You are Marivic Ocampo, the first Chief Analytics Officer of Daloy Mobility, a ride-hailing and parcel-delivery platform in Metro Manila. At four this afternoon the board will fund a single machine-learning initiative, and before then you must say which of three predictions your team will build.
The situation
Daloy coordinates about 46,000 driver-partners and completes roughly 1.9 million trips a week. It books PHP 24 billion a year, keeps a commission of about 18% and sits near break-even; its investors want profit, not growth. It holds one of the richest urban-mobility datasets in the country, and almost none of it enters a decision: dispatch runs on rules written by hand on a whiteboard.
In the last seven days, 14% of booked trips ended in a cancellation. Since January, a venture-funded rival has been quoting arrival times that riders find more reliable than Daloy's. The board will give you PHP 90 million and ninety days, about thirteen weeks, to move the operating metric you nominate by 2 points, which the finance committee values at about PHP 70 million a year. If nothing is running in production on day 90, analytics is contracted out and your team of four is dissolved.
Your readiness review leaves three candidates: forecast demand by zone and hour to position drivers before it arrives; predict the arrival time of each trip, the figure quoted to the rider at booking; or predict which booked trips will be cancelled, so dispatch can act in time. Each has champions inside the company, data in a different state, and a different path to a real decision.