root-cause-pareto

Choose a unit of measure such as minutes or cost rather than raw counts, clean categories to one granularity level, normalize by exposure like machine-hours, then rank categories and drill into the top one with a sub-Pareto or 5-why analysis.

Divide each category's impact by an exposure base such as machine-hours, orders, or units produced before comparing lines or shifts. A line running three shifts will otherwise dominate any raw ranking against a single-shift line.

Count-based ranking orders categories by occurrence frequency, while cost or duration ranking weights each event by its impact. A rare but expensive failure mode can be invisible in a count ranking, so showing both side by side prevents misdirected effort.

Common causes include an oversized Other category above 15 percent, mixed granularity levels, unnormalized cross-line comparisons, and rankings built from a single short period. Checking stability across at least two comparable periods filters out anecdotal rankings.

Avoid acting when the ranking comes from one short time window, when categories mix symptoms with causes, or when the Other bucket is the tallest bar. Stabilize the categorization and confirm the ranking persists across periods before investing.