Key takeaways
  • A Pareto chart ranks defect categories by frequency so effort goes to the few causes producing most of the loss.
  • Rank by cost or by time lost, not only by count, or you will chase the cheapest problem because it happens most.
  • A flat Pareto usually means the categories are wrong, not that every cause is equally important.
The most common Pareto mistake
Ranking by count when the losses differ wildly

Two hundred cosmetic marks and three cracked castings are not the same problem. Counting alone puts the marks first. Weighting by cost or by lost time puts effort where the money is.

Building one

  • Choose a period long enough to be representative but recent enough to still be true. A month is often right.
  • Use defect categories people can assign consistently. If two inspectors would categorise the same defect differently, the chart is measuring opinion.
  • Count occurrences, then weight by cost, scrap value or lost minutes.
  • Sort descending and plot the cumulative percentage.

Reading it

Look for where the cumulative line reaches roughly 80%. The categories to its left are the working list. If that is two categories, you have a clear target. If it takes nine, the categorisation is probably too fine or too coarse and is worth revisiting before you act.

When the chart is flat

  • Categories may be too broad, so everything lands in 'other' and 'process issue'. Split them.
  • Categories may be too narrow, splitting one real cause across five labels. Combine them.
  • You may be charting symptoms rather than causes. Several defect types can share one cause, and a cause Pareto will look different from a defect Pareto.

Stratifying

The same data cut by shift, machine, operator, supplier or product often turns a flat chart into a sharp one. A defect that appears evenly distributed overall is frequently concentrated in one machine or one incoming batch, and that stratification is the actual finding.

After the chart

A Pareto identifies where to look, not what to do. The top category then needs a root cause investigation. Skipping that step and acting on the category name produces countermeasures aimed at a label rather than a mechanism.