The Supply Chain AI Newsletter for People Who Build Decisions from Evidence

India imports 88% of crude oil, 100% of potash. Forex reserves down $38 B in 90 days. Rupee at ₹95/USD. Quantifies the Three F Crisis and gives a 6-pillar short- and long-term solution playbook.

Why a 1% rare event in a calm regime equals a 50% event in a volatile regime. Bayesian volatility-adjusted resilience math with 12 recent disruptions quantified and a 90-day playbook.

How a ₹10-100 cr MSME can cut forecast errors 25%, free 30% working capital and beat enterprise competitors in 90 days — at ₹0-2 lakh, team of one. Includes 6-layer AI stack, ₹50-cr FMCG case (31× ROI), 10 quick wins.

The 2026 Iran-Israel-USA war and Strait of Hormuz closure disrupted 20% of global oil, 34% of helium, 46% of urea. Brent crude +55%, freight +50%, helium prices +200%. Every supply chain risk quantified, with 12 mitigation strategies and ROI.

Six regulations now police Scope 3 greenwashing: EU CSRD (mandatory from FY2025), CBAM (full 2026), SEC Climate Rule, EU Green Claims Directive, ISSB IFRS S2, UK Green Claims Code. Penalties reach 5% of global turnover. Full disclosure mapping and supply chain data needs.

Use the ABC × XYZ × Frequency matrix to decide when to hold, minimise or eliminate safety stock. High-value (A) high-variability (Z) low-frequency SKUs need lean make-to-order. Low-value (C) stable (X) high-frequency SKUs warrant generous safety stock. All 27 cell-level policies with formulas.

Function-by-function map: demand sensing (LSTM/TFT/TimeGPT), inventory (Bayesian + RL), transport (VRP + GNN), procurement (agentic negotiation), warehousing (vision + slotting), S&OP (LLM consensus). Each capability tied to ROI and recommended ML toolkit.

Safety stock, reorder point, EOQ, service level (cycle), fill rate, inventory turnover, days of supply, MAPE, wMAPE, RMSE, OEE, perfect order, cash-to-cash, takt time, forecast bias. Each formula with units, interpretation and a Python one-liner.

Definitions and contrasts for 12 forecast accuracy metrics: MAPE, wMAPE, sMAPE, RMSE, MAE, MASE, MAD, ME (bias), MPE, Tracking Signal, FVA, Theil's U. Each with formula, when it misleads, and recommended excellent/good/poor thresholds.

Geopolitical conflict, climate disruption, supplier concentration, cyber, regulatory (CSRD/CBAM), demand volatility, talent gaps, technology debt, financial shock, ESG/reputation. Each mapped to KPIs (OTIF, fill rate, MAPE, inventory days, Scope 3, cycle time) with impact ranges.

Updating P(stockout) as new evidence arrives. Using odds-form Bayes: posterior_odds = prior_odds × likelihood_ratio. How to set priors, calibrate likelihoods from lead-time data, and translate posterior P(stockout) into safety-stock decisions — with Python worked examples.

Modern wars are won by supply chains, not soldiers. How the Strait of Hormuz blockade, drone economics ($500 drones vs $2M missiles) and military logistics shape geopolitical outcomes — and what SC leaders should learn about resilience, dual-sourcing and route redundancy.