Skip to content
ML Academy

Forecasting

Interview question

  • seniorlevel
  • 12 minsuggested time
  • 3Data scientist, Data analyst, ML engineer

Airline catering must load meals per flight. Running out of meals costs about 300 per missing meal (compensation, complaints), while an unused meal costs about 100 (hypothetical figures). The current process loads the mean forecast. What should be loaded instead, and how do you check the forecast supports it?

Tests: Translating asymmetric costs into a target quantile (newsvendor critical ratio), producing quantile forecasts or prediction intervals, and evaluating calibration and sharpness.

day 1day 26

Sign in to practise: answers are saved before model answers are shown.