Random sampling feels objective until a reviewer asks why you ignored the largest customer. In app-to-ledger sync validation, your sampling story matters as much as the math.
Three strata that usually earn their keep
High-value: items above a materiality threshold you define before looking. High-velocity: SKUs, SKUs’ cousins, or fee types that move often enough to hide systematic drift. High-risk: new connectors, manual overrides, or flows that failed last close.
Write the why before the draw
For each stratum, record population size, selection rule, and what would make you expand the sample. Doing this after you see results invites storytelling. Doing it before invites method.
Batch validation is not line-by-line destiny
When a nightly batch posts thousands of rows, validate the batch envelope first: file receipt time, row count, amount sum, reject log. Only then sample inside the batch. Many “sync bugs” are failed envelopes that never should have entered detail review.
Show your work in the exception board
Link each sample item to its stratum tag. When an exception appears, reviewers can see whether it is a rare high-value miss or a pattern in the velocity band. That distinction changes remediation priority.
We practice this design in the strata lab inside Ledger Sync Audit Foundations.