application of 6-sigma for service improvement – case ... · there are some fixed parameters in...
TRANSCRIPT
Application of 6-Sigma for
Service Improvement –
Case Study: Mala Tang AQM Term Project: Final Report
Koshi, NFor Earnest
Li, Ran
Shen, Yi
June 11, 2008
2
Table of Content
Table of Content ............................................................................................... 2
Executive Summary ......................................................................................... 3
1 Overview: Project Selection ....................................................................... 4
1.1 Background Information ................................................................... 4
1.2 Problems and its Symptoms ............................................................. 4
2 Define ........................................................................................................ 6
2.1 Goals and Expected Results ............................................................ 6
2.2 Framework ....................................................................................... 6
2.3 Assumptions ..................................................................................... 8
2.4 Constraints ....................................................................................... 9
2.5 Variables .......................................................................................... 9
3 Measurement ........................................................................................... 11
3.1 Video Analysis ................................................................................ 11
3.2 Revised Process Diagram .............................................................. 11
3.3 Questionnaire ................................................................................. 13
4 Analyze .................................................................................................... 16
4.1 Cause and Effect ............................................................................ 16
4.2 Correlation & Pareto ....................................................................... 16
5 Improvement ............................................................................................ 19
5.1 Design of Experiment ..................................................................... 19
5.2 Simulation ...................................................................................... 23
6 Conclusion & Outlook .............................................................................. 28
Appendix A: Questionnaire ............................................................................ 30
Appendix B: DOE results ............................................................................... 31
3
Executive Summary
6-sigma is a powerful tool to detect errors within a process regardless
whether it is a manufacturing process or service process. We applied the
6-sigma Methodology with its roadmap DMAIC (Define – Measure – Analyze –
Improve – Control) on the serving process of the Mala Tang window of the
Taoli canteen in order to reduce the service time per customer and therefore
the waiting time.
During the define step we drew our project scope, specified our goal
concerning the improvements we want to make and made assumptions
concerning the process. A CTQ tree and a rough process diagram were drawn
to illustrate the process understanding. Constraints were set limit the range of
the task and Variables (decision and dependency) were defined to give
additional insight into the whole field of the project.
The measurement step contains a video analysis of the service process at
the canteen window and a revised and improved process diagram with times
measured for each single action. Furthermore a questionnaire was designed in
order to find out more about the average number of ingredients ordered by
students and their satisfaction with the present situation concerning waiting
time and serving time. In addition to that suggestions were collected to find out
the factors that need to be changed in order to achieve a higher customer
satisfaction.
The analyze step deals with a cause and effect diagram to find many
factors that might have an influence on the output – the long waiting time. After
a qualitative selection of the most important ones a correlation matrix helped to
find an order of the main factors and their importance.
In the Improvement stage a DOE was conducted to reveal the impact of
those factors which are considered as the most important in a process. The
following simulation on Java basis that was made using data from the
measurement phase provided additional data for another DOE to test the other
factors
4
1 Overview: Project Selection
1.1 Background Information
Tsinghua University is a large community with tens of thousands of people
and the area it takes can almost compare to a town. Besides classroom,
lecture hall and laboratories for students and professors, it also has its own
supply and logistic service systems just like an isolated world. Hence there is
much space for improvement for the service systems in our campus. In this
research project, DMAIC (Define – Measure – Analyze – Improve - Control),
the implementation steps for 6-sigma Projects are followed to analyze and
improve the service quality of the Mala Tang window of the Taoli refectory.
Quality especially service quality is a crucial factor in the fast growing
modern China. 6-sigma is in the large bunch of the quality management
methods and it is very popular nowadays. 6-sigma can help to improve the
service process. The term 6-sigma is based on a statistical measure that
equals to 3.4 or fewer errors of defects per million opportunities. 6-sigma can
be described as a business improvement approach that seeks to find and
eliminate causes of defects and errors in manufacturing and services process
by focusing on outputs that are critical to customers.
6-sigma provides a blueprint for implementation of total quality system. Its
roadmap – DMAIC (Define-Measure-Analyze-Improve-Control) means: Define
the problem of the process; Measure the performance of the process; Analyze
the cause if the process, Improve the process, reduce waste; Control the
process, eliminate the occurrence of the same problem.
1.2 Problems and its Symptoms
A significant phenomenon in our campus is the long queues in the
refectory during the rush hour (i.e. after-class hours). For example, the
average time for a student from joining a queue to getting the of food in the
Mala Tang window of Taoli refectory is about 10 minutes, while the minutes for
enjoying this meal are just 15! Thus the waiting (or wasted) time takes 40% of
the whole time of a student in the refectory (Figure 1)
5
Unit: minute
15
10Eating
Waiting
Figure 1-1: Composition of time spent in Taoli refectory
6
2 Define
2.1 Goals and Expected Results
The 6-sigma team aims at reduce the total waiting time to 7 minutes or
less and remain the same level of customer satisfaction. The critical-to-quality
(CTQ) tree is shown in figure 2. There are 3 CTQs: diversity, waiting time and
serving time. And the diversity is required to be high, while the others are
required to be low.
Figure 2-1: CTQ flow down for Mala Tang serving process
2.2 Framework
Figure 3 shows the SIPOC framework of our project and the scope of
improvements
7
Mala Tang
Serving Process
Input: Ingredients Output: Mala Tang
Staffs & Refectory
Customer: Students
Our project scope
Supplier: the University
(Cannot be controlled and set as constraints)
Figure 2-2: SIPOC framework
Our scope of work focuses on the Mala Tang serving process and the
related output. We do not consider the complete input and output of the
process. Within the scope is furthermore the staff of the refectory. A rough
process diagram (Figure 4) helps to illustrate the general idea of the serving
process at Taoli refectory. As one can see the „Store‟ does not lie within the red
line and is therefore not part of the project scope. We focus on the interaction
between „Students‟ and „Staff‟ which embodies the service process.
8
Malatang serving process
Staff StoreStudents
Order
Order more?
Yes
Determine price
No
CookPay
Previous order
ready?
Yes
Bring to front
Wait
Is it my order?
Take
Yes
No
Yes
Replenish
No
No
Add spices & rice
Is any
ingredient
almost used
up?
Figure 2-3: Mala Tang serving process & the project scope
2.3 Assumptions
The queue of waiting for service is G/G/2, and the distribution of
inter-arrival and service time will be further studies and assigned in our
simulation. This is in accordance with the fact of the Mala Tang window of Taoli
9
refectory.
The patience of customers in the queue is unlimited and they will not leave
until they get the service. This is generally correct since the waiting people
rarely leave the queue before get the service at the Mala Tang window of Taoli
refectory.
The queue follows First-Come-First-Serve (FCFS) criterion. It is true since
the Mala Tang queue hardly has queue jumpers.
The amount of spice is infinite. It is almost true since the usage of spice is
little and it almost needs no replenishment during the process of service.
2.4 Constraints
There are some fixed parameters in the refectory which is difficult or
impossible for the 6-sigma team to change and is considered as constraints:
Total number of cook units is 6 in the Mala Tang window of Taoli refectory.
Total number of front staffs is 2 in the Mala Tang window of Taoli refectory.
Total room space of the Mala Tang window of Taoli refectory is a constraint
and cannot be modified.
Although the proportion of ingredients (vegetables, meat, and rice) can be
changed, the amount remains constant due to the limited inventory space
2.5 Variables
The variables are divided into 2 categories: decision variables, which can
be decided by the 6-sigma team, and dependency variables, or outputs, which
are directly or indirectly determined by the decision variables
2.5.1 Decision Variables
The decision variables are: the process itself, proportion of ingredients
and the layout
10
2.5.2 Dependency Variables
As mentioned before in section 1.3, the dependency variables for our
evaluation are: average / maximum waiting time, average / maximum queue
length, cost, and some subjective factors such as satisfaction of students
11
3 Measurement
3.1 Video Analysis
Given on the 1st version of the process diagram we made videos from the
Mala Tang serving process of the Taoli refectory. We intended to determine the
time for the single actions the cooking operator needed from the beginning of
taking the order over the cooking process till the end of the process the
handing over of the finished dish including the rice.
We shot the videos secretly in to provide an authentic working condition
and without manipulation concerning motivation or anything else which might
have affected the outcome
3.2 Revised Process Diagram
The analysis of the procedures shown in the videos made us redraw the
process diagram. From the video we could determine the times needed for
each action step during the whole service process. Our observations from the
new process diagram with process times were:
This process contains a lot of branches and therefore is has a lack of
standardizations;
Much time is spent in the region between input two and output two of the
operation sub process;
"Replenish sub process" takes almost 1 minute. Proper configuration of
the ingredients in front to reduce the frequency of replenishment, the total
efficiency will improve a lot.
12
Figure 3-1: Process diagram version 2, 1/3
Figure 3-2: Process diagram version 2, 2/3
13
Figure 3-3: Process diagram version 2, 3/3
3.3 Questionnaire
3.3.1 Questionnaire Design
We designed a survey distributed among random students who came to
the Taoli refectory on the May 11th 2008. Altogether we handed out 50 sheets
of questionnaires and among those 47 valid returned back to us.
The questionnaire contained the questions involving the frequency to
have Mala Tang, time, number of ingredient ordered, perception of queue
length, waiting time and satisfaction, etc. Concrete questions can be found in
Appendix A (in Chinese). The aim of this survey was to detect the customers‟
demand towards possible service improvement in Mala Tang process and their
satisfaction regarding the waiting / queering time
The questionnaire contained the listed questions. The aim of this survey
was to detect the customers‟ demand towards possible service improvement in
Mala Tang process, parameters related to customers (for modeling) and their
satisfaction regarding the waiting / queering time.
3.3.2 Statistical Results
From the questionnaire there are some conclusions:
Most people (73%) will not give up queueing even if the queue is longer
14
than 2 windows, and 1/3 of the people will never give up (Figure 8). So in our
model, we added an assumption that servers are always busy.
Figure 3-4: Acceptable queue length
Another main findings from the statistical result of the survey is the fact
that in average the students are not satisfied (Figure 9) with service and
waiting time at the Mala Tang window of the Taoli canteen.
Figure 3-5: Dissatisfaction level
15
Note: If we give the “dissatisfaction” levels some weight:
No 0
Little 1
Dissatisfied 3
Very Angry 9
Then we can get the average dissatisfaction level of 3.563
There are also some improvements regarding the serving time as
suggested by the customer themselves, Figure 10 gives number of the
suggestions by categories.
Figure 3-6: Customers' suggestions for further Improvements
16
4 Analyze
Now we know that customers want to be served faster. This will
automatically lead to higher satisfaction level among students and will reduce
the overall queuing time for everybody.
The next step will be to gather a high quantity of reasons which causes
that long waiting / serving time.
4.1 Cause and Effect
A fishbone chart, also called „Cause and Effect‟ diagram, is helpful to brain
storming many reasons that can lead to an outcome and is the first step to
identify the key-reason and most crucial factors that have the highest influence
on the outcome.
Figure 4-1: Cause and Effect Diagram
Since it is a service process and not a manufacturing process the 5 P‟s
are uses instead of the 5 M‟s. The main outcome is „long waiting time‟ which is
caused by 5 major symptoms and each has several reasons or causes itself.
4.2 Correlation & Pareto
After the Selection of the most crucial factors from the Cause and Effect
Diagram a correlation analysis of these main causes was mode. After serious
consideration, some factors were not taken into further consideration since it is
17
not possible or hardly possible to change the factors in order to achieve
improvements. The analytic hierarchy process (AHP) method is used to
determine the weight of those causes (Figure 12), excluding some reasons
that cannot be further improved.
Figure 4-2 AHP result of weight of causes
From the weight above a Pareto-chart is drawn to illustrate the crucial
factor to concentrate on during further project process.
18
Pareto Chart for Causes
0.341
0.2004
0.1422 0.12490.0922
0.05870.0226 0.0176
1.000
0.9007
0.8085
0.5414
0.6836
0
0.2
0.4
0.6
0.8
1
No std. Layout Pos.of
ingredient
Many
branches
Prop.of
ingredient
Low
space
Fatigue Mood
Causes
Imp
ort
an
ce
Figure 4-3 Pareto chart for causes
The accumulated weight of the main causes shows that a total of 5 main
causes contribute over 90% of the whole amount of causes. So the 5 critical
factors or reasons for a high waiting time at Mala Tang window of Taoli
refectory are found now.
1. No standard
2. Process layout
3. Position of each single ingredient
4. Many branches in the actions and during process
5. Proportion of ingredients
The first and fourth points are more related to psychological issues and
can be further studied by design of experiments (DOE). Our efforts in
improving the process time will concentrate on these 2 main causes. The other
3 are systematic factors and can be studied by modeling and simulation.
19
5 Improvement
5.1 Design of Experiment
5.1.1 Introduction
This DOE should find out about the influence of standardized patterns and
branches on the actual working or process time. The primary objective of this
design of experiment is to find the crucial factors that might change the
process time of a task.
DOE is known as a structured, organized method for determining the
relationship between factors that affect a process and the output of that
process. After the experiment we want to use the Response Surface
Methodology to find the optimum combination of factors that yields to a
minimum working time.
Our objective is to find out how much standardized actions and branches
contribute to the process time of a task (measured in seconds) and ultimately
find a fitted model to best predict the final working time
5.1.2 Methodology
Figure 5-1 DOE Scene Layout
Figure 16 shows a general overview of the scene layout on which we
conveyed our DOE.
There are 9 tasks carried out during the DOE:
1 Walk to one side of the source
2 Pick an item
3 Walk to center table
4 Put the item on the center table
20
5 Walk to the other side of the source
6 Pick an item
7 Walk to center table
8 Put the item on the center table
9 Next cycle, until all books is moved to center table
In order to investigate the influence of the factors we picked from the
Pareto-chart analysis on the outcome of the DOE. The output of the DOE is
the time needed to complete all the tasks under certain working conditions.
The Factors considered in the 23 experiment is as shown in Table 1
Table 5-1 Factors of DOE
Factors Level
- +
Branch A No branch Has branch
Sequence (Std.) B Fixed order Free Order
Unexpected tasks (Std.) C No 3
Description of factor A
“-“: the person can use all books in source 1 and then use books in source 2.
“+”: the person should use one book from 1 and then in the next cycle the person
should use one book from 2, and then continue
Description of factor B
"-": the person should pick one book and then one piece of paper in a cycle
"+": the person may either pick a book or paper first in each cycle, decided by random
variable generator
Description of factor C
"-": no interruptions
"+": when finish a cycle, the person may be interrupted by other things. After which
cycle is decided by random variable generator. The total number of interruptions in one
repeat (16 cycles) is 3
21
5.1.3 Results
Table 5-2 Experiment results of DOE
Factors
A + - + - + - + -
B + + - - + + - -
C + + + + - - - -
Result Mean
time 206.53 191.03 188.02 183.6 186.14 191.63 177.92 175.94
Using data from Table 2, the analyzing of factorial design and factorial
plotting in MINITAB 15 are performed. Figure 5-2, 5-3 and 5-4 are the results
Figure 5-2 General analysis results from MINITAB
Figure 5-3 Main effects plot for process time
22
Figure 5-4 Interaction Plot for Process Time
Figure 5-3 shows the main effects and indicates that the sequence order
and the appearances of unexpected tasks have the main influence on the final
process time. This confirms our findings from the Pareto-Chart where we
identified those as critical factors that have the highest contribution towards a
long process time and therefore leads to high waiting time in a service process.
Figure 5-2 and figure 5-4 underlines the high importance on focusing on
the two factors „unexpected task‟ and „sequence‟ which both are requirements
that stand for a standard process diagram where each action happens
according to a given and predefined sequence.
23
5.2 Simulation
Figure 5-5: Simulation Screenshot
The simulation is based on the process diagram version 2 and the results
from the questionnaire. Some information like the number of ingredients is on
the basis of the statistical result of the questionnaire and is defined as a
random value within a certain range. Each type of ingredients has a probability
to be selected. This probability will called selection probability. The simulation
was developed with JAVA and shows the complete process of the Mala Tang
window of Taoli refectory.
5.2.1 Time as Output
The simulation provided data for another DOE. We picked the values
which are randomized as new factors that influence the total serving time. The
factors are:
Factor A: Capacity of the basket
24
“-“: all is set to 30
“+”: ingredients with high demand are set to 40, the others are set to 20
Factor B: Variance of Selection Probability
“-“: the variance of selecting different ingredients remains small
“+”: the variance of selecting different ingredients is high
Factor C: Ingredient Position
“-“: the probability of selecting ingredients which has a high or low distance to
each other is the same
“+”: the probability of selecting ingredients which are closer to each other is higher
Using data from simulation the different random seed, we get different
replication for our 23 factorial design. Figure 5-6, 5-7 and 5-8 are the results
Figure 5-6 General analysis results for simulation
25
Figure 5-7 Main effects plot
Figure 5-8: Interaction Plot
The plots show that the factor position has the highest contribution
towards the process time. The interaction and strong correlation of the factors
should be taken into account since synchronized their changes all result in a
higher process time.
26
However, p-value is too low, which implies that the unidentified block
factors affecting service time are too many. Hence we turned to another
indicator: number of finished orders.
5.2.2 Finished Orders as Output
Figure 5-9: Simulation DOE Results
Figure 5-10: Main Effects Plot
27
Figure 5-11: Interaction Plot
We can see that all 3 factors have strong positive correlation with number
of finished orders.
28
6 Conclusion & Outlook
6-sigma was a useful method to identify the key factors of the service
process. It could have a great effect on improving the service time.
Using 6-sigma roadmap DMAIC, refectory in general can raise their
throughput and serve a higher number of customers than they do now which
means that they can greatly raise their effectiveness and efficiency.
The major conclusions we drew were:
1. Fewer branches in the process make lower service time
2. Standardization of the tasks, especially fixed task performing order and
fewer occurrences of unexpected tasks can give lower service time
3. The affection of position and proportion of ingredients is not significant
on average service time.
4. However, there is integrated optimization so that the total number of
finished orders in the given hours has a significant increase. In concrete, if
the most demand ingredients are put at front and give more resource like
capacity, total number of finished orders will increase. Moreover, the
increasing will become more significant as the demands for different
ingredients become more different.
It was not possible for us to match our defined goal. Though we think that
the reduction of time is possible, but we cannot predict the queue length and
therefore the waiting time since many factors are with the customer and the
variation is too high to promise a certain level of service speed.
Our report does not include a control stage. We thought that it is not
appropriate to use statistical control methods to see whether the service
process is working well or not. Since variation that comes from the customer
29
which has a great influence on the service time might lead to the conclusion
that the process time is too high, though all improvements were already
implemented.
Since the service quality is theoretically improved by 6-sigma more
customer requirements can be met, but not all. The service still has some
problems to deal with like the variety of customer group. The next step of
improvement would be to take the popular ingredients into account and make
re-arrangement of the ingredients according to their frequency of usage. By
this it is not guaranteed that every process will be accelerated but though the
average throughput-time will shortened and hence this will finally lead to a
smaller queuing time.
.
30
Appendix A: Questionnaire
桃李园麻辣烫服务过程调查
本问卷用于 2008 年清华大学工业工程系质量管理研究生课大作业顾客情况调
查,不会用于任何商业用途及对您产生任何不良影响。谢谢您的支持。
下面是有关清华大学桃李园餐厅麻辣烫窗口服务过程的 8 个问题。请根据您自己的感觉填
写。
1. 您多长时间去桃李园吃一次麻辣烫?________________
2. 您一般是几点去吃?
中午
11:00 - 11:30
晚上
17:00 – 17:30
11:30 - 12:00 17:30 – 18:00
12:00 - 12:30 18:00 – 18:30
12:30 – 13:00 18:30 – 19:00
3. 您一般会点几个菜?________________
4. 在看到队伍大概多长的时候您会放弃排队?
A. 多长都要排
B. 队尾排到旁边第 3 个窗口
C. 队尾排到旁边第 2 个窗口
D. 队尾排到旁边的窗口
E. 只有很少人或没人排队时才考虑去看看
5. 您吃一次麻辣烫在队伍中等待的时间一般有多久?________________
6. 从点好菜到拿到做好的麻辣烫一般会等多久?________________
7. 您感觉吃一次麻辣烫的等待时间长吗?
A. 没感觉
B. 短点当然更好,但现在这样也能接受
C. 非常希望等的时间可以短点
D. 时间太长了,强烈要求缩短
8. 您个人认为桃李园麻辣烫的服务有什么可改进的地方吗?
31
Appendix B: DOE results
Replication 1
Factors Result
A B C Person 1 Person 2 Person 3
Mean time Perf. Seq. Time / s Perf. Seq. Time / s Perf. Seq. Time / s
+ + + 8 219.4 12 215.8 17 209.8 206.50
- + + 6 194.5 13 172.5 24 184.3 183.50
+ - + 4 178.5 15 173.4 21 178.2 187.23
- - + 3 191.7 16 181.8 18 185.2 186.23
+ + - 5 189.8 9 176.6 20 187.4 184.60
- + - 1 195.7 14 191.8 23 203.6 197.03
+ - - 2 174.6 11 178.6 22 180.3 179.13
- - - 7 185.4 10 168.4 19 175.2 179.37
Replication 2
Factors Result
A B C Person 1 Person 2 Person 3
Mean time Perf. Seq. Time / s Perf. Seq. Time / s Perf. Seq. Time / s
+ + + 41 216.3 32 214.3 33 190.5 207.03
- + + 44 204.6 29 204.2 40 189.4 199.40
+ - + 48 191.3 30 186.2 36 194.4 190.63
- - + 46 193.3 27 186 35 177.5 185.60
+ + - 42 201.9 28 189.1 39 184.9 191.97
- + - 43 192.3 31 186.3 37 188.7 189.10
+ - - 47 186.4 25 181.2 38 179.6 182.40
- - - 45 184.6 26 171.9 34 163.8 173.43
Replication 3
Factors Result
A B C Person 1 Person 2 Person 3
Mean time Perf. Seq. Time / s Perf. Seq. Time / s Perf. Seq. Time / s
+ + + 59 209.5 65 211.5 53 197.2 206.07
- + + 62 199.2 67 187.2 55 184.2 190.20
+ - + 64 182.2 68 196.7 54 179.7 186.20
- - + 60 186.5 69 179.4 52 171 178.97
+ + - 58 188.1 70 174.5 50 183 181.87
- + - 57 186.8 66 185.7 51 193.8 188.77
+ - - 61 168.6 72 171.8 49 176.3 172.23
- - - 63 172.7 71 184.5 56 167.9 175.03