Evidence Ledger: The Immutable Record of the Study’s Evidence
The evidence ledger in Quartyl: the immutable, append-only record of every evidence event — the captures, the snippets, the snapshots and the source, per study.
Definition
The evidence ledger is Quartyl’s immutable, append-only record of every evidence event in the study — the captures (the screenshots, the snippets, the source extracts taken at screening time), the snapshots (the point-in-time state of the comparable’s data), and the source (the URL, the filing, the database record the evidence was taken from) — per study, in sequence, with the identity, the time, and the event. The ledger is the record the audit packet exports (the ZIP, the evidence file for the proceeding) and the accept/reject defense runs on (the per-company evidence, the source passages, the decision record). The immutability (the append-only — the events are added, never overwritten or deleted; the correction is a new event, not the edit) is the point: the ledger is the forensic record, the one the authority and the appeal read as the work as it happened, not the work as it is now.
The evidence ledger, in the study:
1. The events (the captures — the screenshot, the snippet, the source
extract; the snapshots — the comparable’s data state, the point in
time; the decisions — the accept/reject/flag, the override, the
rationale)
2. The per-event record (the event type, the comparable, the source —
the URL, the filing, the database record — the identity, the time)
3. The sequence (the events in the order they happened — the screening
sequence, the review sequence, the override sequence)
4. The immutability (the append-only — the events added, never
overwritten / deleted; the correction the new event, not the edit)
→ the audit packet’s export (the ZIP, the evidence file)
→ the accept/reject defense’s record (the per-company evidence,
the source, the decision)| The element | The content |
|---|---|
| The captures | The screenshots, the snippets, the source extracts (the evidence taken at screening time — the comparable’s business description, the filing line, the website paragraph) — the source (the URL, the filing, the database record) recorded with the capture |
| The snapshots | The point-in-time state (the comparable’s data — the financials, the profile, the ratios — at the capture’s time) — the state the screening ran on, the reproducibility anchor |
| The decisions | The accept / reject / flag (the per-company decision, the AI-assisted screening outcome, the human override, the rationale) — the decision record, the override micro-ledger |
| The per-event record | The event type, the comparable, the source, the identity, the time (the who, the what, the where, the when — the forensic fields) |
| The immutability | The append-only (the events added, never overwritten / deleted; the correction the new event, not the edit) — the forensic record, the work as it happened |
The working read (the AI-assisted screening and the audit defense): the evidence ledger is the record the AI controls guide requires (the Layer 4 — the record control, the audit trail) — the AI’s runs, the per-company evidence, the human decisions, the overrides, in sequence, with the source and the time. The ledger is the difference between the controlled workflow (the evidence captured as the work happened, the record intact) and the uncontrolled one (the evidence reconstructed at audit time, the record assembled) — the manual vs AI-assisted guide’s defensibility column runs on the ledger’s integrity. The audit packet is the ledger’s export (the ZIP, the evidence file for the proceeding) — the ledger is the record, the packet is the delivery.
Example
The study (the benchmarking, the screening): the evidence ledger records, in sequence — (1) the capture (the comparable’s business description, the website snippet, the URL, the capture date, the source); (2) the snapshot (the comparable’s financials, the profile, the ratios — the data state at the screening, the reproducibility anchor); (3) the AI assessment (the per-company draft, the source passages cited, the model’s run record); (4) the human decision (the accept/reject/flag, the reviewer’s identity, the time, the rationale); (5) the override (where the reviewer overrode the AI’s recommendation, the new event — the override, the rationale, the identity, the time — the ground-truth registry linkage). The ledger is append-only (the events added, never overwritten — the correction is the new event, the edit is not). The audit packet (the ZIP, the export) carries the ledger (the sequence, the captures, the snapshots, the decisions, the overrides) — the evidence file for the proceeding, the accept/reject defense record.
See also
FAQ
Is the evidence ledger the same as the audit trail? The evidence ledger is the evidence record (the captures, the snapshots, the source — the what was captured, the where from, the when). The audit trail is the activity record (the study’s chronology — the state changes, the reviews, the sign-offs, the who did the what, the when — the audit trails module). The two are complementary (the evidence ledger is the evidence, the audit trail is the activity) — the audit packet carries both (the evidence, the activity, the sequence) — the forensic file, the proceeding’s record.
What is the “immutability” — can the ledger be edited? The ledger is append-only — the events are added (the new capture, the new snapshot, the new decision, the new override), never overwritten or deleted. The correction is the new event (the “correction of the capture X” — the new event, the original intact) — not the edit (the original overwritten, the record broken). The immutability is the forensic property (the work as it happened, not the work as it is now) — the authority and the appeal read the sequence (the captures, the decisions, the overrides, in the order) — the edit would break the sequence, the append preserves it.
How does the evidence ledger support the AI controls? The ledger is the Layer 4 (the record control, the audit trail) of the AI controls guide — the AI’s runs (the model, the configuration, the inputs), the per-company evidence (the source passages, the captures), the human decisions (the accept/reject/flag, the identity, the time), the overrides (the rationale, the ground-truth registry linkage) — in sequence, with the source and the time. The ledger’s integrity (the append-only, the sequence, the source) is the defensibility — the manual vs AI-assisted guide’s column: the controlled workflow (the ledger intact, the evidence captured) vs the uncontrolled one (the evidence reconstructed, the record assembled). The ledger is the difference — the record, the defense.
Run the screens as a study, not a spreadsheet
Quartyl applies the method, PLI and screening steps above as a pipeline — and keeps a documented reason for every exclusion.
Related docs
Audit Packet: The Evidence Export That Defends the Study
The audit packet in Quartyl: the exportable evidence file for a proceeding — the ledger, the trail, the captures and the snapshots in one deliverable, with retention.
Read docDefending Your Accept-Reject Matrix Before the TPO (and IRS)
What the TPO and IRS actually scrutinize in the Accept-Reject matrix: the evidence file structure, the override rationale, the consistency tests, and the questions to rehearse.
Read doc