Every capability in NeutraLedger, in detail — deterministic matching, explainable decisions, exception clustering, root-cause probability, and audit-grade reporting.
Matching Engine
Deterministic match engine — reproducible, scoring-based matching on IDs and amounts; identical inputs always yield identical output.
Semantic key normalization — transaction references are stripped of prefixes and noise for resilient fuzzy matching.
Multi-signal confidence — separate reference, amount, narration, and date scores are combined into one final confidence per decision.
Reference-mutation detection — flags truncated or altered IDs between systems instead of silently failing to match.
Amount-band optimization — value-bucketing accelerates candidate search across large files (up to 20MB).
One-to-many anchor ledgers — configure an anchor "ONE" side against "MANY" counterpart records.
Explainability & Audit
Explainable decision logic — every match shows its signal breakdown, the rule(s) fired, and the final confidence.
Immutable audit trail — a hash-chained log records candidates evaluated, rejected candidates with reasons, and the final selection.
Reproducible decisions — any cleared or flagged item can be replayed and independently verified.
Forensic export — a multi-sheet XLSX with drift heatmaps, decision signals, lineage, and the audit chain.
Exception Intelligence
Clustering engine — groups exceptions by failure pattern, gateway/provider, batch ID, settlement file, and time window.
Cluster cohesion scoring — quantifies how tightly a cluster's transactions belong together.
Root-cause probability model — ranked causes with probability scores and supporting evidence, not a single vague label.
Transaction lineage — full ingestion path and lifecycle stages that answer "where exactly did it break?"
Batch impact analysis — total value impacted, transaction counts, and financial-vs-timing classification per cluster.
Safe auto-reconciliation policy — strict thresholds: high confidence auto-clears, mid routes to a review queue, low or ambiguous goes to manual review.
Never auto-clears ambiguity — partial payments, split settlements, duplicate references, and delayed postings are detected and held.
Quarantine logic — unparseable rows are isolated to preserve the integrity of the audit trail.
Ingestion, Telemetry & Delivery
Multi-format ingestion — bank statements, ledgers, and exports normalized automatically.