Decide whether to cap, exclude, investigate, or keep an outlier.
Outliers are not automatically dirty. They can be data-entry errors, unit mismatches, legitimate rare events, fraud signals, or business milestones. The cleaning decision must ask what mechanism could produce the value and what the analysis will do with it. The data-quality principle is accuracy plus relevance. If the value is impossible under the source system rules, it is a defect. If it is possible but influential, it needs labeling, robust statistics, or a sensitivity view. Blind capping makes visuals calmer while potentially deleting the most important story in the dataset.
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in