Dataset Data Quality

Data Quality

Issue	Detail	Treatment
Missing SLED on 11% of tasks	Expected for non-tracked categories	Keep field nullable; only use expiry KPIs where populated
1 task picked after expiry; 6 within 30 days	Very low absolute risk	Keep visible in risk views
Post-October 2025 volume drops by 97%	Likely extract boundary, not operating collapse	Default trend charts to March-October 2025
2,283 over-picks with fill rate > 1.0	Likely replenishment or transfer work	Segment out of pick-accuracy KPIs
1,253 tasks still active	Open work, not completed work	Include in backlog; exclude from cycle-time completion metrics
Cycle-time standard deviation of 786 min	Extreme long tails distort the mean	Use median and P75 for service views

Topic	Current status
Trend analysis	Reliable through October 2025
Queue segmentation	Strong; cold vs ambient split is usable
Worker benchmarking	Strong for regular pickers, weaker for very low-volume workers
Product-level analysis	Incomplete until product master is joined
Order-impact analysis	Incomplete until order/shipment joins are added

Do not present over-picks as true quality failures until task classification is joined.
Do not compare late 2025 and 2026 monthly volumes to peak 2025 months without flagging the extract boundary.
Do not use mean cycle time as the headline SLA metric.