Infra Data teaches a layered sample: mandatory inclusion for top contributors to P&L and risk, stratified picks across sleeves, then a small random remainder for surprise testing.
Document the bias on purpose
Judgmental sampling is not a flaw if you write down why items entered the sample. Financial auditing apps for investment reporting reviews can store those reasons; humans still decide the thresholds.
Quarterly twist
Rotate which secondary sleeves receive deep review so window dressing in neglected corners eventually surfaces. Keep last quarter’s sample list to prove rotation happened.