Data and Evidence
Data and Evidence
Fund Analyst Intelligence is evidence-first by design.
It treats sources as production artefacts.
It treats provenance as a core data structure.
This is what enables trust.
It is also what enables scale.
Evidence captured once can be reused across cycles and reports.
Objectives
A production evidence system must ensure:
- every key claim is traceable to a source
- sources are versioned and time-stamped
- evidence is preserved across monthly cycles
- outputs are reproducible from stored artefacts
- conflicts and gaps are explicit, not hidden
Source types
Fund Analyst Intelligence supports controlled source classes.
1. User-provided artefacts
These are the primary source set for most partners.
Examples
- DDQ / ODD documents
- factsheets and pitch decks
- manager letters and periodic updates
- fee schedules, term sheets, share class details
- internal notes and committee decisions (optional)
User artefacts are stored with version metadata.
They are attached to a specific monthly cycle.
2. Approved online sources
Online sources are policy-defined.
Allow-lists are the default approach.
Examples
- manager website publications
- official announcements and notices
- regulated filings where applicable
- policy-approved third-party sources
Online sources are recorded with retrieval timestamps.
Where required, snapshots are retained for reproducibility.
Source policy
A source policy defines:
- what domains and source types are permitted
- recency requirements for critical fields
- rules for conflicting sources
- retention and archival expectations
- roles and permissions for adding sources
Source policy is a governance decision.
It should be approved by the owning team.
It should be stable across cycles.
Artefact versioning
Every artefact should be treated like production input.
Fund Analyst Intelligence records:
- artefact identifier
- file name and type
- ingestion timestamp
- cycle identifier
- source origin and metadata
- version relationship to prior artefacts
This prevents “latest file ambiguity”.
It also supports reproducibility and audit.
Claims and evidence mapping
The platform distinguishes two things:
- Fields: structured fund attributes (e.g., fees, liquidity terms).
- Claims: statements extracted from sources (e.g., “monthly dealing with 30 days notice”).
Every field and claim can be linked to evidence:
- source artefact id
- location reference (page, section, excerpt where possible)
- extraction timestamp and method
- confidence indicators
- reviewer notes if edited or overridden
This mapping is the credibility layer of the system.
It is also what enables fast follow-up.
Evidence packs
An evidence pack is a structured output.
It is produced alongside reports.
It is designed for scrutiny.
A standard evidence pack contains:
- source inventory for the cycle
- claim-to-evidence mapping for key statements
- citations and location references
- conflicts and gaps, with status
- reviewer decisions and sign-off references
Evidence packs can be delivered as a document bundle or a structured export.
They are the fastest way to answer “where did this come from”.
Handling uncertainty and gaps
Production systems must handle imperfect inputs.
Fund Analyst Intelligence uses explicit states:
- supported: evidence exists and is consistent
- conflicted: evidence exists but contradicts
- missing: required evidence is absent
- stale: evidence exists but violates recency policy
- needs verification: evidence is weak or ambiguous
These states appear in exceptions and reports.
They prevent silent assumptions.
They also guide follow-up work.
Conflict resolution rules
When sources conflict, the system does not guess.
It surfaces the conflict as an exception.
Resolution can follow a policy such as:
- prefer primary artefacts over secondary references
- prefer newer versions where provenance is clear
- require reviewer decision for high-severity fields
- record rationale and evidence links on resolution
The decision becomes part of the audit trail.
The conflict remains searchable historically.
Reproducibility
A report is reproducible if:
- the input artefacts are stored or referenced reliably
- extracted fields and claims are versioned by cycle
- validation and delta logic are deterministic
- reviewer decisions are recorded
- report generation uses the stored cycle record
Fund Analyst Intelligence is built around cycle records and snapshots.
This is what makes “re-run and reproduce” possible.
Data minimisation and controlled scope
Evidence-first does not mean collecting everything.
It means collecting what is necessary and defensible.
Partners should define:
- required artefacts for baseline and monthly cycles
- the fund fields that must be evidence-backed
- retention periods and allowed storage locations
- who can add sources and when
This keeps operations clean.
It also reduces compliance exposure.
Definition of done
Data and evidence governance is production-grade when:
- source policy is explicit and enforced
- artefacts are versioned and cycle-linked
- claims map to evidence consistently
- conflicts and gaps are explicit exceptions
- evidence packs support scrutiny without manual reconstruction
- reports are reproducible from stored cycle records
This is the foundation of trust in Fund Analyst Intelligence.