Stratified Recommendation Accuracy Divergence (SRAD)

GMI-AGR-009Agriculture and Farm OperationsIntegration-RequiredPattern O-22: Fairness Drift

Definition

The spread in recommendation accuracy (prescription effect, forecast error, detection recall) across strata of soil, climate zone, farm size, irrigation status, and production system, reported as the ratio of worst-stratum to best-stratum accuracy.

The Measurement Gap

There is no metric that tells you how recommendation accuracy varies by soil, climate, farm size, and system stratum, which means a model that performs well on the acres it was trained on is sold as performing well everywhere.

Formula

SRAD = 1 − acc_min_stratum / acc_max_stratum, with the identity of the worst stratum reported

Proposed formula, pending validation pilots.

At a Glance

Unit
Difference (0 to 1)
Trend Direction
Lower; approaching 0
Category
Agronomic Decision Validation

Decision Supported

Whether a model that performs on the acres it was trained on should be deployed, or priced, or trusted on acres that look different, and which strata need their own validation before adoption.

← Prescription-to-Application Fidelity Index (PAFI)Yield Effect Attributability Score (YEAS) →

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