Outcome record
Board-pack reporting draft, timed with a second marker
In June 2026, eleven learners on Reporting and Drafting with Verification assembled a first complete draft of an anonymised monthly pack twice. Median time moved from 74 minutes to 31 minutes. Mean factual gaps found by a second marker moved from 6.2 to 1.8 per pack. Source marks were required on the after-pass; they were optional on the baseline, which is how most of the group already worked.
Course details All outcome records
What the sample looked like
The pack was a ten-page monthly report for a fictional SME: revenue, two operational KPIs, a short risk note, and a decisions list. Source files were a spreadsheet export, a three-page operations memo, and last month’s pack with two figures that had been updated in the export and not in the memo. Names of staff and customers were tokens. Currency was Singapore dollars.
The trap in the pack was the stale memo figure. Learners who drafted from prose first copied it. Learners who locked numbers from the export first caught it. That trap is why we measure gaps as well as minutes. A faster draft that carries a stale total is a worse pack.
Protocol on the day
Baseline: produce a complete first draft in whatever way you already do, including any assistant you already use. The clock started when the folder was opened and stopped when the learner said the draft was ready for a director. Completeness meant every heading had content and every figure had a number, even if the source was not marked.
After the rebuild: same pack, new blank document, source-marking rule in force. Every figure needed a pointer to export tab or memo paragraph. The assistant was allowed to shorten prose only after numbers were locked. The second marker counted gaps: unmarked figures, figures that disagreed with the export, period labels that did not match the cover, and invented commentary that was not in either source.
The numbers
| Measure | Baseline | After |
|---|---|---|
| Median time to complete first draft | 74 min | 31 min |
| Mean factual gaps (second marker) | 6.2 | 1.8 |
| Learners kept | 11 | |
| Date | June 2026, Fusionopolis open class | |
Gaps are a count of discrete faults. A pack with one invented total and one stale KPI scores two. The after-pass still had gaps: mostly unmarked percentages in the risk note, where learners treated a phrase as qualitative. We kept those ticks. A director reading the risk page would still ask where the percentage came from.
What learners built
The skeleton followed the headings of the fictional pack, with placeholders the assistant was forbidden to fill from memory. The source-marking rule was a footnote style: [E:Tab2] or [M:p2]. The fact-check checklist had four forced questions: period, unit, comparison base, and “unknown” when neither source held the claim.
The redraft pass was last. Several baseline packs had fluent prose and a stale total. After the rebuild, prose was plainer and the stale total was gone in ten of eleven files. The remaining file had the correct total in the table and the stale one in the executive paragraph; the second marker counted that as a gap, correctly.
Limits that belong next to 31 minutes
The pack was ten pages with two sources. Live packs have more systems, last-minute comments, and a design template that fights the writer. The room was a training room. Directors were imaginary. We measured time and gaps. We did not score whether a board would have taken the decision on page nine.
Sample size is eleven. We do not generalise from that to “reporting in Singapore”. We also do not have a later re-measure on this cohort. If your organisation’s risk is a regulated disclosure, this classroom file is the wrong object; say so on the scoping call so we can decline the work.
Measure this on your own pack
Bring last month’s pack with names stripped, plus the exports that fed it. Agree the gap definitions before the clock starts. Programme page: Reporting and Drafting with Verification.