Defects per Million Opportunities


The Six Sigma approach aims to get the DPMO below 3.4 per million. Using the number of opportunities for a defect, rather than the number of units with a defect, allows the measure to be applied to simple and complex items. The disadvantage is that the definition of opportunity is hard to pin down. For example it could be (incorrectly, but plausibly) argued that recording an address involves many opportunities for a defect; wrong street number, wrong zip, street name spelled wrong, and so on. The opportunity should be defined in terms the customer cares about.
The total number of defects observed when processing a number of units. Suppose you observe 100 units being made. Five fail, four are reworked, and one is scrapped. The DPU is 0.05.
By conventional thinking the process yield is 99%. However the First Pass Yield (the proportion of units that go through the process the first time) is:
See Defects Per Million Opportunities
See Defects Per Unit
The proportion of units that, on average, go through a process first time without defects. It is calculated from:
See also Defects per Unit
See First Pass Yield
The probability that a unit can pass through a process without defects. It is the product of the first pass yields at each step:
RTY = y_{1} x y_{2} x y_{3} x
..y_{n}
where the y_{i} values are the yields at each step before rework.
The argument is that the perceived yield is misleading because it ignores the rework. In process rework is sometimes known as the 'hidden factory'. Consider an example:

Defects

Scrap

Rework

Units

DPU

FPY

RTY

1

10

5

5

100

0.100

0.905

0.905

2

12

5

7

95

0.126

0.881

0.797

3

8

0

8

90

0.089

0.915

0.730

4

10

5

5

90

0.111

0.895

0.653

5

5

1

4

85

0.059

0.943

0.616

Totals 
45



460

0.485


0.616

See Rolled Throughput Yield
The standard deviation is usually represented by the Greek letter sigma (s). See Sigma Level
The Six Sigma methodology measures the capability of a process using the 'sigma level'. The aim is to achieve a sigma level of at least six, which equates to less than 3.4 DPMO.
The Sigma level originates from the normal distribution. The idea is that the span between the upper and lower specification limits should be at least 12 standard deviations (six standard deviations on each side). Because it is not usually practical to set the processes mean exactly on target, and the mean of most processes is subject to drift, a 1.5 standard deviation offset is assumed in the calculation.
Other sigma levels are:
Sigma Level

Parts Per Million (PPM)

3

66811

4

6210

5

233

6

3.4

The proportion of a process output that is free of defects.
