Case study: ride app analytics
Request
The Finch Rides operations lead needs a fair comparison of city utilization.
Data
North: 12,000 occupied and 20,000 available minutes. South: 3,000 occupied and 10,000 available minutes. Trip completions can arrive late; location uploads vary by phone. These are fictional teaching inputs.
Done
Deliver a correction-capable pipeline, a city-window model, the fleet rate, and a checked operating recommendation.
Finch Rides wants to compare vehicle utilization in two cities. Its operations lead has completed-trip events, canceled requests, and driver availability intervals. Your job is to build a measure of occupied minutes divided by available driver minutes for the same city and time window. More location pings don't mean more availability; a busy phone can upload them more often.
The two cities also have different traffic volumes. Summing their occupied and available minutes gives a fleet rate, while averaging city percentages changes the weighting. Late trip-end events can correct occupied time after a window first appears. Keep those corrections idempotent and identify the version served to the dashboard. The final recommendation must use the supplied evidence, including the reason canceled requests cannot stand in for completed occupied time.
North: occupied 12,000 min / available 20,000 min = 60%
South: occupied 3,000 min / available 10,000 min = 30%
Fleet: 15,000 / 30,000 = 50%
The fleet rate is weighted by available driver minutes. A simple city average would be 45%.