A Pareto chart is a bar chart sorted from the largest bar to the smallest, with a line that adds up the shares from left to right. It answers one question: which few items carry most of the problem? The idea behind it is an observation that holds surprisingly often. A small number of causes, defect types, customers or products accounts for the larger part of the effect.
Build it in five steps
- Add up each defect type over the period.
- Sort from largest to smallest. The category Other always goes last.
- Calculate the share of each type and the cumulative share.
- Draw the bars and the cumulative line, which ends at 100 percent.
- Mark the types that together reach about 80 percent. Start there.
For powder coating line 2, with 1,920 parts with a defect in 20 days:
| Defect type | Parts | Share | Cumulative |
|---|---|---|---|
| Thin coating | 720 | 37.5 % | 37.5 % |
| Inclusions | 480 | 25.0 % | 62.5 % |
| Orange peel | 288 | 15.0 % | 77.5 % |
| Scratches and handling damage | 192 | 10.0 % | 87.5 % |
| Craters and pinholes | 120 | 6.3 % | 93.8 % |
| Color deviation | 72 | 3.8 % | 97.5 % |
| Other | 48 | 2.5 % | 100.0 % |
Three of seven types carry 77.5 percent of the defects.
Chart it again by cost
Counting treats every defect alike. The plant does not. On line 2, a part with thin coating is simply coated again, which costs EUR 10. A part with inclusions is sanded first, which costs EUR 20. Multiply each count by its rework cost and sort again:
| Defect type | Share of cost | EUR |
|---|---|---|
| Inclusions | 33.3 % | 9,600 |
| Thin coating | 25.0 % | 7,200 |
| Orange peel | 20.0 % | 5,760 |
| Scratches and handling damage | 10.0 % | 2,880 |
| Craters and pinholes | 8.3 % | 2,400 |
| Color deviation | 2.5 % | 720 |
| Other | 0.8 % | 240 |
The total is EUR 28,800 in four weeks. By cost, inclusions move to the front with a third of the total. In this case both charts name the same top three, so the decision is easy: thin coating and inclusions first, worth EUR 16,800 of EUR 28,800.
Rules for a useful chart
Seven to ten bars. With thirty categories nobody classifies the same way, and the chart becomes a comb.
Other stays small. If Other grows above 5 to 10 percent, the categories do not fit the process anymore. Split it.
Same period, same scale. A smaller bar after an action means nothing if the second chart covers two weeks and the first one four.
Chart effects first. A Pareto chart of defect types shows what you can see. A Pareto chart of causes that nobody has verified only counts opinions. Causes come later, with data.
When the chart is flat
Sometimes no bar stands out. Then the 80 to 20 pattern does not apply to this view of the data. Try another view: by product, by process step, by customer, by machine. If every view is flat, the problem is general, and the process as a whole needs attention rather than one defect.
The 80 percent is a hint, not a law. Nobody should argue whether the limit is at 77.5 or 80 percent. The purpose is to stop working on seven things at once.
Before and after
The Pareto chart is also the simplest proof that an action worked. Twelve weeks after the baseline, the same check sheet gave this picture for 20 days:
| Defect type | Before | After |
|---|---|---|
| Thin coating | 720 | 168 |
| Inclusions | 480 | 216 |
| Orange peel | 288 | 288 |
| All seven types | 1,920 | 1,104 |
The defect rate fell from 8.0 to 4.6 percent and the rework cost from EUR 28,800 to EUR 18,000 in four weeks. Orange peel, untouched, is now the largest bar. The chart is flatter than before, and the next round starts with a new top defect. That is how the tool is meant to be used: not once for a presentation, but every month at the board.