Coppel cut truck fuel 8% by assigning vehicles to regions by terrain, not habit

Retailer Coppel operates its own truck fleet in Mexico, and transportation is the sector's fastest-growing emissions source. Replacing vehicles with greener models is expensive; the operators realized the cheaper lever was to make sure each existing truck spent its time in the conditions it runs best.

They built the allocation from data: historical GPS, cargo, fuel use, gradient variability, elevation and distance between stops went into machine learning models that estimate each vehicle's fuel efficiency and CO2 emissions per region. Those estimates fed a mathematical optimization that reassigns the whole fleet across regions to minimize total emissions.

In a one-month trial of swapping ten vehicles, the approach delivered roughly 8% savings in fuel efficiency and transportation CO2 — without buying a single new truck.