Field notes · 2 April 2026
Sampling settlement batches without drowning in volume
Fintech ledgers move quickly. Sensible audit samples focus on risk strata, not random screenshots of the busiest day.
Settlement volumes at e-wallet and remittance firms can overwhelm a naive sample plan. Pulling fifty random transactions from peak Friday traffic often duplicates the same low-risk pattern while missing the thin, high-risk strata that actually fail.
We stratify by product, corridor or merchant category, and by exception flag. Within each stratum we select items that hit approval thresholds, reverse entries, and late adjustments. The goal is coverage of decision points, not photographic representation of every yen or New Taiwan dollar moved.
Document the selection rule before you open the extract. Changing the rule mid-test to chase a suspected issue contaminates independence. If something unexpected appears, open a separate inquiry sample and label it clearly in the working papers.
For each selected batch, follow the path from initiation to bank confirmation. Missing a single link—authoriser identity, dual control evidence, or bank advice—should be recorded as a control observation even when the economic outcome was correct.
Close the sample with a summary that states what was covered and what deliberately was not. That paragraph saves hours of confusion when management later challenges the breadth of fieldwork.