You are on page 1of 21

Quantitative Capability

Assessment

Cp, Cpk, Pp Ppk


These are non dimensional constants used to describe capability
In 6 Sigma organizations they are more useful than percentage
yields
Flowserve is Six Sigma capable in only a few processes so we
tend not to use these indices. However it is not uncommon for
Six Sigma aware customers to ask use to describe our capability
using these measures.

Cp

Remember a 3 sigma process is


99% good. For many years in
high volume manufacturing this
was the goal for process
capabilities.
Once you go beyond 3 sigma
process capabilities are measured
in fractions of percentages the

numbers are valid but clumsy so


the sigma (Cp) scale is used. Cp
assumes that you have a normal
process centered halfway between
the specification limits

When you have a 3 sigma


process. (i.e. from the mean to the
spec limits = 3 sigma) Cp 1, your
yield is 99%

USL LSL
Cp
6 * sshort term

If your process is more capable,


then Cp increases. If you halve
the variation in a 3 sigma process
it becomes a 6 sigma process &
Cp = 2

Cpk

The difference between Cp and


Cpk is that Cp assumes the voice
of the process is centered half way
between the sigma limits and Cpk
uses the actual voice of the
process mean.
The bigger the difference between
Cp and Cpk the greater
opportunity there is to improve the
process capability by centering.
For some simple processes this is
valuable information as you only
need change an offset to increase
capability

Short -Term Capability Indices

USL LSL
Cp
6 * s short term
Cpk min (Cpk(USL) , Cpk(LSL) )

Cpk(USL)

(USL X)

3 * sshort term

Cpk(LSL)

(X LSL)

3 * sshort term

Cpk(USL) & Cpk(LSL)


Short -Term Capability Indices

Notice we used Cpk USL and


Cpk LSL to identify Cpk

These figures should be quoted


instead of Cp when you do not
have both upper and lower
specification limits .

USL LSL
Cp
6 * s short term
Cpk min (Cpk(USL) , Cpk(LSL) )

Cpk(USL)

(USL X)

3 * sshort term

Cpk(LSL)

(X LSL)

3 * sshort term

Cpm

You are almost never going to use


this, but for completeness..
In some processes you will not
target the center point.
Example cutting impellers you
want to cut an impeller diameter
between 196 and 200 mm
It may be cheaper to bias the target
cut towards say 199mm instead of
198mm

So in this case if we target


199mm we want to measure the
process against the target
instead of the center point

Short and Long Term Sigma

Remember from modules 1 and 2


over a long period we expect the
capability of a process to
deteriorate
Also we estimate that the
difference between the short term
capability and long term capability
will be 1.5 sigma
So if a project team achieves a
process that is 99% capable (ST)
we expect it to be 80% capable
(LT) and to create a process that is
50% good (LT) we aim for 93%
(ST)

Short Term
Sigma

6.0
5.5

5.0
4.5

4.0
3.5

3.0
2.5

2.0
1.5

Long Term
Sigma

Long term
yield

Yield
4.5
4

99.9997%

3.5
3

99.9767%

2.5
2

99.3790%

1.5
1

93%

0.5
0

69%

99.9968%
99.865%
97.72%
84%
50%

99.865%
99.379%
97.725%
93.319%
84%
69%
50%
31%
16%
7%

Short and Long Term Sigma

What does short term and long


term mean? It depends! as a
guide:
Long term is more likely
to include special causes
Long term is likely to
include mixtures of
batches, parts and include
changing personnel
Long term capability does
not get worse.

Short Term
Sigma

6.0
5.5

5.0
4.5

4.0
3.5

3.0
2.5

2.0
1.5

Long Term
Sigma

Long term
yield

Yield
4.5
4

99.9997%

3.5
3

99.9767%

2.5
2

99.3790%

1.5
1

93%

0.5
0

69%

99.9968%
99.865%
97.72%
84%
50%

99.865%
99.379%
97.725%
93.319%
84%
69%
50%
31%
16%
7%

Pk, Ppk, Ppk(usl), Ppk(lsl)


The only difference between calculating the C and P
is that you use the short term sigma level for C.. And the
long term sigma level for P.
Remember long term capability = short term capability +
1.5

Normal Capability Within Cp


Between Pp

This is another more sophisticated technique for calculating short and


long term sigma levels.
It relies on your ability to group data. For example you collect the
shipments per day. However you know that there is a pattern of
shipments during the week so you group the shipments into weeks.
You can now calculate the average for each week and the standard
deviation for each week. The overall variation is made up of two
components the variation within each week and the variation
between each week.
Next the assumption that the variation within each week corresponds
to short term and the variation between the weeks is long term. By
taking these two variations in turn you derive the standard deviation
and hence Cp, Pp etc

Normal Capability Within Cp


Between Pp

In Minitab you would


select these two
variations in turn you
derive the standard
deviation and hence Cp,
Pp etc
Overall variation = short
term variation + long term
variation
A large difference
between the Overall and
Within Capability indices
may indicate the process
is out of control.

Process Capability of Y=Shipped


LB

USL
Within
Overall

Process Data
LB
0
Target
*
USL
5
Sample Mean 4.81481
Sample N
27
StDev(Within) 2.18221
StDev(Overall) 3.77818

Potential (Within) Capability


Cp
*
CPL
*
CPU 0.03
Cpk 0.03
Overall Capability
Pp
*
PPL
*
PPU 0.02
Ppk 0.02
Cpm
*

-4
O bserved Performance
PPM < LB
0.00
PPM > USL 518518.52
PPM Total
518518.52

Exp. Within Performance


PPM < LB
*
PPM > USL 466185.90
PPM Total
466185.90

12

Exp. Overall Performance


PPM < LB
*
PPM > USL 480453.93
PPM Total
480453.93

PPM parts per million defects

The bottom of the capability


diagram shows predicted ppm
defect rates. The figures are not
measured but calculated assuming
that the VoP will be normal and
using the observed average,
standard deviation, sample size
and the spec limits.
Information is presented as
number of ppm exceeding each
spec limit.
You will probably wish to simplify
into percentages

Process Capability of Y=Shipped


LB

USL
Within
Overall

Process Data
LB
0
Target
*
USL
5
Sample Mean
4.81481
Sample N
27
StDev(Within)
2.18221
StDev(O verall) 3.77818

Potential (Within) Capability


Cp
*
CPL
*
CPU 0.03
Cpk 0.03
O verall Capability
Pp
PPL
PPU
Ppk
Cpm

-4
O bserved Performance
PPM < LB
0.00
PPM > USL 518518.52
PPM Total
518518.52

Exp. Within Performance


PPM < LB
*
PPM > USL 466185.90
PPM Total
466185.90

12

Exp. O verall Performance


PPM < LB
*
PPM > USL 480453.93
PPM Total
480453.93

*
*
0.02
0.02
*

What to share with Champions and


teams?

Unless you are confident that your


Champion understands CPk, Within
variation and parts per million, please edit
the graph and delete that information
Also please make the title legible
It is often best to print the pictures for
sharing with non Minitab users rather than
asking them to look at your screen

Process Capability of Lead Time


Lower Specification Limit

Process Capability of Lead Time


LSL
Within
Overall

Process Data
LSL
200
Target
*
USL
*
Sample Mean
226.976
Sample N
21
StDev(Within)
3.04521
StDev(Overall) 11.1206

200

Potential (Within) Capability


Cp
*
CPL 2.95
CPU
*
Cpk 2.95
Overall Capability
Pp
PPL
PPU
Ppk
Cpm

200
Observed Performance
PPM < LSL 0.00
PPM > USL
*
PPM Total
0.00

210

Exp. Within Performance


PPM < LSL 0.00
PPM > USL
*
PPM Total
0.00

220

230

240

Exp. Overall Performance


PPM < LSL 7637.71
PPM > USL
*
PPM Total
7637.71

250

*
0.81
*
0.81
*

210

220

230

240

Summary

Capability can be shown as a


picture
For GB start with observed
capability
In the long term capability gets
worse

To predict the long term


capability You could
describe the capability
using - % good, parts per
million, sigma level, Cp,
Pp
The one you use will
depend on your audience

Exercise

Over the telephone you are told


that lead time is a problem,
customers want lead times less
than 25 days and here are the
lead times of the last 20 orders

29,15,21,16,30,25,20,28,
21,22,28,30,24,23,45,25,
42,23,27,19
What is the capability?
How do you interpret the EDA
output?
Create a PowerPoint slide
which you would use to explain
the capability to your
champion?

Wrong Answer
Process Capability of C1
LB

USL
Within
Overall

Process Data
LB
0
Target
*
USL
25
Sample Mean
25.65
Sample N
20
StDev(Within)
7.93206
StDev(O verall) 7.57008

Potential (Within) Capability


Cp
*
CPL
*
CPU -0.03
Cpk
-0.03
O verall Capability
Pp
PPL
PPU
Ppk
Cpm

0
O bserved Performance
PPM < LB
0.00
PPM > USL 400000.00
PPM Total
400000.00

10

Exp. Within Performance


PPM < LB
*
PPM > USL 532655.13
PPM Total
532655.13

20

30

40

Exp. O verall Performance


PPM < LB
*
PPM > USL 534212.87
PPM Total
534212.87

*
*
-0.03
-0.03
*

Better Answer

Descriptive Statistics for: C1


Anderson-Darling Normality Test
A-Squared:
0.7439
p-value:
0.0439

10

20

30

40

50

95% Confidence I nterval for Mu

10

20

30

40

95% Confidence I nterval for Median

Mean
Std Dev
Variance
Skewness
Kurtosis
n of data

25.6500
7.4712
55.8184
1.2583
1.9857
20.0000

Minimum
1st Quartile
Median
3rd Quartile
Maximum

15.0000
21.0000
24.5000
28.7500
45.0000

95% Confidence I nterval for Mu


22.1534
29.1466
50

95% Confidence I nterval for Sigma


5.6818
10.9122
95% Confidence I nterval for Median
21.2352
28.0000

Shape is not normal


Which may mean that we are looking at
Granularity perhaps there is a reason that
there is a gap around orders taking 30-40
days
Or perhaps we have two catastrophic
failures
or perhaps there is a mixture of more than
one type of order.
Next step is to investigate these two orders
with the team.
With the two values to the right excluded
the remaining data is normal. 75% of
orders are shipped in 28 days or less and
we can expect an average lead time
between 21 and 26 days std deviation 7
days.. continued

Descriptive Statistics for orders less outliers


Anderson-Darling Normality Test
A-Squared:
0.2279
p-value:
0.7800

15

20

25

30

95% Confidence I nterval for Mu

15

20

25

95% Confidence I nterval for Median

Mean
Std Dev
Variance
Skewness
Kurtosis
n of data

23.6667
4.5244
20.4706
-0.3082
-0.6675
18.0000

Minimum
1st Quartile
Median
3rd Quartile
Maximum

15.0000
20.7500
23.5000
28.0000
30.0000

95% Confidence I nterval for Mu


21.4167
25.9166
30

95% Confidence I nterval for Sigma


3.3951
6.7828
95% Confidence I nterval for Median
21.0000
27.4820

Process Capability outliers removed


LB

USL

Process Data
LB
0
Target
*
USL
25
Sample Mean
23.6667
Sample N
18
StDev(Within)
5.00626
StDev(O verall) 4.59143

Within
Overall
Potential (Within) Capability
Cp
*
CPL
*
CPU 0.09
Cpk 0.09
O verall Capability
Pp
PPL
PPU
Ppk
Cpm

0
O bserved Performance
PPM < LB
0.00
PPM > USL 333333.33
PPM Total
333333.33

Exp.
PPM
PPM
PPM

10

Within Performance
< LB
*
> USL 394991.25
Total
394991.25

15

20

25

30

Exp. O verall Performance


PPM < LB
*
PPM > USL 385756.67
PPM Total
385756.67

35

*
*
0.10
0.10
*

Remaining system is stable and


shows long term Capability
61%

Even Better Answer

Capability outliers removed - 66% of lead times within cust expectations


LB

USL

10

15

20

25

30

Having investigated with the team


there is a reason that there is a gap
around orders taking 30-40 days as
we pull orders into this month if
possible. The two orders later than
this were delayed by the customer
Unless we have more customer
delays we can expect around 66%
of orders to be within customer
expectations
To achieve 90% or better we need
to make significant process
changes to reduce the average lead
time by around 5 days

Practice
Using data from a project you or your Green Belt are
working on create power point slides to describe the
Capability

Quantitative Capability
Assessment

You might also like