Outlier Detection and Correction
This article is a comprehensive guide to managing sales outliers. It begins by dissecting the root causes of outliers in sales order data. Then, it proceeds to critique the limitations of conventional outlier management methods. More importantly, our guide introduces our unique approach to addressing outliers: a dual solution centered on identifying and removing erroneous data while integrating business drivers into our predictive model.
Outliers in supply chain data are more than mere statistical aberrations. These disruptive elements threaten to make your demand forecasts irrelevant, ultimately hindering supply operations. Anomalies in sales data present significant challenges to supply chain planners, who routinely find themselves immersed in laborious manual data-cleaning activities.
Causes of Outliers: Errors and Business Events
Outliers in sales data are often more than meets the eye. They can be traceable to many causes, broadly falling into two principal categories: Erroneous Data Entries and Legitimate Business Events.
- Erroneous Data Entries: One of the common culprits behind outliers are simple human errors or inexact transaction processing. Even meticulously maintained transactional order data is not immune to inaccuracies and inconsistencies. For instance, consider a situation where an order entry clerk inadvertently types in a sales quantity of 10,000 units instead of the intended 100.00 units. Such a drastic error significantly distorts the demand pattern and manifests as a major outlier in your data set.
- Legitimate Business Events: A variety of business drivers, such as inventory shortages, promotional campaigns, spot deals, and price changes, can considerably influence sales data, generating peaks and valleys in demand. (I usually refer to these events or factors as demand drivers.) Inventory shortages, for example, can lead to unmet demand, which later manifests as a sales surge when the product is back in stock, creating two…