Achieving Customer Defined Quality

1
Achieving Customer Defined Quality

• Amit Deokar
• M.S. (Industrial Engineering)
• Systems and Industrial Engineering
• University of Arizona
• Tucson, AZ

2
Way to Customer-Defined Quality
Loss
Capability Indices
Conformance/ Non-Conformance
3
Capability Indices

• Characterizes what a process will produce in
  future
• Statistically stable process necessary

Specifications Minimally acceptable Product
Natural Process Limits What the Process will
produce
4
Some Capability Indices

• 
  - Juran (1974)
• Indicates the potential proportion conforming,
  due to the centering assumption
• 
  - Kane (1986)
• Accounts for the lack of centering assumption
• Reinforces the description of Cp as the
  potential capability
• 
  - Chen Spiring (1988)
• 
  - Hsiang Taguchi (1985)
• Relatively new and still not used very frequently
  in industry

5
Capability Indices

• Statistics that vary over time even though the
  underlying process does not change
• Easy to misinterpret making process monitoring
  more difficult
• Non-Linear Relationship between Capability
  Indices and Percent Nonconforming
• In complex products, combining specifications on
  various inter-related dimensions/components to
  get optimal values is difficult

6
Why not Capability Indices?

• Goal-Post Loss Zero-Defects Philosophy
• No incentive to improve the process
• Not inline with the Lean Manufacturing
  methodology
• Are we achieving Customer-Defined Quality?

7
Cost of Poor Quality

• Failure Costs
• Internal Failure Costs (e.g. scrap, rework)
• External Failure Costs (e.g. warranty charges)
• Appraisal Costs
• Prevention Costs

Reference Jurans Quality Handbook Juran
Godfrey
8
Loss (Quadratic)

• On Target with Min. Variance concept
• Incentive for Continual Improvement
• Incorporates Cost of Using Conforming Products

9
Loss (Quadratic)

• General Form of Quadratic Loss Function (QLF)
• Loss for deviation, , from target
• Average Loss in case of QLF
• (Independent of the probability distribution of
  the underlying process)

10
General Form of Average Loss

• where process distribution
• Loss Function used for
  the model

Expected Loss
Voice of the Customer
Voice of the Process
11
Graphical Representation
Figure 2
Figure 1
Reference Beyond Capability Confusion Donald
Wheeler
12
General Class of Loss Functions

• Reflected Normal Loss Function
  (Spiring, 1993)
• Modified Reflected Normal Loss Function
  (Sun, Laramee Ramberg, 1995)
• General Class of Loss Functions
  (Spiring Yeung, 1998)

13
Capability Index Motivated by QLF

• Cpm Average Quadratic Loss
• Relation to Expected Quadratic Loss
• Process Incapability Index (Greenwich (1995))
• Another form by Ramberg (2002)

14
Example in Automotive Transmissions

• 3 processes Frictional Clutch Plate
  Manufacturing
• None make more than 0.27 scrap relative to specs
• Scraping Cost - 2.50/part

Reference Taguchi Techniques for Quality
Engineering Phillip J. Ross
15
Example in Automotive Transmissions
Reference Taguchi Techniques for Quality
Engineering Phillip J. Ross
16
What do we learn?

• On-Target with reduced variation ? Competitive
  Product with Reduced Loss to Society (Both
  Producers Customers)
• Missing the target value - a serious loss
  compared to hitting the target with increased
  variation
• Capability Indices Cp Cpk - Bottom-Line
  improvements not motivated (Cpm - motivated by
  Loss Function concept itself)
• Process 2 seems to be expensive for the
  producers. But, with Loss
  Function at work, they are actually cheaper

17
Loss Function for Multiple Components

• For a system of independent components having
  similar functions
• Expected Losses due to deviations are given by,

18
Example in Tech Support Call Center

• Criteria
• Degree of problem resolution (R)
• Length of call time (X)
• Incremental Billing Plans
• 4 per min
• Customer Satisfaction Scores (VOC) Product
  Sales

19
Call Time Criterion

• Loss Linear
• Call Time Distribution Exponential (20, 25)

20
Problem Resolution Criterion

• Loss Quadratic
• Call Resolution Distribution Discrete

21
Loss Computation

• For Call Time length criterion,
• For degree of Problem Resolution criterion,

22
Loss Computation

• Total Loss
• Average Loss

23
Conclusions

• Conformance to Specifications Necessary but not
  sufficient
• Capability Indices Useful but not sufficient
• Loss Functions
• Focus on reducing the Loss to the Society
• Incorporate the True Voice of Customer
• Helps in designing better products processes
• Need to take research to industry