ReviewDesk
← Back

ReviewDesk guide · By the Ingenious AI team · Published 14 November 2026 · Reviewed 14 November 2026

What should a review case's decision history record?

Generate a free adverse-outcome letter — no sign-up Build a free pre-use notice See pricing

The short answer. Every case's story on one page: the request as filed, the assignment journey, the evidence weighed, the reviewer's outcome and note, the communication sent — each entry timestamped, each correction appended rather than overwritten. A decision history kept that way lets you demonstrate your review programme instead of describing it. ReviewDesk keeps the history per case inside the audit trail, and its completion metrics count from those real case records.

What is the review decision history?

A decision history is the per-case record of a human review: the request, the case materials, assignment and reassignment events, escalation states, the decision, the deciding reviewer's note, and the resulting communications. In ReviewDesk the history lives on the case within the workflow's audit trail, is retained with the decision file for three years, and feeds the real-case completion metrics on paid plans.

Why is the history a separate artefact from the audit trail?

Altitude. The audit trail is the organisation's whole event stream — every workflow, every case, every configuration change. The decision history is one case's distilled story, readable by a human in two minutes: what was asked, what was considered, what was decided, what was said. Reviewers work from histories; auditors sample histories; applicants' advisers are shown histories. When the two are confused — when reading one case means querying the global log — cases get summarised from memory at exactly the moments the record was built for.

ReviewDesk keeps both: the trail as the system of record, the history as the case's view of it. The distinction disciplines what the history must contain — everything a reader needs to judge the review's quality without leaving the page. If a material fact requires a trip to another system, the history has a hole, and the hole is a design defect rather than a filing error.

What entries does a complete history contain?

The request verbatim with its arrival timestamp. The case assembly: which decision, which letter, which notice version it answers. Assignment events with names and times. Escalation state changes with their notes. The evidence added along the way — corrections submitted, verifications performed, third-party confirmations — each dated. The decision: outcome, deciding reviewer, the note in the reviewer's words. And the communications generated from that decision, with their sending logged. Read top to bottom, the history should leave no question a fair reader would ask unanswered.

Two absences speak loudest in weak histories. The missing reviewer note — outcome recorded, reasoning absent — reduces the review to an assertion. And the missing communication record leaves the organisation unable to prove the applicant was told. Both absences are prevented structurally in ReviewDesk: a case cannot close without its outcome note, and the decided state requires the reply to be generated from the record, the discipline from the inbox guide.

How should the reviewer's note be written?

As the case's centrepiece: what the person asked, what the reviewer examined, what weighed most, what was decided and why — in the reviewer's own sentences, at a length proportional to the case's difficulty. The note is simultaneously the applicant's explanation, the quality sample leadership reads, and the evidence exhibit if the case is ever examined. Three audiences, one text: that is why templates must not write it. Templates carry the letter around the note; the note itself is the human judgement the entire apparatus exists to produce, and its authenticity is detectable — applicants can tell, every time, whether a person was there.

Reviewers improve fastest by reading each other's notes. Build the history into induction and calibration: new reviewers read twenty closed histories before taking a case; monthly sampling discusses two notes in depth. The standard in the review checklist stops being a poster and becomes a corpus. Organisations that treat notes as private paperwork lose their best training asset and their best quality signal in the same decision.

How do corrections and re-decisions appear in the history?

As continuation, never as replacement. A corrected record adds its entry — what was corrected, on what evidence, by whom — and a re-run assessment adds its decision entry referencing the original. The history of a case that changed its mind is longer than one that did not, and visibly so. That visibility is protective: a file showing an error, its verification and its remedy reads as a functioning system; a file silently showing only the final correct state reads, to any experienced examiner, as a file that has been prepared.

The append-only chain discipline of the traces layer — the hash-chain guide — is the machine version of the same ethic. The human layer should honour it identically: histories are written in ink, with corrections in later ink, attributed and dated. Your future selves, reconstructing a case under pressure, will be grateful for every predecessor who resisted the tidy edit.

How do histories feed programme metrics honestly?

By counting what they contain. ReviewDesk's completion metrics are built from real case records — cases opened, decided, their times and outcomes — which is why the landing page's coverage counter reports only real answered-over-total figures and shows an empty state rather than a sample number when nothing has been routed yet. Apply the same honesty internally: overturn rates, correction shares, median days and breach causes computed from histories are management information; the same numbers assembled for a slide from memory are decoration.

The metrics' deepest use is comparative over time. Is the overturn rate drifting? Are notes shortening? Is one decision type's correction share climbing? Each trend is a question the histories can then answer case by case, because the aggregate and the evidence are the same records at two altitudes. Programmes that separate metrics from records end up managing the metrics. Keep them unified, and managing the records is managing the programme.

How long should histories be kept, and who may read them?

With the decision file: three years, per the retention model in the records guide, exported through the evidence vault when a case must be shown outside. Read access follows the case's sensitivity: the assigned reviewer and the inbox owner in the ordinary course; leadership in aggregate and in samples; the person themselves, for their own case, where their rights require it. Every exceptional read — counsel preparing a response, an auditor sampling — is itself the kind of event the audit trail exists to note.

At retention's end, histories are deleted with the file, on schedule, by the vault's horizon rather than by individual decision. A history kept forever 'just in case' is a liability wearing prudence's clothes; a history deleted on its recorded schedule is a retention policy being visibly honoured. Both halves of the discipline — keep completely, delete punctually — are what make the middle period's records believable.

Your five actions for the review decision history

  1. Keep one readable history per case: request verbatim, case assembly, assignments, escalations, evidence, decision, note, communications.
  2. Make closure structural: no outcome note, no close; no reply generated from the record, no decided state.
  3. Write the reviewer note for its three readers — the applicant, leadership sampling, and any future examiner — in the reviewer's own words.
  4. Append corrections and re-decisions with references; never let a history be silently rewritten into its final state.
  5. Compute programme metrics from the histories themselves, keep them for three years, and delete on schedule.

Worked example: the case that defended itself

A fictional applicant challenges a lender's decline twice: first through review, which upholds it with a detailed note addressing his documents; then, months later, through a formal complaint alleging his review was a rubber stamp by the original decision-maker. The history answers in one reading: the request verbatim, assignment to a reviewer outside the original chain, two evidence entries — a bureau confirmation and the applicant's own further statement — an escalation when the bureau was slow, and a four-paragraph note engaging each of his grounds before upholding.

The complaint is resolved on the history alone; nobody is interviewed, nobody reconstructs. The review itself was unremarkable — a careful person doing careful work. What made it unanswerable was that the work left its record as it went, in the ordinary course, months before anyone knew the record would be needed. That is the entire argument for decision histories: you cannot know which case will need to defend itself, so every case is kept as if it will.

Frequently asked questions

What is a decision history?

The per-case record of a human review: the request as filed, the case materials, assignment and escalation events, evidence added, the decision with the reviewer's own note, and the communications sent — each entry timestamped and kept with the decision file.

How is it different from the audit trail?

Altitude. The audit trail is the organisation-wide event stream; the decision history is one case's complete, readable story drawn from it. If judging a review requires leaving the case page to hunt through other systems, the history has a hole.

What makes a good reviewer note?

The person's grounds addressed, the evidence weighed named, the decisive factors identified, and the outcome explained in the reviewer's own sentences. It serves three readers at once: the applicant, leadership sampling quality, and any future examiner.

Can a decision history be edited?

Corrections and new decisions are appended with references, never overwritten. A history that shows an error and its remedy is more credible than one silently presenting only the final state.

Is this guide legal advice?

No. ReviewDesk is templates and workflow tooling, not legal advice. Record-keeping is described as ReviewDesk implements it; consult counsel for your retention obligations.

Keep reading: How does the 3-year evidence vault export a complete decision file? · What should you ask a software vendor before you rely on their tool for consequential decisions? · ADMT duties explained for software vendors and the businesses that deploy their tools · How do you set up your first ADMT workflow, step by step? · What makes a hosted pre-use notice page work better than a PDF or a policy link? · All ReviewDesk guides · ReviewDesk home · Try the free desk · Free letter generator · Free notice builder · Pricing

ReviewDesk is built by Ingenious AI. Templates and workflow tools — not legal advice.