Demand pattern screen

Intermittent Demand Pattern Classifier

Measure average demand interval and nonzero-size variability before choosing a forecasting method.

Demand history

Paste 6–240 equally spaced periods. Keep zeroes for periods with known availability and no demand.

Method

ADI is total periods divided by nonzero-demand periods. CV² is the population variance of nonzero demand sizes divided by their squared mean. This screen uses the traditional 1.32 ADI and 0.49 CV² classification boundaries: smooth, intermittent, erratic, or lumpy.

Interpretation

The class describes this sample; it does not automatically select a method or inventory policy. A stocked-out period can look like zero demand, so run the censoring audit first when availability was incomplete. Then compare methods with the rolling backtest.

Worked example and limits

The sample has 24 monthly periods and eight nonzero observations, so ADI is 3. A small sample, aggregation choice, new item, supersession, or changing installed base can move the class. The thresholds are a published convention, not a service-level or stocking rule.