Car Fleet Forcasting
-
Client
Helvetic Motion AG
-
Date
sep 2025 - feb 2026
-
Company link
with cooperation of Zurich University of Applied Sciences,
Dr. Nima Riahi
Problem definition
Optimal fleet planning for car rental agencies requires highly accurate, segment-specific demand forecasts spanning multiple months. While foundational research successfully modeled demand using static seasonal factors and holidays, it overlooked a critical real-time market signal: Booking Pace.
By relying solely on historical trends, current forecasts cannot dynamically adapt to the actual, accumulating velocity of reservations or the non-linear cancellation risk associated with booking lead time. This results in forecast error, hindering the capacity for agile, profit-maximizing decisions in inventory control and dynamic pricing.
My solution
This project advances the forecasting capability by shifting from static analysis to a dynamic, pace-informed methodology.
The core solution involves engineering novel time series features: the Pacing Curve Deviation (quantifying real-time booking momentum against historical S-curves) and a Risk-Adjusted Forecast Buffer (estimated via Survival Analysis of booking lead time).
Integrating these dynamic regressors into segmented Prophet models will yield a continuous, risk-mitigated forecast, providing a significant improvement in accuracy within the tactical 30-to-90-day horizon essential for optimal fleet mix allocation and inventory profitability.
