Understanding Outliers in Supply Chain Data

In supply chain analytics, data-driven insights drive optimization and efficiency. However, outliers—data points that deviate significantly from the norm—can skew these insights, leading to poor decision-making if unaddressed. Let’s explore outliers in supply chain data, methods to detect them, and handling strategies.

Types of Outliers in Supply Chain Data

Outliers can be classified as univariate, multivariate, and time-series outliers. Each type impacts data interpretation in different ways and demands distinct detection techniques.

• Univariate Outliers: Deviations in a single variable.
• Multivariate Outliers: Unusual patterns across multiple variables.
• Time-Series Outliers: Spikes or drops in sequential data over time.

Univariate Outliers

These outliers can be detected by examining a single variable’s distribution, such as unusually high order quantities. To detect univariate outliers, use methods such as Standard Deviation and IQR.

Standard Deviation Method: Calculate the mean (μ) and standard deviation (σ). Then, flag any data point that falls outside [μ?3σ,μ+3σ] as an outlier.

Interquartile Range (IQR) for Detecting Extreme Values: The Interquartile Range (IQR) focuses on the middle 50% of data, using the difference between the 75th percentile (Q3) and the 25th percentile (Q1) to identify outliers. Data points below Q1?1.5×IQR or above Q3+1.5×IQR are flagged.

Winsorization reduces the effect of outliers by capping extreme values at a specified percentile (e.g., 95th percentile for upper outliers and 5th percentile for lower outliers).

Multivariate Outliers

These result from unusual combinations of variables, examples might include high shipping costs paired with low order quantities.

In this data, Customer C’s combination of low quantity and high shipping cost could signal an outlier.

Mahalanobis Distance calculates the distance between a point and the mean of a distribution, factoring in correlations between variables. It’s especially useful for detecting multivariate outliers.

Time-Series Outliers

In sequential data, time-series outliers could be spikes in demand or sudden inventory shortages. seasonal decomposition methods like ARIMA can help identify such anomalies.

Why Outliers Matter in Supply Chain Data

Outliers can distort supply chain insights, impacting demand forecasting, inventory levels, and transportation costs. Recognizing and handling outliers accurately improves data reliability and operational efficiency.

Summary of Outlier Handling Techniques

• Standard Deviation: Effective for identifying univariate outliers, especially with normally distributed data.
• IQR: Useful in detecting extreme values by setting bounds.
• Winsorization: Caps extreme values to minimize outlier impact, ideal for univariate outliers caused by data errors.
• Mahalanobis Distance: Identifies multivariate outliers in correlated data.