Blended Lean Six Sigma Black Belt Training – ABInBev ©2010 ASQ. All Rights Reserved. Answers to DDC Study “Improve”

Blended Lean Six Sigma Black Belt Training –

ABInBev

©2010 ASQ. All Rights Reserved.

Answers to DDC Study “Improve”

© 2010 ASQ. All Rights Reserved.2

About this Module

• You are the coach for the Green Belt doing the Brazil DDC project.

• He is planning to conduct a DoE• This is his first DoE so he has asked you for help filling

out the DoE planning sheet.

© 2010 ASQ. All Rights Reserved.3

Where He Is So Far

He indicates that he has completed the form as far as he can go without your help. (DoE Plan DDC.doc).He also indicates he has budget approval for 16 runs.What would you recommend and why?

© 2010 ASQ. All Rights Reserved.4

Determining the Factors

From the Pareto we know the four leading causes to test those we decided to use the following factors

1. Number of orders

2. Route in minutes

3. Truck loading time

4. Release time

© 2010 ASQ. All Rights Reserved.5

What Types of Experiment Could We Do?

Stat>DoE>Create Factorial Design>Display Available Designs

Viable designsViable designs

© 2010 ASQ. All Rights Reserved.6

The Viable Options

• A full factorial design without replication– Advantage

• No alias or confounding– Disadvantage

• No measure of error (saturated model) if all main effects and interactions are significant

• A half faction factorial design with two replicates– Advantage

• The replication provides a measure of error– Disadvantage

• Aliasing may be difficult for an inexperienced Green Belt to understand and fully explain

© 2010 ASQ. All Rights Reserved.7

Design Choice

Consensus was a full factorial design with the factor levels set at.

1. Number of orders 17 32

2. Route in minutes 56 73

3. Truck loading time 5.5 7.5

4. Release time 3 9

Complete the DoE Matrix

© 2010 ASQ. All Rights Reserved.8

Create the DoE Matrix

© 2010 ASQ. All Rights Reserved.9

Analyze the Data

• The experiment was performed and the data saved in DDC DoE.mtw

• Analyze the data and prepare to present the results

© 2010 ASQ. All Rights Reserved.10

Analyze the DoE

Stat>DoE>Factorial Analyze Factorial Design

© 2010 ASQ. All Rights Reserved.11

Analyze the DoE

For the initial model include all main effects and interactions

© 2010 ASQ. All Rights Reserved.12

First Model

ABCAC

ABDC

CDA

ABACDAD

ABCDBCDBDBCDB

121086420

Term

Effect

3.23

A Number of OrdersB Route in MinC Truck Loading in MinD Release Time in Min

Factor Name

Pareto Chart of the Effects(response is Number of Non Delivered Orders, Alpha = 0.05)

Worksheet: DDC DoE.MTWLenth's PSE = 1.25625

Reduce the model

© 2010 ASQ. All Rights Reserved.13

First Reduction

© 2010 ASQ. All Rights Reserved.14

Second Model

AC

C

CD

A

AB

AD

BD

BC

D

B

1086420

Term

Standardized Effect

2.57

A Number of OrdersB Route in MinC Truck Loading in MinD Release Time in Min

Factor Name

Pareto Chart of the Standardized Effects(response is Number of Non Delivered Orders, Alpha = 0.05)

Worksheet: DDC DoE.MTW

Reduce the model again

© 2010 ASQ. All Rights Reserved.15

Final Model

© 2010 ASQ. All Rights Reserved.16

Final Model

Release Time in Min

Route in Min

1086420

Term

Standardized Effect

2.16

Pareto Chart of the Standardized Effects(response is Number of Non Delivered Orders, Alpha = 0.05)

Worksheet: DDC DoE.MTW

© 2010 ASQ. All Rights Reserved.17

Check the Residuals

5.02.50.0-2.5-5.0

99

90

50

10

1

Residual

Per

cent

N 16

AD 0.249P-Value 0.703

20151050

2

0

-2

-4

Fitted Value

Resi

dual

3210-1-2-3-4

4.5

3.0

1.5

0.0

Residual

Frequen

cy

16151413121110987654321

2

0

-2

-4

Observation Order

Resi

dual

Normal Probability Plot Versus Fits

Histogram Versus Order

Residual Plots for Number of Non Delivered Orders

Worksheet: DDC DoE.MTW

The residuals are normally distributed with a mean of zero and constant variance. No reason to reject the model.

© 2010 ASQ. All Rights Reserved.18

Optimize the Performance

Stat>DoE>Analyze Factorial Design>Response Optimizer

© 2010 ASQ. All Rights Reserved.19

Setup the Optimizer

© 2010 ASQ. All Rights Reserved.20

Optimizer Output

CurHigh

Low1.0000D

Optimal

d = 1.0000

MinimumNumber o

y = 2.4937

1.0000DesirabilityComposite

3.0

9.0

56.0

73.0Release Route in

[56.0] [3.0]

Non delivered orders are minimized by using the lower route time and lower release time. The model shows we should be able to achieve an average of ~2.5 non delivered orders.

© 2010 ASQ. All Rights Reserved.21

How Much of the Variation is Explained?

Stat>ANOVA>General Linear Model

© 2010 ASQ. All Rights Reserved.22

Session Window

Source DF Seq SS Adj SS Adj MS F PRoute in Min 1 537.08 537.08 537.08 99.43 0.000Release Time in Min 1 312.41 312.41 312.41 57.83 0.000Error 13 70.22 70.22 5.40Total 15 919.71

Remove the spaces that are interior to the column and source names.

© 2010 ASQ. All Rights Reserved.23

Prepare to Graph

Source DF SeqSS AdjSS AdjMS F PRouteinMin 1 537.08 537.08 537.08 99.43 0.000ReleaseTimeiMin 1 312.41 312.41 312.41 57.83 0.000Error 13 70.22 70.22 5.40Total 15 919.71

Your session window should like like this.

Copy this data excluding the total row to the clipboard then paste starting in the title row of the worksheet.

© 2010 ASQ. All Rights Reserved.24

Create the Graph

Graph>Pie Chart

© 2010 ASQ. All Rights Reserved.25

Specify the Labels

© 2010 ASQ. All Rights Reserved.26

Pie Chart

Error7.6%

ReleaseTimeiMin34.0%

RouteinMin58.4%

Pie Chart of Source

Worksheet: DDC DoE.MTW

Over 92% for the change in Non delivered orders is explained by release time and route time.

© 2010 ASQ. All Rights Reserved.27

Conclusions

The DoE has identified factors that must be controlled to minimize the non-delivered orders which in turn should increase the sales rate.