You are on page 1of 4

3.

4 Per Million

BY T.M. Kubiak

Perusing Process
Performance Metrics
Selecting the right measures for managing processes
Often, I have students ask for clarifica-

Percentage defective.

For example, if a unit was declared defec-

tion about the subtopic level in the 2007

Defects per unit (DPU).

tive because it met the criteria, but a mate-

Six Sigma Black Belt body of knowledge

Defects per million opportunities

rial review board later found it to be us-

(BoK) that deals with process perfor-

able, the units classification as defective

(DPMO).

mance metrics. Perhaps the students are

Parts per million (PPM).

should remain. Some organizations are

confused because of the sheer number of

Rolled throughput yield (RTY).

reluctant to embrace this position because

metrics or maybe because of some subtle

Process sigma.

it adversely affects their quality numbers.

Before examining each metric, however,

differences among the metrics.

Because a unit is usable, some organiza-

its important to define defect accurately

tions conclude it must not be defective. A

the confusion surrounding the various

to provide a foundation for interpreting

defect may or may not affect usability.

process performance metrics. Based on

these metrics in a meaningful way.

I hope this article can resolve a lot of

For example, a new car may contain


surface imperfections. By definition, the

my review of the literature, Im offering my interpretation of these metrics.

Building the foundation

car is defective because it contains one or

I suspect many differing opinions will

To use the defect-based metric effectively,

more defects, but it is still usable. Reclas-

continue to exist, however, and there may

it is important to answer:

sifying a defective unit as a nondefective

be a flurry of letters to the editor after this

What is a defect?

does nothing to help resolve the underly-

appears.

How can a defect occur?

ing cause of the defect.

Before I attempt to clear things up, its

What is a defect? I have always lived

How does a defect occur? To ad-

worthwhile to revisit each of the process

by the mantra that anything not done cor-

dress this question, many organizations

performance metrics identified in the

rectly the first time is a defect. Of course,

have compiled a list of defect families

BoK, explore the relationships among the

this means understanding what it takes

and defect types within families. Such a

metrics and look at examples of each. The

to do it right the first time. Simply put, a

list should be as complete as possible in

performance metrics include:

defect occurs during any process (for ex-

identifying all possible defect types. Also,

Example of fraction
defective / Figure 1
Distribution plot
Normal, mean = 0, standard deviation =1

0.3
Density

each defect type should be independent

manufacturing,

and mutually exclusive of others. This

chemical or

allows you to recognize the occurrence of

paperwork) when

multiple defects on any given unit.

the outcome of

0.4

0.2

0.00135

-3

52 QP www.qualityprogress.com

0.00135

0
X

clude known defect types because they hap-

not the expected

pen infrequently. In other words, if a defect

outcome. Of

family or type is known to occur, include it

course, expected

on your list. In addition, it is useful to have

outcome means

a defect family or type deemed other

the conditions of

because there may be a lack of foresight or

a defect are speci-

wisdom to define everything in advance.

Its important
3

Furthermore, avoid the temptation to ex-

the process is

fied in advance.

0.1

0.0

ample, assembly,

As you develop your list of defect


types, it is often useful to define them

to hold true to

in pairs (for example, too high and too

these definitions.

low, or too long and too short), particu-

larly when you are examining physical,


mechanical or electrical characteristics.

Example of rolled throughput yield

You might argue that a type such as too

Process FPY1
1

long or too short should simply be

Process FPY2
2

Process FPY3
3

/ Figure 2

Process FPY4
4

RTY

defined as one defect (for example, the


wrong length). This is a viable argument

Defective

Defective

Defective

Defective

and worth considering. I would suggest,


however, looking beyond the defect to the

FPY = first-pass yield RTY = rolled throughput yield

action the defect creates.


For example, if the too long defect
results in a unit requiring further trimming

containing one or more defects. Note that

unit be scrapped, the consequences of the

the ratio,

defect occurrence are different. Different

Total number of defective units


Total number of units

consequences may require identifying and


tracking different defect types. Classify-

Total number of defects


=
Total number of units
165
= 0.004125
40,000

Of course, a defective unit is any unit

and rework, and too short requires the

is known as the fraction defective.

ing the defect occurrence as two different

Consider a process in which the

defect types allows for future root cause

output is normally distributed with a

analysis.

mean of 0 and a standard deviation of 1.


Specifications are set at +/- 3. The fraction

DPMO
The DPMO metric is a measure of capability for discrete (attribute) data found by:
(Total number of defects)(1,000,000)
Total number of opportunities

The DPMO metric is important because

Defining the metrics

defective for the process is shown by the

it allows you to compare different types

When selecting meaningful metrics, con-

tail areas in Figure 1. The total fraction

of product. Developing a meaningful

sider the audience and how the metrics will

defective is the sum of the tail areas, or

DPMO metric scheme across multiple

drive action. In the following example, the

0.0027. Therefore, the percentage defec-

product lines, however, can be very time

PPM metric might be more understandable

tive is 0.27%.

consuming because it is necessary to


accurately determine the number of ways

to an organizations management that compares processes at a high level. A quality

DPU

(or opportunities) a defect can occur per

engineer who has oversight responsibility

The DPU metric is a measure of capability

unit or part. This can be an enormous

for the process, however, may consider the

for discrete (attribute) data defined by the

task, particularly when dealing with highly

DPU metric to be more actionable at the

following:

complex products and subassemblies, or

specific process level.

Total number of defects


Total number of units

As you read through each of the process performance metrics below, consider

For example, a process produces

even paperwork.
Continuing with the pencil example,
lets calculate the number of opportuni-

how they may apply within your own

40,000 pencils. Three types of defects can

ties. First, determine the number of ways

organization.

occur. The number of occurrences of each

each defect can occur on each item. For

defect type is:

this product, blurred printing occurs in

Percentage defective

Blurred printing:

The percentage defective is simply defined

Too long:

by the following equation:

Rolled ends:

Total number of defects: 165

Total number of defective units


x 100
Total number of units

36
118
11

A straightforward application of the


DPU formula provides this:

only one way (the pencil slips in the fixture), so there are 40,000 opportunities for
this defect to occur.
There are three independent places
where dimensions are checked, so there
are (3) (40,000) = 120,000 opportunities
for this dimensional defect.

Debate on performance metrics

Recently, ASQs Six Sigma Forum discussion board has hosted lively exchanges on
the confusion over DPU, PPM and DPMO vs. Sigma level. Read the topic thread at
www.asq.org/discussionBoards/thread.jspa?threadID=8655&tstart=0&forumID=37.

Rolled ends can occur at the top and


the bottom of the pencil, so there are (2)
(40,000) = 80,000 opportunities for this
defect to occur. Thus, the total number of

August 2009 QP 53

3.4 per million


we say defects, we are really referring to

opportunities for defects is:


40,000 + 120,000 + 80,000 = 240,000.
Likewise, the total number of opportunities per unit is:
1+3+2=6
Applying the DPMO formula, you can
readily determine the DPMO metric:

i=1

defectives.
PPM is also used to refer to contami-

165,000,000
= 687.5
240,000

Note that n= number of subprocesses,

nants. For example, suppose 0.23 grams

and FPYi = first-pass yield of the ith sub-

of insect parts are found in 25 kilograms

process.

of product.

Similarly, the FPY represents the percentage of units that completes a subpro-

0.23
1,000,000 = 9.2
PPM =
25,000

(Total number of defects)(1,000,000)


=
Total number of opportunities

RTY = FPYi

cess and meets quality guidelines without

Finally, in the more traditional scien-

being scrapped, rerun, retested, returned

tific context, PPM may simply refer to the

or diverted to an offline repair area. The

various ratios of components in a mixture.

FPY is calculated as:

For example, the oxygen component of

PPM

air is approximately 209,000 PPM. In this

In a typical quality setting, the PPM metric

case, the idea of defective isnt even a

usually indicates the number of times a

consideration.

Number of units entering the


process - number of defective units
Number of units entering the process

x 100

Note the FPY and RTY values are often

defective part will occur in 1 million parts

Table 1 illustrates the links among

produced. By contrast, the DPMO metric

multiple metrics, including PPM, sigma

expressed simply as the fractions or probabilities.

reflects the number of defects occurring in

level, percentage in specification and per-

1 million opportunities. It is important to

centage defective. The familiar 3.4 PPM

trated by the example given in Figure 2,

note that some authors say the PPM and

corresponds to a 6-sigma level of quality,

(p. 53) which depicts an overall process

DPMO metrics are identical. If we follow

assuming a 1.5 shift of the mean. Sigma

comprised of four subprocesses. Suppose

the definitions above, however, this would

level of a process and the 1.5 shift of the

the FPY of each subprocess is 0.95. Then,

only be true when the number of opportu-

mean will be addressed later.

the RTY is easily computed as:

The concept of the RTY is best illus-

nities for a defect per unit or part is 1.

RTY = FPYi =

RTY

Perhaps additional confusion can

i=1

(0.95)(0.95)(0.95)(0.95) = 0.81 or 81%

surround the PPM metric because of a

The RTY metric represents the percent-

laxness in the terminology applied. In the

age of units of product passing defect free

Six Sigma context, PPM is also referred to

through an entire process. It is determined

as the PPM defect rate. Similarly, 3.4 PPM

by the multiplying first-pass yields (FPY)

are relatively high, the total process yield

is often stated as 3.4 defects per million

from each subprocess of the total process

has dropped significantly. A significant

parts. In both examples, however, when

as follows:

advantage of using the RTY metric is that

Although individual subprocess yields

The relationship among several process


performance metrics / Table 1
Without sigma shift (centered)
Sigma level
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
5.50
6.00

Percentage in
specification
68.2689
86.6386
95.4500
98.7581
99.7300
99.9535
99.9937
99.9993
99.99994
99.999996
99.9999998

PPM = parts per million

54 QP www.qualityprogress.com

Percentage
defective
31.7311
13.3614
4.5500
1.2419
0.2700
0.0465
0.0063
0.0007
0.00006
0.000004
0.0000002

With 1.5 sigma shift


PPM
317311
133614
45500
12419
2700
465
63.3
6.8
0.6
0.04
0.002

Sigma level
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
5.50
6.00

Percentage in
specification
30.2328
49.8650
69.1230
84.1313
93.3189
97.7250
99.3790
99.8650
99.9767
99.9968
99.9997

Percentage
defective
69.76721
50.13499
30.87702
15.86869
6.68106
2.27504
0.62097
0.13499
0.02326
0.00317
0.00034

PPM
697672
501350
308770
158687
66811
22750
6210
1350
233
31.7
3.4

it provides a more complete view of

fore difficult to depict

the process. Subprocess yields that run

graphically.

high arent likely to garner the attention

This results in a total

necessary to drive improvement. Often,

fraction defective of

it is only when the total process yield

0.0668106, a percentage

becomes visible does real action occur.

defective of 6.68106%

Process shifts 1.5 sigma to


the right / Figure 3
Distribution plot
Normal, standard deviation = 1
0.4

and a PPM level of


66,811. From Table 1,

When theres talk of the process sigma of

we can see these values

a process, youll often hear it described

equate to a 3-sigma level.

as a 3-sigma or 4-sigma process or some-

same approach as above,

described as the sigma level of a process.

you can quickly gener-

What does this mean and how do you

ate the values shown in

interpret it in the context of Six Sigma?

Table 1.

Assume the output of a process is op-

Mean
0
1.5

0.2

If you extend the

thing similar. Sometimes youll hear it

A quick review of

USL

0.3
Density

Process sigma

LSL

Fraction
defective

0.1

Fraction
defective

Process
shift
0.0

-4

-3

-2

-1

LSL = lower specification limit


USL = upper specification limit

erating as a standard normal distribution

Table 1, along with

with a mean of 0 and standard deviation

understanding the

of 1, with an upper specification limit

1.5-sigma shift, explains why Six Sigma

than others, while some have more of an

(USL) and lower specification limit (LSL)

uses 3.4 PPM for a 6 sigma process, and

emotional impact. For example 66,811

set at +/- 3, respectively. This is depicted

not 2 PPB.

PPM may be more startling to manage-

by the blue curve in Figure 3. From basic


statistics, you know that:
P (Z 3 = USL) = 0.00135 (the area to
the right of the USL and below the blue

In addition to the use of Table 1, the

P (Z 3 = LSL) = 0.00135 (the area


to the left of the LSL and below the blue
curve in Figure 3).

ment than using a corresponding 3-sigma

sigma level associated with the 1.5-sigma

level. Regardless of which metrics you

shift can be approximated based on the

choose, each one must be based on a

PPM metric using the following formula:

clear operational definition of a defect.

curve in Figure 3).

Sigma level =
0.8406 + 29.37 - (2.221) ln (ppm).

By the way, Ive had numerous students ask me what happens to the defective products shown in Figure 2 (p. 53).

The above equation very closely ap-

I tell them they are packaged as com-

proximates the actual sigma value when

plex derivatives and other high-quality

0.0027 or percentage defective of 0.27%

the PPM is below 309,000, or when the

securities and sold on the world financial

as we determined previously.

sigma value is expected to exceed 2.

markets. Some just sit there and wonder.

This gives a total fraction defective

The underlying philosophy of Six

The subject of the 1.5-sigma shift is

Sigma, however, assumes a 1.5-sigma

highly controversial. You may or may not

shift of the mean either to the right or

accept its validity. The intent of this sec-

left over the long term. If you assume the

tion was simply to explain how the shift

shift is to the right as shown in Figure 3,

relates to PPM and associated process

the process distribution is normal with

sigma level.

a mean of 1.5 and a standard deviation


of 1. Applying basic statistics again, you

Choices abound

know that:

Clearly, there are a lot of process per-

P (Z 3 = USL) = 0.0668072 (the area

formance metrics to consider. Though

to the right of the USL and below the red

they take different forms, some of them

curve in Figure 3).

are equivalent. Selecting the appropriate

P (Z 3 = LSL) = 0.0000034 (the area

ones for your organization depends on

to the left of the LSL and below the red

your audience and how the metrics will

curve in Figure 3).

be used to drive improvement actions.

Note that the area is small and there-

Some metrics are more understandable

Others think, Good idea! QP


bibliography
Breyfogle, Forrest W. III, Implementing Six Sigma: Smarter
Solutions Using Statistical Methods, second edition, John
Wiley & Sons Inc., 2003.
Kubiak, T.M. and Donald W. Benbow, The Certified Six Sigma
Black Belt Handbook, second edition, ASQ Quality Press,
2009.
Kubiak, T.M., Expert Answers, Quality Progress, June 2008,
p. 12.
Schmidt, S.R. and R.G. Launsby, Understanding Industrial
Designed Experiments, Air Academy Press, Colorado
Springs, CO, 1997.

T.M. Kubiak is an author and


consultant in Weddington, NC. He
is a co-author of The Certified Six
Sigma Black Belt Handbook. Kubiak,
a senior member of ASQ, serves
on many ASQ boards and is a past
chair of ASQs Publication Management Board.

August 2009 QP 55

You might also like