Defect Pareto
90 days · 342 defect units across 9 codes · the cumulative line crosses 80% at the third code, so three conversations fix most of this
Traced to batch through serials. 61 of the 110 STITCH_FAIL units carry EPCs encoded against GRN-2026-0404, batch B-2604, PO-2026-0112. That trace exists only because the encode row carries a grn_line_id. In rfid-app today S01 encodes with no inflow or PO link at all, so this join is currently impossible and the defect just looks like a supplier average.