Demand Forecast Accuracy at Decision Horizon (DFADH)

GMI-TF-028Fleet & TransportationPilot-Ready

Definition

Mean absolute percentage error between AI-predicted freight demand and actual freight volume, measured at the time horizon where fleet capacity decisions are made, quantifying whether the demand model produces forecasts accurate enough to justify fleet sizing and lease commitments.

The Measurement Gap

Most fleet operations track demand through lagging volume metrics that never connect forecast quality to capacity cost. The blind metrics include:

Fleet utilization rate: tells you what percentage of capacity was used but never explains why utilization was off. A fleet running at 85% utilization looks healthy whether the forecast was perfect or wildly wrong; you cannot distinguish planned efficiency from accidental right-sizing.

Spot market spend: captures the cost of emergency capacity but does not trace that cost back to forecast error versus operational disruption. A spike in spot spending could reflect demand model failure or a single facility closure.

Load acceptance rate: measures how often the fleet accepts tendered loads but reveals nothing about whether the fleet was sized correctly for the loads that materialized. High acceptance on low volume still masks a forecast miss.

None of these metrics answer the fundamental capacity planning question: did the demand model's prediction, at the time horizon where we made fleet commitments, match what actually happened? Without DFADH, organizations adjust fleet size reactively: chasing utilization numbers after the cost of over- or under-provisioning has already been incurred.

Decision Supported:

Is the demand model accurate enough to justify fleet sizing decisions based on it, or should capacity planners discount the AI forecast and rely on simpler heuristics?

Formula

Plain language: Take each forecast the AI made at the decision horizon (e.g., 30 days before capacity commitments). Compare it to what actually happened. Calculate the average percentage error, then subtract from 1 so the result reads as an accuracy score. A score of 1.0 means perfect forecasting; 0.70 means 30% average error.

Notation:

$$DFADH = 1 - \frac{1}{N} \sum_{i=1}^{N} \frac{|Demand_predicted_i - Demand_actual_i|}{Demand_actual_i}$$

Supplemental: Directional Bias:

$$DirectionalBias = \frac{1}{N} \sum_{i=1}^{N} \frac{Demand_predicted_i - Demand_actual_i}{Demand_actual_i}$$

Positive bias = systematic over-forecasting (fleet oversized). Negative bias = under-forecasting (spot market exposure).

Supplemental: Capacity Cost of Forecast Error (CCFE):

$$CCFE = \sum |Capacity_provisioned_i - Capacity_needed_i| \times UnitCost_i$$

CCFE translates forecast error into dollars: the cost of excess capacity sitting idle plus the premium paid for emergency capacity to cover shortfalls.

Variable Description Source
Demand_predicted_i AI-forecasted freight volume for period i at decision horizon Demand planning / TMS
Demand_actual_i Actual freight volume realized in period i TMS / load board
N Number of forecast periods evaluated Calculated
Capacity_provisioned_i Fleet capacity committed based on forecast Fleet management / procurement
Capacity_needed_i Capacity actually required by realized demand TMS / operations
UnitCost_i Cost per unit of excess or deficit capacity Financial system / spot market rates

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher; more accurate forecasts
Reporting Frequency
Monthly; per forecast horizon
Category
Fleet Economics and Lifecycle Integrity
← Defensive Braking False Positive Rate (DBFPR)EV Assignment Reality Gap (EVARG) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment