Evidence Repository: Search, Review and Export

The Evidence Repository is the firm-wide evidence workspace. It aggregates four sources into one searchable surface: the decision record of every comparable (its disposition, rationale and rejection category), the timestamped captures stored in blob storage, the write-once per-comparable ledger, and the study-level milestones from the study audit log. Search, review, capture, ledger recording and packet export all live on this one page.

What the repository holds

Plan feature: This capability requires the evidence_repository plan feature (see Plan Features).

Source	What it carries
Comparable + study rows	The disposition per company (accepted, rejected, flagged, almost accepted), the recorded reason, the rejection category
Evidence snapshots	Timestamped screenshots, extracted documents and text snippets, with source URL and capture time
The evidence ledger	Write-once events per comparable — captures, reviews, overrides — detailed in the immutable ledger
The study audit log	Study-level milestones (creation, transitions, sign-off), the latest fifty per study

Searching evidence across studies

The search is global: it spans every study in the firm, and the filters narrow the candidate set before anything is matched.

• Study-level filters — financial year, jurisdiction, and TP method (the PLI the study benchmarks, read from the study parameters).
• Reviewer — matches comparables carrying a ledger event by that actor, or studies with an audit-log entry authored by them.
• Decision category — the ten standardized categories, exactly as the pills label them: Accept, Non Comparable Product, Non Comparable Function, Non Comparable Financials, Other - Cooperative, Potential RPT, High Intangible Assets, Insufficient Business Information, Manual Review Required (Flagged), Almost Accepted Companies.
• Free text — a lowercased substring match against the comparable name, the AI rationale and the qualitative rationale, plus reviewer-authored text that does not live on the comparable row: snippet text, source URLs and ledger justification notes.

Results are the matched comparables, most recently updated first, paginated at fifty per page. The filter options (years, jurisdictions, methods, reviewers, categories) are derived from the firm’s own studies, so the dropdowns only ever offer values that exist in your data.

Reviewing one comparable

Selecting a result opens the drill-down for that company: the Digital War Room gallery of captures, the ledger timeline alongside, and the study’s latest milestones below. A reviewer walking one comparable reads the wall top to bottom: the AI recommendation event (what the screen decided and why), each capture with its source and capture time, the recorded disposition, and any verification or override with its justification. Timeline dots are severity-coloured — info, success, warning, danger — so an override is visible at a glance.

Capturing evidence and recording events

Two write paths live on the page:

• Capture a snapshot — a screenshot, an extracted document or a text snippet, recorded with its type, source URL and capture time. The file upload is optional: a snippet can be stored as text alone, in which case the snapshot has no blob behind it and the hash field holds only whatever you typed in. When the capture is scoped to a comparable, the ledger records an “Evidence Captured” event automatically in the same step.
• Record a ledger event — one of the seven event types (AI Recommendation, Evidence Captured, Analyst Verified, Manager Reviewed, Manager Override, Partner Reviewed, Partner Locked). Manager Override is the only type whose note is enforced: the form disables save without justification notes, and the backend rejects an empty note (422) as the real boundary.

Exporting to the audit packet

Partner and Firm Admin (and the platform Superadmin) see the packager: generate the audit packet for a study, which assembles a ZIP of the study record, the audit log, the ledger timeline and the snapshots, stores it, lists the packets you have generated before, and hands out a presigned download URL with a five-minute lifetime. The packet record keeps a SHA-256 of the whole archive, and the manifest inside it lists a hash per captured file. Each run writes a new archive with its own timestamp, so two packets of one study are not byte-identical. The packet is the export you hand over; the repository is where its contents are curated.

Who can use it

Role	Scope
Analyst	No access — the page is not in the sidebar and direct navigation redirects to the dashboard
Manager	Search, review, capture, record ledger events
Partner / Firm Admin	Everything above, plus packet export and retention purging
Superadmin	The same surface, scoped to the studies the superadmin created

FAQ

Why can analysts not see it? The repository is the review-and-defense surface; it is role-restricted to the decision-makers who defend studies and plan-restricted to firms that hold evidence_repository.

What happens to evidence when a study is re-run? A re-analysis replaces the study’s comparable set, and with it the comparable-scoped ledger events and snapshots; study-level captures (not tied to a comparable) survive. The ledger then repopulates as the new run’s recommendations are saved.

Where do captured files physically live? In tenant-scoped blob storage under the evidence/snapshots prefix; the database keeps the key, the source and the capture time, and the packet packager pulls the bytes when it builds the ZIP.