Building a Reconciliation Engine for Currency Mismatch

A payment gateway settles in USD. Your ledger records the transaction in IDR at the mid-market rate. The provider's settlement arrives 48 hours later at a different rate. The amounts do not match. Your reconciliation report flags a discrepancy. Someone on the finance team spends two hours tracing a $0.47 gap that no one caused and no one can fix.

I have watched this scenario repeat across payment systems. The root cause is the assumption that reconciliation means matching exact amounts. When multiple currencies and FX rates enter the picture, exact matching breaks. You need a reconciliation engine designed for imprecision.

Why Currency Mismatch Happens

Three forces create mismatches: rate timing (authorization vs settlement rate), provider markups (FX spreads), and rounding differences (different precision levels).

Designing the Reconciliation Engine

The engine has four layers:

Normalization. Convert both sides to a common currency using a reference rate from a neutral provider.

Threshold-based matching. Define acceptable variance thresholds per currency pair and transaction size. Use both percentage (for large amounts) and absolute (for rounding artifacts).

Rate attribution. Record the implied rate for every matched transaction with variance.

Exception classification. Not all exceptions carry the same weight. FX_VARIANCE routes to a dashboard for bulk approval. SUSPICIOUS triggers an alert.

Lessons from Production

Thresholds need calibration (FX markets shift). Audit every decision (regulators will ask). Separate FX variance from genuine errors (finance teams should not investigate 30-cent gaps with the same urgency as missing $5,000 settlements).