S12 · decision value simulation

How many meals to load, when running short costs more than waste?

People, data and outputs in this scenario are simulated. Each stage is recorded when you commit; the reveal comes after.

5stages
35 minabout
0stages committed

The brief

A catering planner (simulated) at the synthetic airline orders meals for each IST-LHR departure 24 hours ahead. Today the order equals the number of bookings held at that time. Unused meals are thrown away; a passenger who finds no meal is given a voucher and a note (hypothetical costs below).

Data from flight_day and booking_curve (synthetic): IST-LHR, flights ZX114 and ZX115, 1,460 flight-days in 2022-2023 for fitting and 732 in 2024 for testing. In 2024, bookings one day before departure averaged 162.4 and passengers who flew 150.7. The show-up ratio (passengers over bookings one day out) averages 0.928 with a standard deviation of 0.013.

Stage 1 of 5

The planner asks the forecasting team what it should deliver each day.

Which quantity should the forecast provide for the meal order?

The quantile of the passenger distribution given by shortage cost over shortage cost plus waste cost.The expected number of passengers, as accurate as the model can make it, so that the order is right on average across flights.The upper limit of the 95% prediction interval, so that no passenger is ever left without a meal on any flight in the schedule.