The drafting room.

Victoria reads the design first and settles what it is — not human-subjects research, quality improvement, exempt, expedited, full board — before a single form is opened. Her reading is against the design as drawn, not the box a candidate hoped to tick: a comparison group, a validated instrument, a plan to publish generalizable findings all tell her which path the study is actually on, and she says so plainly even when it is not the answer the candidate wanted.

Every file that leaves the room passes her tolerance reading last: the same dimension — the sample, the site count, the instrument names, the recruitment wording — identical in every document the board will read together. She leads the practice's Purdue Global work and has read its files, at one stage or another of a candidate's study, every week since the practice began preparing them.

• OwnsThe determination; the document plan; the final tolerance reading on every file
• Reads firstThe proposal's method chapter — the protocol of record everything else measures from
• Her line"Claim the study you actually drew, and the category question becomes the easiest one on the form."

Grace draws the pages reviewers stop finding things in. The consent with every required element — purpose, procedures, risks, benefits, the limits of confidentiality, the right to decline or withdraw without consequence, whom to contact — and the right version stamped to the protocol it belongs to. Each instrument named once, and used with those exact words in the application, the consent and the recruitment pieces. Data-security answers with a name in them: which encrypted service, which named roles, when identifiers come off and where the key is kept, when and how the data is destroyed.

A practising advanced-practice nurse, she came to the IRB work from the clinical side of DNP projects — the side that has to live with a consent form in a real clinic — and it shows in the consents she draws: readable by the person signing them, and complete for the person reviewing them.

• OwnsThe consent; instruments alignment; the data-security plan; recruitment materials exactly as they will run
• Reads forMissing elements and version drift — the return boards are trained to find first
• Her line"Data will be kept secure is the one sentence that always comes back, because it measures nothing."

Nora drafts the site permission letters for signature so they name the actual data — which records reviewed, which people interviewed, how many minutes observed — on the site's letterhead, signed by someone with the authority to grant access. "Happy to support this study" is goodwill; she drafts permission. Then she assembles and submits the file in the form the university's portal asks for, and when the board writes back she answers it: revision responses that address each point on the record, amendments whenever the drawing changes, continuing review when the board calls for it — until the approval letter exists.

She is the consultant most candidates hear from most often, because she carries the file through the part of the process where the candidate would otherwise be alone with the portal. Her replies to the board are drawn to the same discipline as the file itself: each point answered where it was raised, each change made everywhere it appears.

• OwnsSite permission letters; the submission; revision responses, amendments and continuing review
• Reads forLetters that describe different data from the application — one dimension measured two ways
• Her line"A return is not a verdict on the study. It is a question the file could have answered before it went in."