on a partner site, a dataset, or a facility, a letter is the difference between an assertion and an arrangement.
What parses cleanly, and what does not.
The proposal and solicitation can be up to 15 MB each. As many as ten supporting documents can go alongside them, up to 5 MB each, and the whole run can be up to 20 MB and 200 pages, which is what the review models can read at once. If a document cannot be read, the run fails and is returned to your proposal automatically, so nothing is charged.
Four things not to upload.
Describe your clinical study in the narrative. Do not upload the patient records, identifiable datasets, or anything covered by a HIPAA obligation — we do not sign BAAs.
No ITAR, EAR-controlled, or CUI content. If your programme office has flagged the work, GrantPanel is not the right environment for it.
Including one you are reviewing for an agency. Confidential peer-review material is not yours to process, and our acceptable-use policy prohibits it.
They dilute the reviewers' attention across text no panelist would read. Cite them instead — the reference audit handles the rest.
Run it early enough to act on it.
Don't wait for the final version. One proposal includes three runs, so the first can be a mostly complete draft. Start about two months before your deadline: run 1 on the draft you have, run 2 after you fix the structural findings, and run 3 in the week before submission. Runs 2 and 3 read the revised proposal against what the earlier run found, so you see what moved.
A run four days out still catches compliance triggers, but a structural weakness found four days out is a weakness you submit with.