Who asked: Tomasz Adeyemi, operations lead at Larkspur Freight, a regional courier.
Why: Depot managers argue about yesterday's numbers every morning because each one counts parcels differently. Tomasz wants one daily figure per depot and a late-delivery rate everyone trusts.
The data: fleet.shipments_for(day) returns the day's shipments with depot, parcels and late. Some days have no shipments.
Done looks like: daily partitioned daily_shipments, depot_parcels (parcels per depot, 0 for a quiet depot) and late_rate (late share, 0.0 on an empty day) from 1 November, a warning when more than half are late, a job with a 1:00 schedule, and history backfilled.
Build the courier's daily metrics pipeline step by step. Each step starts from your previous code.