Supplier Reliability Score

Measures the percentage of orders delivered by the originally committed date, not a revised date. This metric penalizes suppliers who artificially extend lead times to appear reliable.

• Calculated as On-Time In-Full (OTIF) adherence
• Excludes orders where the buyer requested a delay
• Weighted by order criticality or revenue impact
• Example: A supplier with 98% OTIF but frequent date renegotiation scores lower than one with 95% OTIF and zero revisions

Quantifies the statistical dispersion of historical lead times, rewarding suppliers with consistent, predictable delivery windows. High variability forces buyers to hold excess safety stock.

• Uses coefficient of variation (standard deviation / mean)
• Penalizes both positive and negative variance
• Segmented by SKU class and seasonality
• Example: Supplier A (14 days ± 1 day) is ranked higher than Supplier B (10 days ± 7 days) despite a longer average

Evaluates the time-to-resolution when disruptions occur. A reliable supplier proactively communicates delays and has a documented recovery process.

• Tracks mean time to acknowledge (MTTA) an alert
• Measures mean time to recovery (MTTR) from disruption
• Scores the quality of root cause analysis provided
• Example: A supplier that detects a quality hold and notifies the buyer within 2 hours scores highly, even if the shipment is delayed

Compares the supplier's internal commit dates against independent machine learning predictions. A high correlation indicates the supplier's promises are grounded in operational reality.

• Uses Pearson correlation between supplier commits and model forecasts
• Flags suppliers whose commit dates consistently diverge from predictive models
• Integrates external signals like port congestion data
• Example: A supplier whose commit dates align with a Temporal Fusion Transformer forecast is trusted more than one ignoring systemic port delays

Adjusts the composite score in real-time based on external risk factors that are outside the supplier's direct control but impact their operating environment.

• Applies geopolitical risk indices to suppliers in unstable regions
• Factors in financial health signals (e.g., credit rating changes)
• Integrates weather and climate risk for agricultural commodities
• Example: A supplier with a perfect delivery record receives a temporary score downgrade if a hurricane is forecast to hit their primary shipping port

Combines all sub-metrics into a single 0-100 scale using configurable weights that reflect organizational priorities. The normalization ensures cross-category comparability.

• Weights are adjustable by procurement category (direct vs. indirect)
• Uses min-max scaling or z-score normalization per metric
• Provides a breakdown of contributing factors for auditability
• Example: A strategic semiconductor supplier might weight lead time stability at 40%, while a commodity packaging supplier weights cost predictability higher

The statistical dispersion of a supplier's historical delivery times, representing inconsistency in performance. High variability forces buyers to hold excess safety stock.

• Measured by standard deviation or coefficient of variation
• A primary input into the reliability score's stability component
• Distinguishes a consistently late supplier from an unpredictable one

A critical Key Performance Indicator measuring a supplier's ability to deliver the correct quantity of goods by the originally committed date. It is a binary metric—a delivery either meets both conditions or it fails.

• Combines delivery precision and order accuracy
• Often used as a contractual threshold for supplier penalties
• Forms the foundational precision dimension of the reliability score

An unsupervised machine learning technique used to identify unusual shipment patterns or outlier transit events that deviate significantly from expected behavior. It flags hidden risks before they become systemic.

• Uses algorithms like Isolation Forests and autoencoders
• Detects early signals of supplier distress or port disruptions
• Feeds real-time risk flags into dynamic reliability scoring models

The phenomenon where the statistical properties of a supplier's delivery performance change over time in unforeseen ways. A previously reliable supplier may silently degrade, rendering static scores obsolete.

• Requires continuous monitoring of input data distributions
• Triggers automated model retraining and score recalibration
• Ensures the reliability score reflects current, not historical, reality

A statistical branch for analyzing the expected duration until an event occurs, such as a delivery being completed. It uniquely handles censored data—shipments still in transit at the time of analysis.

• Uses Kaplan-Meier estimators and Cox Proportional Hazards models
• Predicts the probability of on-time delivery at any future point
• Provides a probabilistic foundation for the reliability score's predictive component

A game-theoretic approach to explain the output of any machine learning model by computing the marginal contribution of each feature to a specific prediction. It reveals why a supplier received a particular score.

• Quantifies the impact of transit time, port delays, and carrier performance
• Provides auditable transparency for vendor scorecards
• Enables procurement teams to negotiate with data-backed evidence