Intermittent Demand

The foundational forecasting technique for intermittent demand that decomposes the time series into two separate exponential smoothing estimates: the demand interval (time between non-zero periods) and the demand size (magnitude when demand occurs). This prevents the forecast from being biased downward after zero-demand periods, a critical flaw of standard exponential smoothing. The Syntetos-Boylan Approximation is a widely adopted variant that corrects for Croston's inherent positive bias.

A two-dimensional inventory segmentation matrix that classifies items by value contribution (ABC) and demand variability (XYZ) to differentiate stocking strategies. Intermittent items typically fall into the Z-category (erratic, lumpy demand). This classification prevents planners from applying standard, continuous-demand formulas to items with fundamentally different statistical behavior. For example, a high-value, intermittent item (AZ) requires a fundamentally different buffer strategy than a low-value, stable item (CX).

A phenomenon where large demand fluctuations tend to be followed by more large fluctuations, creating temporal dependencies in variance. For intermittent demand, this means a single large spike is statistically likely to be followed by another spike rather than a return to zero. Adaptive safety stock algorithms detect these volatility regimes and automatically increase buffer sizes during turbulent periods, preventing cascading stockouts that naive models would miss.

A probabilistic prediction method that estimates specific percentiles of the future demand distribution rather than a single point forecast. For intermittent demand, this is essential because the mean is a poor representation of a distribution dominated by zeros. By forecasting the 95th or 99th percentile, planners can directly size safety stock to achieve a target service level without relying on parametric assumptions that break down with sparse, irregular data.

A computational technique that runs thousands of randomized demand-supply scenarios to empirically determine the safety stock required to achieve a target service level. For intermittent demand, this approach is superior to formula-based methods because it makes no assumptions about the underlying distribution. The simulation can model complex, real-world constraints like minimum order quantities, review periods, and supplier reliability that analytical formulas cannot capture.

The desired probability of not stocking out during a replenishment cycle, expressed as a percentage that directly drives safety stock requirements. For intermittent demand items, selecting the appropriate service level requires careful cost-benefit analysis because the holding cost of slow-moving inventory can quickly outweigh the stockout cost. Many organizations apply service differentiation, assigning lower cycle service levels to Z-class items while using alternative fulfillment strategies like drop-shipping.