Outlier

An outlier is a measured or recorded value, such as a cycle time or inspection result, that differs significantly from the expected pattern of similar data. In manufacturing data, an outlier can indicate a genuine process problem worth investigating, an operational disruption (a tool change, an interruption) unrelated to the underlying process, or simply normal statistical variation, and distinguishing between these before acting matters.¹

Why distinguishing outlier causes matters before acting on them

Recent research on analyzing cycle-time anomalies in CNC production specifically addresses this distinction: filtering out operational disruptions before further analyzing what remains, rather than treating every unusual value the same way.

Where outliers are identified

Not on the drawing; an outlier is identified in production or inspection data (such as production actuals), not something specified by a part's requirements.

Common mistakes

Reacting to every outlier the same way, either always discarding it as noise or always treating it as a process problem; the appropriate response depends on distinguishing genuine process issues, operational disruptions, and normal variation.

Does an outlier always indicate a problem?

Not necessarily; it can reflect a genuine process issue, an operational disruption, or normal statistical variation, and telling these apart matters before acting.

Should outliers be removed from data before analysis?

Not automatically; a recurring outlier pattern can reveal a real process issue worth investigating rather than data to discard.

How does this relate to production actuals?

Outliers are often identified within production actuals data, such as an unusually long or short cycle time compared to similar parts.