Definition
An analysis method that breaks down data by business-meaningful dimensions such as shift, equipment, material, personnel, product, or time, to avoid overall averages masking local differences.
In detail
Stratification is a preliminary step in the seven quality tools. The core is to preserve the context when data is generated, allowing the same check sheet, Pareto chart, histogram, or control chart to be redrawn by key conditions. Stratification dimensions must be defined before data collection and kept mutually exclusive; you cannot arbitrarily cut out favorable groups after seeing results. It is used to identify where differences come from, but further verification is still needed to confirm root causes.
Example in practice
The plant-wide first-pass yield averages 95%, which appears close to the target. After stratification by product family and equipment, it is found that the main product's yield on machine 2 is only 80%, while other combinations exceed 98%. The team therefore stops discussing the plant-wide average and shifts to investigating the process conditions on machine 2.