Digital War Room: Screenshots, Snippets and Source Evidence

The Digital War Room is the visual face of a single comparable: every piece of evidence attached to one company in one study, in one wall. It is reached from the Evidence Repository search by selecting a comparable, and it sits next to the company’s ledger timeline.

What a war room shows

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

The gallery lists every snapshot for the comparable — plus the study-level captures filed against the study with no company attached, which the same query returns so you can see the study’s own supporting documents here too. Each card carries:

• The content — the screenshot image, the document extract, or the text snippet itself.
• The type — Timestamped Screenshot, Extracted Document, or Extracted Snippet.
• The source — the source URL or filing the capture was taken from.
• The capture time — the timestamp on the record. The dialog lets you set it; left blank it defaults to the moment the capture was stored, so the wall is a dated record, not a collage.
• The file facts — size and a short SHA-256 hash of the stored blob, when the capture actually uploaded a file; a snippet stored as text alone has no blob to hash.

Snapshots fall into three types:

Type	Label	Typical use
SCREENSHOT	Timestamped Screenshot	A screen capture of the source as it stood at review time
DOCUMENT	Extracted Document	A filing or document extract relevant to the comparability call
SNIPPET	Extracted Snippet	A short text extract — a line from a filing, a disclosed figure

How captures get onto the wall

One path is automatic, and it feeds the timeline rather than the gallery. When the run’s screening pass saves comparables, each newly saved company gets a write-once “AI Recommendation” ledger event carrying the AI rationale as its recorded reason, attributed to the system. The “why” behind the disposition is on the company’s record from the first run — but it is a ledger row, not a capture.

Every snapshot in the gallery comes from the capture dialog, and nothing files captures there on its own: the pipeline does not write screenshots or document extracts into the repository. The dialog accepts a screenshot, a document or a snippet as a file upload (or as text) with its type and source, scoped to the study or to one comparable in it. A capture scoped to a comparable also writes an “Evidence Captured” ledger event automatically, so the gallery and the ledger never drift apart.

Walking one comparable’s evidence wall

A reviewer’s pass over a single comparable, in order:

1. Read the AI recommendation event. What the screen decided, and the recorded rationale — the claim to be verified.
2. Verify the captures. Does the screenshot or snippet actually show the source it claims? Is the capture time in the right period?
3. Check the disposition. The recorded decision (accepted / rejected / flagged / almost accepted) and its category, against what the wall shows.
4. Read the human events. Verification, manager review, override — each with its justification notes; an override without a justification is the gap the ledger exists to expose.
5. Note the severity dots. Warning and danger dots mark the events that changed the outcome; a wall with no human events after the AI recommendation is a company nobody looked at.

Re-run semantics

Re-running the analysis produces a fresh comparable set. The wall for replaced comparables is rebuilt: comparable-scoped snapshots and ledger events for the study are removed and re-created with the new run, while study-level captures (snapshots not tied to a comparable) survive the re-run. Treat a re-run as a new evidentiary record for that study.

FAQ

Who can capture evidence? Any viewer of the repository — Manager, Partner, Firm Admin, Superadmin (scoped to own studies). Analysts do not see the page at all.

What is the difference between a snippet and a document? A document is an extracted filing or file attached whole; a snippet is a short text extract kept as plain text. A capture that uploads a file is stored under the tenant’s evidence/snapshots prefix with a timestamped key; a snippet recorded as text has no key, only the row and its type, source and capture time.

Are captures shared across the firm? Within a study, yes — anyone in your firm who can open the repository sees the same wall for that company. Across studies, no: snapshots are scoped to one study (and, for company captures, one comparable). Across firms, no: storage keys are tenant-scoped and every query is tenant-filtered.