10 · Publication and Correction Policy
Use: What we publish, how, where, when — and exactly what happens when we get a number wrong. Owner: Head of Impact. Reviewed: Annually + after every material correction.
What we publish
| Artefact | Audience | Cadence | Owner | Location |
|---|---|---|---|---|
| Quarterly impact report | Per client | Within 10 working days of quarter end | Account Manager → Head of Impact sign-off | Direct to client + Evidence Room |
| Annual impact report (file 03) | Public | Within 90 days of FY end | Head of Impact | memacademy.org + downloadable PDF + anonymised dataset |
| Board impact dashboard (file 06) | Board + staff | Monthly | Head of Impact | Internal |
| Methodology (file 04) | Public | On change, with versioning | Head of Impact | memacademy.org |
| Financial proxies table | Public | Annual Q1 refresh | Head of Impact | memacademy.org |
| Corrections log | Public | Continuous | Head of Impact | memacademy.org/corrections |
| Anonymised dataset (annual) | Public | With annual report | Head of Impact | memacademy.org |
Every public artefact lives at a stable URL, is versioned (v1.0, v1.1, …), and links back to its Evidence Room records (file 08).
What we do not publish
- Numbers that do not have an Evidence Room record.
- Numbers below the suppression threshold (file 05).
- Named participants without consent (file 09).
- Named clients without written consent on the exact wording.
- Forecasts of SROI.
- "Industry benchmarks" we can't source.
- Marketing-only "case studies" that bypass file 09.
Sign-off chain — every report, no shortcuts
| Step | Owner | Output |
|---|---|---|
| Draft prepared | Account Manager (quarterly) / Head of Impact (annual) | Draft + evidence_refs for every claim |
| Methodology check | Data Lead | Confirms every claim resolves to an Evidence Room record |
| Safeguarding check | Head of Coaching | Confirms no participant/coach is identifiable |
| Legal/compliance check | Compliance Lead | Confirms consent envelope and FCA boundary |
| Final sign-off | Head of Impact | Signs the Evidence Room record set as published |
| Publication | Ops | Posts, versions, and archives prior versions |
Any step skipped = the report is held. No exceptions, including for client deadlines. We tell the client.
Correction policy
A "material" correction is any of:
- A published outcome number changing by more than 5% in either direction.
- A published SROI ratio changing by more than 0.2 in either direction.
- A confidence factor changing band (file 05).
- A suppression decision being reversed (we said n<5; turns out it was 7) or vice versa.
- A named participant or named client appearing or disappearing.
- Any safeguarding-related claim being adjusted.
Non-material corrections (typos, layout, broken links) are fixed in place with a changelog entry, no separate announcement.
Material correction workflow (5 working days)
| Day | Action | Owner |
|---|---|---|
| 0 | Issue identified, logged in Evidence Room, draft correction prepared | Whoever spotted it → Head of Impact |
| 1 | Confirm scope: which artefacts contain the bad number; who already received them | Head of Impact + Data Lead |
| 2 | Draft corrected version with strikethrough on old number, new number, one-paragraph explanation | Head of Impact |
| 3 | Sign-off chain (as above) on the corrected version | All sign-off owners |
| 4 | Publish corrected version; update corrections log; notify every recipient of the original | Ops |
| 5 | Post-mortem note added to Evidence Room: what went wrong, what we changed in the process | Head of Impact |
The original version is not deleted — superseded with a banner and a link to the corrected version. Public-facing PDFs are re-issued; the old PDF carries a watermark "Superseded — see v1.1 dated [date]."
Correction log entry format
Correction CR-YYYY-NNNN
Date: 2026-MM-DD
Affected artefact: <Report name, version, page>
What was wrong: <one paragraph, plain English>
What was changed: <old number → new number, with units and n>
Why it happened: <root cause, no blame language>
What we changed in the process: <process / system change>
Linked Evidence Room record: ER-...
Published at memacademy.org/corrections, oldest first, never deleted.
When the client wants us to soften a number
We won't. The same numbers go to the client, the board, and the public report. We will:
- Explain the methodology more clearly.
- Add context (e.g. "n was low this quarter because the cohort was paused for restructuring").
- Show direction of travel alongside the absolute number.
- Suggest the headline number for the next quarter that the programme is genuinely on track to hit.
What we won't do is publish a number we don't believe.
When we are asked to keep a correction private
A client may ask us not to publish a correction on their named engagement. We respect the client-named version of the correction (rephrase to "a UK client in [sector]"). We do not skip the correction itself. If a client refuses any public correction at all, that becomes a board-level relationship issue (file 06).
Annual review
Once a year, Head of Impact reviews:
- Number of corrections issued, by category.
- Time-to-correction (target: median ≤5 working days; ceiling 10).
- Repeat root causes.
- Whether any process changes from prior corrections have actually held.
Findings go into the annual impact report (file 03), Section D.