NextDemand knowledge hub

Demand Forecasting Wiki

By MLAIA Data Science Ltd. · Published 13 July 2026

The Demand Forecasting Wiki is a concise, vendor-neutral reference on forecasting and inventory planning for intermittent demand — the sporadic, zero-inflated demand patterns typical of spare parts, aftermarket components, and MRO inventories. It condenses the established operations-research literature (Croston, 1972; Syntetos & Boylan, 2005; Teunter, Syntetos & Babai, 2011, among others) into short encyclopedic articles: precise definitions, the standard formulas, and the known failure modes of common practice.

It is written for inventory planners, supply-chain and operations managers, and data practitioners who need correct working definitions — what ADI and CV² measure, why Croston's method is biased and what SBA fixes, when the classic safety-stock formula misleads — without wading through the original journal papers. Articles are maintained by MLAIA Data Science Ltd., the team behind NextDemand, and kept deliberately free of marketing.

Articles

Quick definitions

Intermittent demand
Demand that occurs sporadically, with zero demand in many time periods.
Lumpy demand
Intermittent demand whose non-zero order sizes are also highly variable.
ADI
Average inter-Demand Interval — the mean number of periods between demand occurrences; values of 1.32 or more indicate intermittency.
CV²
Squared coefficient of variation of non-zero demand sizes; values of 0.49 or more indicate high size variability.
Croston's method
The classical intermittent-demand estimator (1972): separate exponential smoothing of demand sizes and inter-demand intervals.
SBA
Syntetos–Boylan Approximation — the (1 − α/2) correction that removes Croston's positive bias.
TSB method
Teunter–Syntetos–Babai variant that smooths demand probability every period, so forecasts decay for items approaching obsolescence.
Safety stock
Buffer inventory held above expected lead-time demand to absorb demand and supply variability.
Reorder point
The inventory level that triggers replenishment: expected lead-time demand plus safety stock.
Service level
The stockout-protection target — probability of no stockout per cycle (cycle service level) or fraction of units served from stock (fill rate).
Dead stock
Inventory with no demand over an extended horizon and no realistic future use.
Excess inventory
Stock held above what forecast demand and policy buffers justify.