Forecast Bias Segmentation Index (FBSI)

GMI-RET-001Retail OperationsPilot-Ready

Definition

Ratio of signed forecast error by product segment to overall absolute forecast error, revealing whether AI demand forecasts are systematically biased toward overforecasting or underforecasting specific product categories while aggregate accuracy appears healthy.

The Measurement Gap

AI demand forecasting systems report aggregate accuracy metrics that mask systematic bias at the product segment level. A demand planning system can report 92% MAPE accuracy while simultaneously overforecasting new product launches by 40% and underforecasting core replenishment items by 15%. Inventory turnover, the standard efficiency metric, cannot distinguish between overstock caused by forecast bias and overstock caused by slow-moving product decisions. GMROI aggregates margin performance across the entire assortment without isolating the contribution of forecast-driven inventory misallocation. Forecast accuracy metrics like MAPE and WAPE treat positive and negative errors symmetrically, so a segment that is consistently 30% over and another that is consistently 30% under produce the same aggregate accuracy score as a segment that is randomly distributed around zero. The result is a forecasting system that appears healthy on every dashboard while systematically destroying value in specific product categories through invisible, directional bias.

Decision Supported:

Should the organization trust the AI forecast equally across all product segments, or do specific segments need manual adjustment, different models, or different safety stock rules?

Formula

Plain language: For each product segment (new/trending, core basics, seasonal, clearance), compute the average signed percentage forecast error (predicted minus actual, divided by actual). Compare the magnitude of signed percentage errors by segment to the overall MAPE. Segments with large signed percentage errors relative to the overall MAPE reveal systematic bias the aggregate metric hides.

Notation:

FBSI(s) = MSPE(s) / Overall_MAPE

Variables:

Variable Definition Unit
FBSI(s) Forecast Bias Segmentation Index for segment s Signed ratio
MSPE(s) Mean Signed Percentage Error: (1/N_s) x SUM((Forecast_i - Actual_i) / Actual_i) for all SKU-store-day observations in segment s; positive = overforecast, negative = underforecast Ratio (dimensionless)
Overall_MAPE (1/N) x SUM( Forecast_i - Actual_i
N_s Number of forecast-to-actual observations in segment s Count
s Product segment grouping (new launches, core replenishment, seasonal, promotional, clearance) Category

Proposed formula, pending validation pilots.

At a Glance

Unit
Ratio (-1 to 1)
Trend Direction
Lower; approaching 0
Reporting Frequency
Rolling 13-week windows
Category
Demand Intelligence and Inventory Risk
← Autonomous Reorder Variance Index (ARVI)Safety Stock Tail-Risk Exposure Score (STRES) →

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