Designing a Settlement & Reconciliation Engine

Most founders underestimate reconciliation. Internal ledger and external gateway (or bank) records must match. This note covers settlement windows, matching, mismatch handling, and dispute tracking.

Internal ledger vs external gateway records

Your journal is the source of truth. Gateways and banks send statements or APIs with their view of transactions. Reconciliation matches each external record to one or more internal entries by reference id, amount, currency, and date. Unmatched records are exceptions.

T+1 settlements

Many schemes settle next day (T+1). Design batch jobs that run after cut-off, ingest statement files or API responses, normalize to a common schema, and match against the journal. Store settlement results for audit.

Batch vs real-time reconciliation

Batch is simpler and aligns with bank cycles. Real-time reconciliation (e.g. per transaction) adds complexity but shortens the window to detect discrepancies. Start with batch; move to real-time only if required.

Mismatch handling

When no match is found or amounts differ, flag the record and route to an exception queue. Support manual resolution and re-run. Never auto-adjust the ledger without an audit trail and approval path.

Dispute tracking

Link disputes to reconciliation exceptions and to the original transaction. Keep a timeline of resolution steps. This supports compliance and investor due diligence.