What is Pareto analysis and how is it done?

Why it matters

In a loss meeting, the event described most often may not be the event that loses the most time. Many short stops may catch a worker's attention while a few long breakdowns may consume more of the total production time. Pareto analysis ranks categories from largest to smallest by one chosen measure. The columns show category values, and the cumulative line shows the share accumulated up to that row. When frequency is chosen, the number of events is used; when duration is chosen, lost minutes; when cost is chosen, a defined money amount. The same data can give different priorities for different questions. The aim is not to make the same category first in every ranking, but to make visible which loss the decision focuses on.

The record unit must be explained before ranking. Two defects on one part can be two separate defect records; this is not the total number of defective parts. A piece of equipment stopping for two reasons in the same minute also does not create two separate full time losses. The event causes can be noted separately, but the total lost time must be counted once. The start, end, product and equipment scope must be the same. Accepting a record with a missing duration as zero minutes unknowingly changes the duration Pareto result. Category names must also be consistent. If the same event is split under different names it looks small; if unrelated events are collected under a broad Other heading, a category that is large but cannot be acted on is created.

A cost ranking is not the duration ranking drawn in a different color. How material spent, repair expense and production impact are valued must be defined clearly. Adding the same loss both as missed production value and, separately, as the contribution of the same production can create double counting. Here the costs are synthetic event amounts given for comparison. For them to be interpreted as a real expense reduction, which expense will really change is examined separately. Minutes and liras are not combined in a single total. If several aims are to be compared, separate tables are prepared; safety and mandatory quality conditions are not postponed by converting them to a money amount. These limits determine the authority of the ranking over the decision.

The common 80% description is not a strict law of nature. The data is not expected to show exactly 80% of the loss in exactly 20% of the categories. If the threshold is a coverage decision, the category where the accumulated share first passes the threshold is also taken into account. The largest column is not the root cause; it is the starting area for investigation. Different causes can exist within the same category, and the chosen solution may affect only some of them. A fall in the category share after an improvement is not counted as success by itself. Total production volume, observation time and actual loss amount are also compared. A change in share sometimes comes from another category growing; with the ratio, the numerator and denominator are kept. Data notice for this lesson's figures, examples, counterexamples and three short quiz questions: This dataset was generated synthetically for training; it does not represent a real business, employee or customer study.

Formula