Definition
Net economic impact of driver deviations from AI-recommended fuel stops, combining fuel price differentials, route time costs, and estimated avoided delay costs into a single dollar value per deviation event.
The Measurement Gap
Fleets measure fuel economics through metrics that cannot distinguish good deviations from bad ones.
Fuel cost per mile
aggregates all fuel purchases without segmenting by whether the driver followed or deviated from recommendations.
Fuel card compliance rate
treats every deviation as a failure, punishing drivers who skip a recommended stop to avoid a 45-minute truck queue.
Driver fuel stop adherence rate
measures obedience to the algorithm rather than the economic value of that obedience.
These metrics assume the AI recommendation is always correct, that every deviation destroys value. In practice, fuel recommendation engines often lack truck-specific constraints: they don't know that Stop A has a 30-minute diesel queue at 2 PM, that the recommended station's canopy clearance won't accommodate the trailer, or that the driver's fuel card isn't accepted at that network. When drivers deviate for legitimate reasons, compliance-based metrics label them as non-compliant while the fleet actually saved money.
FDVI reverses the lens. Instead of measuring whether drivers obeyed, it measures whether obedience would have been worth it: converting each deviation into a net dollar impact that judges the algorithm's recommendations, not just the driver's behavior.
Decision Supported:
Are drivers who deviate from fuel recommendations being wasteful, or are they compensating for impractical AI recommendations? Should the fleet enforce fuel stop compliance more strictly, or invest in improving the recommendation engine's truck-specific inputs?
Formula
Plain language: For each deviation event, FDVI calculates three components: how much more (or less) the driver paid for fuel compared to the recommended stop, how much additional (or saved) time the deviation cost at the driver's per-minute rate, and an estimated value of delays the driver may have avoided by deviating. These three values are summed per event and averaged across all deviations in the measurement period. A negative average means deviations save money; positive means they cost money.
Notation:
FDVI = (1/N) × Σ(FuelCostDiff_i + TimeCostDiff_i + AvoidedCostEstimate_i)
Where:
FuelCostDiff_i = (Price_actual − Price_recommended) × Gallons_i
TimeCostDiff_i = (RouteTime_actual − RouteTime_recommended) × DriverCostPerMinute
AvoidedCostEstimate_i = Estimated delay avoided at recommended stop (conservative default: 15-minute delay × DriverCostPerMinute, applied when deviation explanation is queue avoidance or access constraint)
Variables:
| Variable | Description | Source |
|---|---|---|
| Price_actual | Per-gallon price paid at actual fuel stop | Fuel card transaction records |
| Price_recommended | Per-gallon price at AI-recommended stop at time of recommendation | Fuel management / TMS |
| Gallons_i | Gallons purchased in deviation event | Fuel card transaction records |
| RouteTime_actual | Total route time using actual fuel stop | Telematics / ELD |
| RouteTime_recommended | Estimated route time using recommended stop | Routing engine / TMS |
| DriverCostPerMinute | Loaded driver cost per minute (wages + benefits) | Payroll / fleet cost system |
| AvoidedCostEstimate | Conservative estimate of delay avoided by deviating | Default 15-min estimate; override with actual queue data when available |
| N | Total deviation events in measurement period | Fuel management system |
Sign convention: positive FuelCostDiff means the driver paid more than recommended. Positive TimeCostDiff means the deviation took longer. Positive AvoidedCostEstimate offsets costs when the driver avoided a known delay.
Proposed formula, pending validation pilots.