Sampling transfers when the trail starts in the app

A transfer that begins as a thumb-print confirmation in a mobile banking application still needs a population definition an auditor can defend. We usually start from the core banking transfer journal filtered by channel code, then reconcile that list to the mobile middleware event store for the same period.

Disagreements between the two lists are themselves findings. Common causes include retries that create duplicate middleware events, staff-assisted transfers miscoded as mobile, and corporate bulk payments that share an SDK with the retail app.

Once the population is stable, stratification by amount band and rail (internal book transfer, PromptPay, cross-bank) keeps high-impact items visible without drowning the sample in micro-payments. Each selected item is walked from the confirmation screen metadata through authorisation to posting.

Document the channel filter logic in the working papers. When supervisors ask why a celebrity-amount transfer was omitted, you need the stratum math, not a screenshot alone.