Outlier
Definition
An outlier is a data point that falls far outside the normal pattern of a series — such as a demand spike from a one-time bulk order — and can distort forecasts, safety stocks, and performance metrics if left untreated.
In Practice
In demand planning, outliers are the enemy of clean history: one 5,000-unit tender in a SKU that normally sells 400 units a month will inflate both the average and the variability estimate, driving the forecast and safety stock up for months after the event. Standard filters flag points beyond about 3 standard deviations (or 1.5× the interquartile range) for review.
The key discipline is to correct, not delete — replace the spike with typical demand for forecasting, but keep the true history for financials, and record the cause. If the event will recur, like an annual promotion, it belongs in the forecast as a planned event rather than being scrubbed out.
Frequently Asked Questions
How do you identify outliers in demand history?
Common tests flag observations more than about three standard deviations from the mean, or beyond 1.5 times the interquartile range for skewed data. Good planning software flags candidates automatically, but a planner should confirm the cause — a data error, a one-off deal, or a genuine step change in demand.
Should outliers be removed from demand history?
Adjust rather than delete. Replace the abnormal value with representative demand in the forecasting history so models and safety stock are not distorted, but preserve the original record and note the cause. If the event repeats — promotions, seasonal tenders — model it as a planned event instead of scrubbing it.
Related Calculators
Related Terms
Demand forecasting is the practice of estimating future customer demand using historical data, statistical models, and market knowledge. It provides the quantitative foundation for demand planning, inventory targets, and capacity decisions.
Demand VariabilityThe degree to which actual demand fluctuates around its average over time, commonly measured by standard deviation or the coefficient of variation. It is the primary driver of how much safety stock an item needs.
Pyramid ForecastingPyramid forecasting is a technique that reconciles forecasts made at different levels of aggregation — item, product family, and total business — so that detailed and summary plans always roll up and force down to consistent numbers.
Pareto's LawPareto's law, also called the 80/20 rule, is the principle that a small share of causes drives most of the effect — in supply chains, typically about 20% of SKUs, customers, or suppliers account for roughly 80% of value or volume.