Lean Six Sigma

Black Belt Mock Exam

Name of the Black Belt:

Please check you are taking the correct exam. This is the Black Belt Exam for
Services/Industry.
This is an open-book exam. You are allowed to use books and notes.
You are not allowed to use a calculator, telephone, tablet or computer.
Please circle your answer.
Calculation of points
Score = 10(#correct 10) / 30
Your score will be rounded to halves, with the exception of 5.5.

This material is intellectual property of

the Institute for Business and Industrial Statistics
of the University of Amsterdam (IBIS UvA)
http://www.ibisuva.nl.

1
Process Capability of Gespreksduur (min.)

USL
Within
Overall

Process Data
LSL
*
Target
*
USL
5.00000
Sample Mean
3.03327
Sample N
250
StDev(Within)
0.92945
StDev(Overall)
0.91889

Potential (Within) Capability

Cp
*
CPL
*
CPU
0.71
Cpk
0.71
CCpk 0.71

Observed Performance
% < LSL
*
% > USL 1.60
% Total
1.60
Exp. Within Performance
% < LSL
*
% > USL 1.72
% Total
1.72

Exp. Overall Performance

% < LSL
*
% > USL 1.62
% Total
1.62

Overall Capability
Pp
PPL
PPU
Ppk
Cpm

0.75

1.50

2.25

3.00

3.75

4.50

*
*
0.71
0.71
*

5.25

This is a process capability analysis of the duration of telephone calls in a call

center. What can be concluded from this analysis?
a) 1.62% of the calls is longer than the norm of 5 minutes.
b) The average duration of a call is too long.
c) The call duration does not follow a normal distribution.
d) A call center employee can handle around 12 calls per hour.

Incoming
requests

Step 1

Step 2

Step 3

Takt time: 3 min.

20 jobs / hour

Cycle time: 5 min.

Cycle time: 2 min.

Cycle time: 11 min.

Requests are processed in three steps. The cycle time of each process step is given
(the cycle time is the processing time per employee per request).
The takt rate can be assumed constant (so, no fluctuations in workload), and the
variation in processing times is marginal.
How many employees are required in total for these three steps?
a) 3.
b) 5.
c) 7.
d) 18.

3
USL

90%
90%

50%
50%

95%
95%

75%
75%

99%
99%

10

12

14

Throughput times:
50% shorter than 2.6 days
75% shorter than 5.1 days
90% shorter than 8.3 days
95% shorter than 10.6 days
99% shorter than 16.0 days

16

The graph above indicates that 50% of throughput times are shorter than 2.6 days,
75% is shorter than 5.1 days, et cetera.
By means of which techniques has the BB determined these percentages?
a) Descriptive statistics (Basic Statistics > Display Descriptive statistcs).
b) Process capability analysis (Quality tools > Capability analysis).
c) Empirical CDF (Graph > Empirical CDF).
d) Normal distribution (Calc > Probability distributions > Normal).

Why was the name Six Sigma chosen for the programme?

a) Because all processes are allowed to fluctuate in a range of 6 around the mean.
b) Because it was invented by statisticians.
c) For normally distributed variables, a distance of six times the (long-term) standard
deviation between the specification limits and the processs average ensures a
defect rate of 3.4 PPM.
d) It is symbolic for the systematic pursuit of extremely low defect rates in all
processes.

The length of a certain component has a normal distribution with mean 12.3 mm. and
standard deviation 0.21. We produce 1000 components and measure their length.
Which of the following statements is false?

a) Of the 1000 components, around 25 have a length smaller than 11.88 mm.
b) Around 500 have a length smaller than 12.3 mm.
c) Around 950 have a length smaller than 12.72 mm.
d) The average of the 1000 lengths is about equal to their median.

A BB has performed a process capability analysis with the Stat > Quality tools >
Capability Analysis (Normal) procedure in Minitab.
Suppose that the data have a strongly non-normal distribution.
Which ones of the results of the Minitab analysis are likely to be wrong?

a)
b)
c)
d)

Only the estimated StDev (both within and overall).

The estimated StDev (within and overall) and the observed performance.
Only the estimated Within (= short term) performance and the Cp and Cpk.
The estimated Within performance and the Cp and Cpk, but also the estimated
Overall (= long term) performance and the Pp and Ppk.

A BB has done a regression analysis to study the effect of an X onto his CTQ. The
result is displayed below.

Regression Analysis: CTQ versus X

The regression equation is
CTQ = 0.675651 + 0.695515 X
S = 1.39516
R-Sq = 34.8 %
R-Sq(adj) = 32.5 %
Analysis of Variance
Source
DF
SS
MS
F
P
Regression
1 29.1469 29.1469 14.9743 0.001
Error
28 54.5010
1.9465
Total
29 83.6478
Which is not true?
a)
b)
c)
d)

X has a significant effect on the CTQ.

The predicted value for X = 6 equals CTQ = 4.85.
The model gives precise predictions about the CTQ.
For fixed X, the CTQ has a standard deviation of about 1.40.

Typical questions for Black Belt for Industry

Optimization of the production of decaf
Questions 8 through 24 all relate to this case.
We consider a project at a manufacturer of coffee. The process under study is the
extraction of caffeine from raw coffee beans in order to make decaffeinated coffee
(decaf). According to food laws, the caffeine percentage of decaf should be below
0.1%. The objective of the project is to reduce the cycle time of decaffeinating a
batch of coffee, thereby increasing the capacity of the production facility (that is,
increasing the number of batches that the facility can process in a week). Processing
one additional batch per week would represent extra revenues of k15. Of course,
the final product should meet the norms as stated in the food laws.

What is a suitable operational definition of the CTQ(s) in this project?

a) CTQ-1: cycle time (measurement procedure: elapsed time registered in process log on
the computer / unit: batch / requirement: as low as possible). CTQ-2: caffeine%
(measurement procedure: chemical analysis / unit: batch / requirement: USL=0.1).
b) CTQ-1: cycle time (measurement procedure: lapsed time registered in process log on
the computer / unit: batch / requirement: as low as possible).
c) The project is successful if it succeeds in reducing the cycle time sufficiently to allow
the processing of one additional batch per week, thus creating k15 in extra revenue.
d) The CTQ is cycle time. The relevance of the CTQ is that a reduction allows an
increased capacity without additional investments (k15 for each additional batch).

9
Optimization of the production of decaf
The BB does a gauge R&R study. The set-up is as follows:
10 batches are randomly selected, and from each batch a sample is taken. The 10
samples are measured three times (on three different occasions), in a randomized
order, by a single operator.

What is a consequence of the chosen set-up?

a) This is the standard set-up of a gauge R&R study, and the results will allow a good
assessment of the measurement system.
b) With the chosen set-up, it is not possible to estimate the part-to-part spread.
c) It is not wise to select 10 batches randomly. As a consequence, the estimated GR&R%
is unreliable.
d) With the chosen set-up, it is not possible to determine the reproducibility of the
measurements.

10
Gage R&R
Source
Total Gage R&R
Repeatability
Part-To-Part
Total Variation

VarComp
0.0000294
0.0000294
0.0000605
0.0000899

%Contribution
(of VarComp)
32.74
32.74
67.26
100.00

Upper process tolerance limit = 0.1

Study Var
Source
StdDev (SD)
(6 * SD)
Total Gage R&R
0.0054252 0.0325515
Repeatability
0.0054252 0.0325515
Part-To-Part
0.0077763 0.0466581
Total Variation
0.0094818 0.0568909

%Study Var
(%SV)
57.22
57.22
82.01
100.00

%Tolerance
(SV/Toler)
45.59
45.59
65.35
79.68

Number of Distinct Categories = 2

How precise are the caffeine% measurements?

a) The 99.7% confidence margins on the true caf% are: 0.0000294.
b) The 99.7% confidence margins on the true caf% are: 0.0000882.
c) If a single person measures a single batch multiple times, the standard deviation of the
results is 0.0054252.
d) If a single person measures a single batch multiple times, the standard deviation of the
results is 0.0094818.

11
Optimization of the production of decaf
The BB does a process capability study. She has the caffeine% determined for 50
batches. The resulting analysis is on the next slide.

What should the BB conclude from the analysis?

a) The Cp cannot be calculated because the data do not have a normal distribution.
b) In the long run, around 0.00% total of all batches will be nonconforming (meaning:
caffeine% > 0.1%).
c) Over a longer period, the standard deviation of the caffeine percentages is around
0.013342.
d) This is a 0.040026 sigma process.

Process Capability of Caf%

USL
Process Data
LSL
*
Target
*
USL
0.1
Sample Mean
0.0562153
Sample N
50
StDev(Within) 0.010803
StDev(Overall) 0.013342

Within
Overall
Potential (Within) Capability
Cp
*
CPL
*
CPU 1.35
Cpk 1.35
Overall Capability
Pp
*
PPL
*
PPU 1.09
Ppk 1.09
Cpm
*

0.024 0.036 0.048 0.060 0.072 0.084 0.096

Observed Performance
% < LSL
*
% > USL 0.00
% Total 0.00

Exp. Within Performance

% < LSL
*
% > USL 0.00
% Total 0.00

Exp. Overall Performance

% < LSL
*
% > USL 0.05
% Total 0.05

12
Optimization of the production of decaf
The BB wishes to establish whether the caffeine percentages have a normal,
lognormal or Weibull distribution, or maybe even an altogether different distribution.

What is the best tool to study the distribution of the data?

a)
b)
c)
d)

The control chart.

The histogram.
The empirical CDF (cumulative distribution function).
The probability plot.

13
Optimization of the production of decaf
Suppose the BB is worried about possible contaminations in the dataset, and wants
to screen the data for outliers.

What is the best tool to identify outliers in the dataset?

a)
b)
c)
d)

The histogram.
The empirical CDF (cumulative distribution function).
The probability plot.
The 1-sample t-test.

14
Optimization of the production of decaf
The largest component of the batch cycle time, is the extraction time (ET). In normal
production, ET is set by the operators at 400 minutes per batch. To investigate the
effect of ET on the caffeine percentage, the BB does an experiment, in which ET is
varied from 200 to 700 minutes. The results are analyzed using regression analysis
(see next slide).

Is this analysis acceptable?

a)
b)
c)
d)

No: the BB should remove the outliers from the dataset.

No: the BB should try to fit a transfer function with a quadratic term.
No: 87.1% is not an acceptable value for the R-sq.
No: the value S = 0.0161376 indicates that there is too much random scatter in the
dataset.

Regression Analysis: Caf% versus ET

The regression equation is
Caf% = 0.1788 - 0.000261 ET
S = 0.0161376

R-Sq = 87.1%

R-Sq(adj) = 86.3%

Analysis of Variance

Fitted
Line Plot
Plot
SourceLine
DF

SS

1 0.0315351
0.1788 - 0.000261
CT
Caf% =Regression
0.000261
CT
Error
Total

0.16

0.0046876
0.0362227

S
R-Sq
R-Sq
Sequential Analysis of Variance
R-Sq(adj)
R-Sq(adj)

0.14
0.12

Source
Linear
Quadratic

0.10
Caf%

18
19

MS
0.0315351
0.0002604

DF
1
1

SS
0.0315351
0.0031627

0.08
0.06
0.04
0.02
0.00
200

300

400

500
CT
ET

600

700

F
121.09
35.26

F
121.09
0.0161376
0.0161376
87.1%
86.3%

P
0.000
0.000

P
0.000

15
Optimization of the production of decaf
Assume for the sake of this question that the analysis on the previous slide is correct
and acceptable.

Based on the fitted line plot on the previous slide, what would be a suitable extraction time
ET?
a)
b)
c)
d)

Caf% = 0.1%.
ET = 230 min.
ET = 300 min.
ET = 430 min.

16
Optimization of the production of decaf
In the regression analysis of Caf% versus ET (two slides back) it said:
Analysis of
Source
Regression
Error
Total

Variance
DF
SS
1 0.0315351
18 0.0046876
19 0.0362227

MS
0.0315351
0.0002604

F
121.09

P
0.000

The value for P is 0.000

What does this mean?

a) It means that the slope of the fitted line differs significantly from zero (that is, the fitted
line is not horizontal).
b) It means that the effect of Caf% is significant.
c) It means that the extraction time determines almost 100% of the caffeine percentage.
d) It means that the dataset is not large enough to prove that there is a relationship
between Caf% and ET.

17
Optimization of the production of decaf
The BB runs an experiment in which she investigates the following factors:
- Temperature
- Soaktime
- Heattime
Besides the Extraction Time ET, the Soak-time and heat-time are two other
components of the total batch cycle time. Temperature is a setting on the machine.
The CTQ in the experiment was the Caffeine%.
(see next slide).
What experimental design did the BB use?
a)
b)
c)
d)

Box-Behnken design with four center points.

25 factorial design with center points
251 fractional factorial design with center points
24 factorial design

StdOrder RunOrder PtType

Blocks
TemperatuSoaktime Heattime Caffeine%
9
1
2
1
77.5
30
60
0.11
6
2
2
1
85
40
60
0.18
14
3
0
1
77.5
40
75
0.06
5
4
2
1
70
40
60
0.17
8
5
2
1
85
40
90
0.08
12
6
2
1
77.5
50
90
0.03
7
7
2
1
70
40
90
0.08
2
8
2
1
85
30
75
0.06
11
9
2
1
77.5
30
90
0.03
3
10
2
1
70
50
75
0.12
15
11
0
1
77.5
40
75
0.07
10
12
2
1
77.5
50
60
0.15
16
13
0
1
77.5
40
75
0.06
13
14
0
1
77.5
40
75
0.08
4
15
2
1
85
50
75
0.09
1
16
2
1
70
30
75
0.12

18
Optimization of the production of decaf
The results of the first analysis are given on the next slide.
1. There are insignificant effects that should be dropped.
2. The R2 (R-sq) indicates a poor fit.
3. The analysis should have been done in uncoded units.

Which of the claims 1., 2. or 3. are valid?

a)
b)
c)
d)

Only 1.
Only 3.
Both 1. and 3.
All three 1., 2. and 3.

Response Surface Regression: Caffeine% versus

Temperature; Soaktime; Heattime
The analysis was done using coded units.
Estimated Regression Coefficients for Caffeine%
Term
Constant
Temperature
Soaktime
Heattime
Temperature*Temperature
Soaktime*Soaktime
Heattime*Heattime
Temperature*Soaktime
Temperature*Heattime
Soaktime*Heattime

S = 0.0180728
R-Sq = 94.96%

Coef
0.067500
-0.010000
0.008750
-0.048750
0.038750
-0.008750
0.021250
0.007500
-0.002500
-0.010000

PRESS = 0.0208889
R-Sq(pred) = 32.05%

SE Coef
0.008036
0.005683
0.005683
0.005683
0.008036
0.008036
0.008036
0.008036
0.008036
0.008036

T
8.399
-1.760
1.540
-8.579
4.822
-1.089
2.644
0.933
-0.311
-1.244

P
0.000
0.129
0.175
0.000
0.003
0.318
0.038
0.387
0.766
0.260

R-Sq(adj) = 87.40%

19
Optimization of the production of decaf
Consider this analysis:
Term
Constant
Temperature
Heattime
Temperature*Temperature
Heattime*Heattime
Temperature*Heattime
S = 0.0195891

Coef
0.063125
-0.010000
-0.048750
0.038750
0.021250
-0.002500

R-Sq = 89.94%

SE Coef
0.007616
0.006219
0.006219
0.008795
0.008795
0.008795

T
8.288
-1.608
-7.839
4.406
2.416
-0.284

P
0.000
0.139
0.000
0.001
0.036
0.782

R-Sq(adj) = 84.91%

Besides the constant, which terms should be included in the transfer function?
a) Only Heattime and Heattime*Heattime
b) Only Heattime, Heattime*Heattime, and Temperature*Temperature
c) Heattime, Heattime*Heattime, Temperature, and
Temperature*Temperature
d) All the terms given in the table.

20
Optimization of the production of decaf
Remember the objective of the project: batch cycle time should be reduced, while
keeping the caffeine percentage safely below 0.1%. Heat-time is a component of the
total batch cycle time.
The results of the experiment are summarized in the contour plot (next slide).

Is (heat-time = 66; temperature = 78.5) a recommendable process setting?

a) No: (heat-time = 90; temperature = 78.5) is the optimal setting.
b) No: with that setting, around 50% of the batches will have an unacceptable caffeine%.
c) No: (heat-time = 60; temperature = 70) gives maximal caffeine percentages of more
than 0.16%.
d) Yes: the given setting meets the projects objectives.

Contour Plot of Caffeine% vs Heattime; Temperature

90

0.04
85

0.08

Heattime

80

0.06

75
70

0.12
0.10

65
60
70

0.14
72

74

0.14
76
78
80
Temperature

82

84

21
Optimization of the production of decaf
The current process settings are:
Temperature = 73
Heat-time = 85
Suppose it is acceptable if at most 2.5% of the batches is above 0.1% (in caffeine%).

Based on the results of the experiment, how much can the batch cycle time be reduced?
a)
b)
c)
d)

No reduction in batch cycle time possible based on the results of the experiment.
By about 9 minutes.
By about 15 minutes.
By about 19 minutes.

22
Optimization of the production of decaf
The following principle is called Robust Design:
If a nuisance variable has an interaction effect with a control variable,
choosing a suitable setting for the control variable may reduce the effect of the
nuisance variable onto the CTQ, and thus, the CTQ is less sensitive to variability in
the nuisance variable.
The following principle is called Tolerance Design:
If much of the variation of a CTQ is caused by a nuisance variable,
then variability of the CTQ may be reduced by setting tolerance limits on the
nuisance variable, which limit the variability that is allowed in the nuisance variable.
Based on the results of this experiment, can the BB develop further improvement actions?
a) Robust design: by selecting a suitable value for heat-time, the caffeine percentages can
be made less sensitive to variation in temperature.
b) Tolerance design: by setting suitable tolerance limits to the caffeine percentages,
variability in temperature may be reduced.
c) Robust design: by selecting a suitable value for caffeine%, the temperature can be
made less sensitive to variability in heat-time.
d) None of these three ideas is feasible.

23
The BB has executed an experiment with 2 factors at 2 levels with cps. Below is the
analysis of the data. So far, the BB has only removed the interaction term.
Term
Constant
A
B

Effect
-0.2723
0.3533

Coef
42.1251
-0.1362
0.1767

SE Coef
0.2173
0.2173
0.2173

T
193.86
-0.63
0.81

Analysis of Variance for Response (coded units)

Source
DF
Seq SS
Adj SS
Main Effects
2
0.39799
0.39799
Curvature
1
0.33031
0.33031
Residual Error
8
3.02182
3.02182
Lack of Fit
1
0.02496
0.02496
Pure Error
7
2.99687
2.99687
Total
11
3.75011

P
0.000
0.548
0.440

Adj MS
0.19899
0.33031
0.37773
0.02496
0.42812

F
0.53
0.87

P
0.610
0.377

0.06

0.816

Here are 4 explanations for the analysis above.

1) Curvature has ruined the analysis.
2) The pure error is overestimated, due to irregulaties in the experiment.
3) A and B were studied in too small a range.
4) A and B have no or minor effects.
Which possible explanations make sense?
a) 4
b) 2 and 4
c) 2, 3 and 4
d) 1, 2, 3 and 4

24
Term
Constant
A
B
C
A*B
A*C
B*C

Effect
3.7509
-0.0413
0.8467
-0.3650
1.1588
-0.4096

Coef
49.7576
1.8754
-0.0207
0.2734
-0.1825
0.5794
-0.2048

SE Coef
0.2656
0.2656
0.2656
0.2656
0.2656
0.2656
0.2656

T
187.32
7.06
-0.08
1.93
-0.69
2.18
-0.77

P
0.000
0.000
0.940
0.130
0.509
0.057
0.460

Analysis of Variance for Response (coded units)

Source
Main Effects
2-Way Interactions
Residual Error
Lack of Fit
Pure Error
Total

DF
3
3
9
1
8
15

Seq SS
57.478
6.576
10.161
1.173
8.988
74.214

Adj SS
57.478
6.576
10.161
1.173
8.988

Adj MS
19.159
2.192
1.129
1.173
1.123

F
16.97
1.94

P
0.000
0.193

1.04

0.337

Having collected the results of an experiment, the BB runs the Analyze factorial
design procedure for the first time. The Minitab output is shown above. What is a
sensible next step?
a) Conduct steepest ascent.
b) Do additional runs to fit a quadratic model.
c) Eliminate the highly insignificant terms (p > 0.2).
d) Draw contour plots and find optimal settings.

Typical questions for Black Belt for Services

Reducing turnaround time for mortgage offers
Questions 25 through 40 all relate to this case.
We consider an administrative process in a bank. The process handles applications
for mortgages, and in case the application is accepted, produces an offer.
On average 40 applications enter the process each day. In the first stage, they are
handled by 4 coordinators, who check applications for completeness, and collect
additional information from the Real Property Registration (Kadaster).
Next, the coordinators send the application to one of 3 rate officers. They judge
wether the applicant qualifies for a loan, and they determine the interest rate that is
offered.
Applications involving a higher loan sum (25% of the applicactions) are then sent to
an underwriter, who assesses the risk and decides on a rejection or acceptance.
In the end, 10% of the applications are rejected, in which case a secretary writes a
rejection letter. For the accepted applications (90%), the secretary prepares the offer
and sends it to the applicant.
The company is losing business due to the rather long turnaround times (total
throughput time) of the process. If applicants do not receive an offer within a few
working days, they simply give up and apply at another bank. This results in a poor
conversion rate (= the percentage of offers accepted by applicants).

25
Revenue

Operational
cost

Conversion rate
(= % offers accepted by clients)

Turnaround
Time (TT)

Man-hours per
application

Waiting times (WTs)

in between process steps

Processing times (PTs)

for the process steps

Is the CTQ flowdown above acceptable?

a) Yes, the flowdown is good as it is.
b) No, Waiting Times (WTs) should not be connected to Man-hours per application.
c) No, Processing Times (PTs) should not be connected to Man-hours per application.
d) No, Conversion Rate should be replaced with Quality and Efficiency.

26

What is a practical and effective way to measure the PTs and WTs?
a) Travel-sheet: follow a sample of applications through the process, while employees
place time-stamps when they start or complete a task.
b) Day-in-the-life-of (multi-moment-opname): follow a number of employees over a
day. Each 15 minutes, they indicate on a check sheet what sort of task they are
performing.
c) Time-study: Have an employee perform the first task in the process for 10
applications and measure the processing times with a stopwatch. The same for the
second step in the process, and so on.
d) Process-study: Measure on what moments an application arrives in the first process
step, then in the second step, and so on.

27

DMAIC 2: Validate the measurement procedures

The BB wants to validate the measurement methods and measurement plan that she
has devised for determining the WTs and PTs.

Which of these techniques is not relevant for validating the measurement plan?
a) Gauge R&R study.
b) Do a test measurement.
c) After the data have been collected: check face validity of the results.
d) Control chart.

28
Gage R&R
%Contribution
(of VarComp)
6.20
4.79
1.41
1.41
93.80
100.00

Source
Total Gage R&R
Repeatability
Reproducibility
Measurer
Part-To-Part
Total Variation

VarComp
0.32292
0.24965
0.07327
0.07327
4.88506
5.20798

Source
Total Gage R&R
Repeatability
Reproducibility
Measurer
Part-To-Part
Total Variation

StdDev (SD)
0.56826
0.49965
0.27068
0.27068
2.21022
2.28210

Study Var
(5.15 * SD)
2.9265
2.5732
1.3940
1.3940
11.3826
11.7528

%Study Var
(%SV)
24.90
21.89
11.86
11.86
96.85
100.00

Given the results above, what are the 99% uncertainty margins (precision) for the time
measurements?
a) 0.323
b) 0.568
c) 1.46
d) 2.28

29
I Chart of PT (coordinator)
35
UCL=32.71

Individual Value

30
25
_
X=19.26

20
15
10

LCL=5.80

5
1

10

13
16
19
Observation

22

25

28

On the basis of 30 measurements, the BB makes a control chart for the PT of the
coordinators (see above). What can she conclude?
a) The processing times of the coordinators are acceptable.
b) The PTs of the coordinators are not in statistical control.
c) There is no evidence for trends or outliers (the PTs are stable).
d) All of the PTs of the coordinators are within the norms as agreed in the SLA (Service
Level Agreement).

30

The BB wishes to determine whether the WTs have a normal, lognormal or Weibull
distribution, or maybe even an altogether different distribution.

What is the best tool to study the datas distribution?

a)
b)
c)
d)

The empirical CDF (cumulative distribution function).

The histogram.
The control chart.
The probability plot.

31

The BB makes an empirical CDF graph of 50 observations of the Turnaround Time (TT);
see next slide.

What can she conclude from this graph?

a) About 4% of applications are longer than 10 work days in the process.
b) About 24% of the applications has a turnaround time of 5 days.
c) The Weibull distribution is not a good fit, as the data are not on a straight line.
d) The effect of TT is not linear.

Empirical CDF of TT
Weibull
Shape 3.575
Scale 7.100
N
50

100

Percent

80

60

40

20

0
2

6
7
8
TT (work days)

10

11

32

DMAIC 4: Identify potential influence factors.

Trying to get the TTs down to more acceptable levels, the BB considers doing an FMEA
with the employees working in the process.

Is it useful, in this project, to do an FMEA?

a) No, since the objectives of the project are not about risks.
b) No: an FMEA is used to identify failure modes (possible malfunctions) in products.
c) Yes, provided that part of the long TTs are caused by errors and mistakes.
d) Yes, as it helps to identify the bottleneck in the process.

33

DMAIC 4: Identify potential influence factors.

The BB organizes a brainstorming meeting with the employees in the process. The goal
of the meeting is to identify the causes of long TTs and to devise possible solutions.
The meeting is quite successful and results in 64 ideas.

According to the DMAIC model, what should the BB do with these ideas?
a) Participants vote about the value of these ideas, and the ideas with the most votes are
implemented (multi-voting).
b) Find evidence for the effects of these candidate causes and ideas.
c) Delegate: let the participants decide which ideas they want to implement.
d) Brainstorming sessions are not based on data. Therefore, it is not a very suitable
technique, and the BB should be very suspicious about the results.

34

Based on the measurements, the BB makes a value stream map (next page).

What are the capacities of the coordinators and the secretary?

a) Coordinators: Cap = 3 applications/day;
Secretary: Cap = 6.
b) Coordinators: Cap = 21;
Secretary: Cap = 6.
c) Coordinators: Cap = 21;
Secretary: Cap = 42.
d) Coordinators: Cap = 84;
Secretary: Cap = 42.

Applications

Client

Taktrate:40/day

(offeror
rejectionletter)

Coordinator
WT

PT:20min
#:4
Cap:

75%

Rate officer
WT

PT:15min
#:3
Cap:

WT
25%

WT

Work hours: 7 hrs / day

Secretary

Under
writer
PT:15min
#:1
Cap:

PT:10min
#:1
Cap:

35

Which resource is the bottleneck in this process? (That is, which bottleneck will have the
highest utilization?)
a) The coordinators.
b) The rate officers.
c) The underwriter.
d) The secretary.

36
The arrival rate of 40 applications per day is an average. Arrivals are not spread evenly
over a day and over a week. Also the given PTs are averages, but there is some
variability in the processing times of applications.

Will there be long waiting queues in the process (large WTs)?

a) No, as the capacity of each of the processs resources is larger than the workload.
b) No; some resources have a capacity lower than the workload, other resources have a
capacity higher than than the workload. But the average capacity is large enough to
handle the workload.
c) Yes, there will be a long waiting time especially before the bottleneck resource.
d) Yes, there will be long waiting times before all resources (all WTs are the same order
of magnitude).

37
Two-sample T for Offer vs Rejection
Offer
Rejection

N
6
6

Mean
8,754
12,063

StDev
0,820
0,846

SE Mean
0,33
0,35

Difference = mu (Offer) - mu (Rejection)

Estimate for difference: -3,309
95% CI for difference: (-4,396; -2,221)
T-Test of difference = 0 (vs not =):
T-Value = -6,88 P-Value = 0,000 DF = 9

The BB does a small study. For six accepted applications, she records how long it takes
(PT) the secretary to compose the offer. Also, for six rejected applications, she records
how long it takes to write the rejection letter.
Assuming a normal distribution for the data, what can she conclude from the analysis
above?
a) The time it takes to compose an offer is significantly shorter than the time to write a
rejection letter.
b) The sample sizes in the study are too small to make reliable conclusions.
c) The means and the variances differ significantly between the two groups.
d) No conclusions are possible, as the BB should have used the ANOVA technique.

38
The BB wants to determine the relationship between the waiting time WT in the workqueue before the coordinators, and the workload WL (the number of applications
submitted per day).
On 20 different days, she records the number of applications that have been submitted.
And she also records, for a randomly selected application, how long it spends in the
coordinators queue.
See the regression analysis on the next slide.
Is this a good fit?
a) Yes, there is no evidence for a poor fit.
b) No, a quadratic fitted line is not needed, a linear line would have been just as good.
c) No, the relationship is not significant.
d) No, there are outliers.

Polynomial Regression Analysis: WT versus WL

The regression equation is
WT = 24,64 - 1,091 WL + 0,01276 WL**2
S = 1,15449

R-Sq = 96,3%

Analysis of Variance
Source
DF
SS
Regression
2 308,243
Error
9
11,996
Total
11 320,238

R-Sq(adj) = 95,4%

MS
154,121
1,333

F
115,63

P
0,000

Fitted Line Plot

14
Sequential Analysis of Variance
Source
DF
SS
F
P
Linear
1 278,632 66,97 0,000
12
Quadratic
1
29,611 22,22 0,001
10

WT (hours)

8
6
4
2

S
R-Sq
R-Sq(adj)

0
40

45

50
55
60
65
WL (applications per day)

1,13149
94,1%
93,4%

70

39

Is the workload (WL) the most important influence factor for the waiting time (WT) before
the coordinators?

a) Yes, the small p-value indicates that this is the vital influence factor having the largest
impact.
b) The high R2-value indicates that this factor explains almost all differences in WT
observed in the regular process.
c) No, there is considerable random scatter of the observations around the fitted curve.
d) To make that conclusion, one should include other potential influence factors in the
regression analysis as well.

75

40

Suppose that the workload is WL = 62.5 applications per day.

What can you say about the waiting time in the coordinators queue?

a) In that case, 95% of all applications will spend 6 hours in the queue before the
coordinators.
b) 95% of the applications will have a waiting time below 8.3 hours.
c) There is a 50% chance that WT will be above 6 hours.
d) 95% of the applications have a waiting time in between 4.85 and 7.15 hrs.

Answers
Question Answer

Question Answer

Question Answer

Question Answer

11

21

31

12

22

32

13

23

33

14

24

34

15

25

35

16

26

36

17

27

37

18

28

38

19

29

39

10

20

30

40