Definition
The variance amplification ratio across autonomous-reorder decision echelons: measuring how much small fluctuations in end-use demand are magnified through plant to DC to vendor to upstream supplier when AI agents make replenishment decisions in series. Originates Universal Pattern O-11 (Autonomous Variance Amplification).
The Measurement Gap
The classic bullwhip effect (small demand fluctuations at the customer end magnified into large order swings upstream) was identified by Lee et al. in 1997 as a consequence of information distortion, batching, and lead-time signal processing by humans. AI was supposed to fix this. Automated reorder agents see demand signals in real time, process them instantly, and act without the cognitive biases that amplified variance in the human era. But AI introduces a new amplification mechanism: speed and synchronization. When multiple autonomous agents at multiple echelons all react to the same signal simultaneously, without dampening logic, the collective response amplifies variance faster than humans ever could. A 5% uptick in end-use demand becomes a 15% order increase at the plant level, a 30% order increase at the DC level, and a capacity-straining spike at the vendor. No single agent is wrong: each is optimizing locally. But the collective behavior produces amplification that no plant-level metric can see. Inventory turns, fill rate, PO cycle time: all computed at the single-node level. The bullwhip is a network phenomenon, and the metrics are node metrics. The AI-amplified version is faster, steeper, and harder to detect than the human version because it happens within hours rather than weeks.
Decision Supported:
Whether auto-reorder thresholds, smoothing rules, or coordination layers are needed between echelons; whether to install dampening logic; and whether AI agents should observe upstream signals to suppress amplification.
Formula
Plain language: Measure the variance of demand at the end-use point, then the variance of the orders the plant places on its supplier, then the variance the supplier sees from all plants combined. The ratio at each step tells you how much the signal is being amplified.
Notation:
ARBA = sigma_squared(orders_to_vendor) / sigma_squared(end_use_demand)
Compute per echelon transition. ARBA > 1 indicates amplification; ARBA = 1 is pass-through; ARBA < 1 is dampening (good).
Variables:
| Variable | Definition | Unit |
|---|---|---|
| sigma_squared(orders_to_vendor) | Variance of order quantities placed on vendor over rolling window | Units squared |
| sigma_squared(end_use_demand) | Variance of end-use demand (material issues) over same rolling window | Units squared |
| Rolling window | Default 13 weeks; deseasonalized | Weeks |
Proposed formula, pending validation pilots.