ReviewDesk
← Back

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

How does the 3-year evidence vault export a complete decision file?

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

The short answer. The vault holds each decision file for three years and exports it whole: the notice version the person saw, the decision trace, the letters sent, the requests received and the review outcomes — a file a stranger can read without phoning you. Paid plans include the vault with full export; the sealed three-regime packs draw on the same records. The export is the moment your record-keeping either proves itself or doesn't, so test it quarterly on a random case.

What is the evidence vault and its export?

The evidence vault is ReviewDesk's retention layer on paid plans: decision files — notices, traces, letters, inbox cases and outcomes — held for a three-year horizon with full export. An export assembles one decision's complete file in order, suitable for the person themselves, counsel, or an examiner, and the three-regime evidence packs are sealed projections of the same underlying records.

What does a complete export contain?

Everything the decision touched, in the order it happened. The workflow and notice version in force. The decision record — for trace-filing plans, the sealed trace with its factors, model version and flags. The explanation letter as sent, numbered and dated. Every request the person filed, verbatim, with the case history: assignments, escalations, evidence, the reviewer's note, the outcome communication. Configuration events that bear on the decision — a threshold or notice change mid-story — included rather than left to confuse the reader.

Completeness is judged by the stranger test: a competent reader with no access to your staff reconstructs the decision's story from the export alone, in minutes. Every export format decision serves that test — chronological order, plain labels, no internal codes without a key. An export that needs a guided tour is a database dump wearing an export's name, and it will fail at the exact moment it is needed: in someone else's office, against a deadline, in an adversarial light.

When do exports actually get used?

In four recurring moments. The person's own request for their file, where the export is the answer. A complaint or claim, where counsel works from it. An examination or audit sample, where its readability is the organisation's first impression. And internal reviews — the annual risk assessment, a portfolio fairness study — where scoped exports supply the evidence base. Notice that three of the four are moments of scrutiny by readers who begin neutral-to-sceptical; the export's job in each is the same: make scrutiny easy to complete and hard to misread.

That is also why exports should be scoped by default: one person, one workflow, one period — the question's shape, not the archive's. The discipline mirrors the status link's scoping in the status guide: give the reader exactly the file their question concerns, completely, and nothing beyond it. Over-broad exports create new exposures while answering old questions.

How does the vault's three-year horizon work?

Retention as a system property: files enter the vault as the workflow runs, the horizon counts from the decision, and expiry is executed by the vault on schedule — not by an annual human decision about what to delete. That symmetry is the policy: complete retention inside the horizon, punctual deletion beyond it, both demonstrable. Where a matter is live — an open case, a pending complaint — the file's retention follows the matter; the vault's job is to make the ordinary case automatic so the exceptions get human attention.

The horizon aligns the vault with the regimes' record horizon as ReviewDesk models it — the framework in the records guide. Businesses with sector rules imposing longer retention should treat the vault export as the transfer mechanism into their own archive at the horizon, with the transfer itself recorded. What no business should do is let 'we might need it' quietly convert a three-year policy into indefinite accumulation of people's decision files.

How do the sealed packs relate to the vault?

The packs are projections; the vault is the source. The three-regime engine, described in the evidence-pack guide, projects a workflow's traces into Colorado, California and preparatory EU arrangements, sealed against the underlying records. The vault export is the per-decision, human-readable counterpart: same facts, narrative form, one person's file. A reader who doubts a pack can be taken to the vault records; a reader who doubts an export can check its trace against the chain, per the chaining guide. The two layers corroborate each other, which is what makes either of them strong evidence.

Keep the relationship clean in your own processes: never edit a vault record to improve a pack. Gaps in the pack render as 'not recorded' because the vault honestly lacks the element; the remedy is to fix the process that failed to record it, prospectively. The day someone backfills the vault to beautify a pack, every pack the organisation has ever produced becomes suspect — the single most expensive sentence in evidence management is 'we tidied the records'.

How do you test the vault before it matters?

Quarterly, cold, on a random closed case. Export it, hand it to a colleague who did not handle it, and time the reconstruction: can they state what was decided, why, what the person was told, what they asked, and what happened — from the export alone? Log the test itself. The exercise finds the same classes of gap repeatedly in young programmes: a letter version not captured, a phone outcome never logged, a configuration change invisible. Each finding is a routing fix, in the spirit of the inbox design guide — make the right capture the path of least resistance.

Test the uncomfortable direction too: export a case you know went badly — a breach, an overturn, a complaint. If the vault flatters it, something in the pipeline is editing reality, and you need to know that in a drill rather than in a disclosure. A vault that faithfully preserves your worst case is a vault a sceptical reader can trust with your best.

Who operates the vault day to day?

Nobody — and one accountable owner. Nobody, because capture is automatic: the workflow writes the file as it works, and a vault needing daily human feeding will starve. One owner, because horizons, exports and access decisions need a name: who approves an export's scope, who runs the quarterly test, who owns transfers at the horizon. In ReviewDesk's workspace model that owner is naturally the inbox owner's counterpart — sometimes the same person at small scale — with the role recorded rather than assumed.

Access discipline mirrors the trail's: exports are events, logged with their scope and requester. An export that leaves no trace of itself would be a strange citizen in a system built on traces. Keep the list of exports with the same care as the files they carry; in any serious examination, 'who has seen this file?' is among the first questions, and it deserves a recorded answer.

Your five actions for the evidence vault and its export

  1. Judge every export by the stranger test: the whole story, in order, readable in minutes without your staff.
  2. Scope exports to the question — one person, one workflow, one period — completely, and nothing beyond.
  3. Let the vault enforce the three-year horizon automatically, with live matters excepted deliberately and on record.
  4. Never edit vault records to improve a pack; fix the recording process prospectively instead.
  5. Export-test a random case quarterly — and once a year, deliberately test your worst case.

Worked example: an examination sample answered in a day

A fictional lender receives a request for a sample of twenty decision files across a quarter, with a response window measured in weeks. The vault exports run in an afternoon: twenty files, each with notice version, sealed trace, letter, case history and outcome. The reviewer's follow-up questions number three, all about criteria interpretation — none about missing records, unexplained gaps or version confusion. The examination's remaining weeks are spent on substance, which is where a well-run lender wants them spent.

The lender's own debrief notes the quieter victory: assembling the sample required no staff interviews, no spreadsheet archaeology, no apologies for the filing system. The hours the team did not spend reconstructing the past were spent on the portfolio's future — which is, in the end, the return the vault was bought for. Evidence you never have to think about until the day it introduces itself, complete, is infrastructure in the truest sense.

Frequently asked questions

What is the evidence vault?

ReviewDesk's paid-plan retention layer: each decision file — notice version, trace, letters, requests, outcomes — held for three years with full export. The sealed three-regime evidence packs are projections of the same records.

What does an export contain?

The complete file for one decision in chronological order: the notice version, the decision record or sealed trace, letters as sent, every request verbatim, the full case history and the outcome communications — readable by a stranger without assistance.

How long are files kept?

Three years from the decision under ReviewDesk's model, enforced by the vault on schedule, with live matters held while they are live. Punctual deletion at the horizon is part of the discipline, not an oversight.

Can vault records be edited to fill a gap in a pack?

No — gaps render in packs as 'not recorded', and the remedy is fixing the recording process going forward. Editing source records to beautify a projection would undermine every pack ever produced.

Is this guide legal advice?

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

Keep reading: 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? · How should the review-request form on a notice page be designed? · 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.