The Econoic and Scientific Context of Quality Iproveent and Six Siga By Soren Bisgaard The Eugene M. Isenberg Professor of Technology Manageent Eugene M. University of Massachusetts-Aherst Ljubljana, Slovenia April 7, 4 Copyright 4 by Søren Bisgaard. All rights reserved.. Copyright 4 by Søren Bisgaard. All rights reserved.. Overview. Econoic Context A. Innovation B. Inforation. Scientific Context A. Role of Deduction and induction B. Role of statistics and experientation C. Manageent of change 3. Quality iproveent and Six Siga A. DMAIC for existing products, services and processes B. DFSS for new products, services and processes 4. Statistical research 5. Conclusion
. The Econoic Context A. Innovation B. Inforation Copyright 3 by Søren Bisgaard. All rights reserved..3 SB&A Schupeter On Innovation Joseph Schupeter: The fundaental ipulse that sets and keeps the econoic engine in otion coes fro the innovation of new products, new ethods of production or transportation, new arkets,new fors of industrial organization The priary reason for profits is as a preiu for the risk of innovation Copyright 4 by Søren Bisgaard. All rights reserved..4
Today's Econoic Realities: An abundance of new copeting products and technologies Copetitors with siilar products and technologies invade arkets Shorter product life cycles bigger arket pressure Learning effects: copetitors world wide learn faster to execute better, faster and cheaper Copetitors in other countries eerge with significantly different cost structures World wide counication technology and logistics ake the arket global and extreely copetitive Copyright 4 by Søren Bisgaard. All rights reserved..5 Product Life Cycle Model Ebryonic Growth Maturity Ageing $ s Sales Profits Cash flow Tie Copyright 4 by Søren Bisgaard. All rights reserved..6 3
Sales The Necessity of Innovation to Sustain Profitability Tie Copyright 4 by Søren Bisgaard. All rights reserved..7 The Innovative Process Quality is not just about quality! It s about innovation! Product innovation, process innovation, radical and increental innovation! Radical innovations: entirely new products, processes and services Exaple: the light bulb, wireless counication, Increental innovations: the adaptation, refineent and enhanceent of existing products, processes and services Exaple: next generation of a icro processes, next years autoobile odel, Modern society needs innovations to stay copetitive, growth and prosper! Copyright 4 by Søren Bisgaard. All rights reserved..8 4
Rate of ajor innovations Changing Character of Innovations Process innovations Product Innovations Early Product Life Late Tie Adapted fro Abernathy and Utterback (978) Copyright 4 by Søren Bisgaard. All rights reserved..9 Iproved Features Econoics of Six Siga Larger Market Share Higher Sales Price + Iproved Revenues Iproved Quality Reduction of Deficiencies Iproved Productivity & Cycle tie Reduced Scrap and Waste Reduced nuber of Warranty Clais Reduced Costs Higher Profit Copyright 4 by Søren Bisgaard. All rights reserved.. 5
Quality Manageent in a Nutshell Quality Planning Quality Control During Operations Cost of Poor Quality Sporadic Spike Original zone of Quality Control Quality Iproveent New zone of Quality Control Chronic waste Lessons Learned Tie Adapted fro Juran Copyright 4 by Søren Bisgaard. All rights reserved.. Quality, Econoics, Inforation, Knowledge Generation and Learning Copyright 3 by Søren Bisgaard. All rights reserved.. SB&A 6
Transaction Costs Ronald Coase -- Transaction costs: Like friction in echanical systes Defects, rework, snags and delays are frictions Iproving delivery quality is about reducing certain transaction costs But to be copetitive long ter, we ust also iprove the design quality Whether iproving design or delivery quality, it is achieved through knowledge and inforation. Copyright 4 by Søren Bisgaard. All rights reserved..3 Inforation and Knowledge We need to generate inforation and knowledge about Who the custoers are, What is the value proposition; what is value to the custoers How we can eet and possibly exceeds their needs and expectations And learn How to iprove processes How to develop products that satisfy the custoers Copyright 4 by Søren Bisgaard. All rights reserved..4 7
Hayek on Inforation Friedrich von Hayek: the proble [of econoic success] is ainly one of rapid adaptation to changes ultiate decisions ust be left to people who are failiar with these circustances. Copyright 4 by Søren Bisgaard. All rights reserved..5 The Invisible Hand & Local Knowledge The knowledge we need to generate is local knowledge Decision-aking local to where the action is, is ore effective than centralized decision-aking Much the sae way as the invisible hand of the arket decisions has proven to be ore effective than a centralized plan econoy Copyright 4 by Søren Bisgaard. All rights reserved..6 8
Local Decision Making To support decision-aking and help eployees fro factory floor worker to the CEO we need local knowledge Knowledge needs to be generated where it is needed, where it can be interpreted and understood and where it intelligently can be acted upon But to generate inforation and to learn, we need skills in the application of scientific ethod Copyright 4 by Søren Bisgaard. All rights reserved..7. The Scientific Context Copyright 3 by Søren Bisgaard. All rights reserved..8 SB&A 9
Good Science is Inductive and Deductive A replica of Galileo s apparatus ade in 775 for the Grand Duke of Tuscany. Instituto e Museo di Storia della Scienza, Florence in: Harré, Ro, Great scientific experients. Oxford University Press, Oxford/New York 98, p. Copyright 4 by Søren Bisgaard. All rights reserved..9 Data Iterative Process of a Scientific Investigation: We Need Induction and Deduction Deduction Induction Hypothesis Adapted fro Box, Hunter and Hunter (979) Copyright 4 by Søren Bisgaard. All rights reserved..
An Epirical Foundation For the best and safest ethod of philosophizing sees to be, first diligently to investigate the properties of things, then establish the by experient, and then to seek hypotheses to explain the. Sir Isaac Newton Copyright 4 by Søren Bisgaard. All rights reserved.. Six Siga: A Systeatic Prograed Innovation Approach A successful progra spearheaded by copanies such as General Electric and Allied Signal Originated at Motorola as their quality iproveent progra Different fro traditional TQM and ISO 9; really about systeatic innovation in general, not just quality control Copyright 4 by Søren Bisgaard. All rights reserved..
What is Six Siga? A progra for near eliination of defects fro every product and process A disciplined quantitative approach for iproveent of defined etrics aligned with the overall strategy Can be applied to all business processes: anufacturing, transactions, finance and services Focused on carefully selected Projects In a nutshell: Six Siga is the application of scientific ethod to iprove processes and products Copyright 4 by Søren Bisgaard. All rights reserved..3 DMAIC: A Siple Proble Solving Discipline Define: Select relevant probles to work on» Stating a proble in easurable and actionable ters Measure: easuring the variation of the perforance data of the proble Analyze: finding the sources of variation to the perforance data Iprove: eliinating or enhancing the highest drivers of the perforance data variation Control: establishing controls to anage the gains of the proble solution Copyright 4 by Søren Bisgaard. All rights reserved..4
Phase : Define Scope and Boundary Define Defects Develop Tea Charter and Select Chapions Estiated $ Ipact Get Leadership approval The Six Siga Strategy Phase : Measure Map the process and Identify Inputs and Outputs Make Cause and Effects Matrix Establish Measureent Syste Capability Establish Process Capability Baseline Phase : Analyze Perfor Multi-vari Analysis Develop Input-output relations Identity Critical Process Inputs Develop FMEA Phase 3: Iprove Verify Critical Process Inputs Optiize Critical Process Inputs Reduce variability Phase 4: Control Develop Control Plan Ipleent Control Plan Verify Long Ter Capability Transfer to operations Continuously Iprove the Process Copyright 4 by Søren Bisgaard. All rights reserved..5 Is Six Siga A Fad? 9 Defects in % 8 7 6 5 4 3 Index Copyright 4 by Søren Bisgaard. All rights reserved..6 All Jan 3 Feb 3 Mar 3 April 3 May 3 June 3 July 3 Not to the CEO of this copany!!! 3
Quality Iproveent J. M. Juran: All quality iproveent takes place project by project and in no other way! Copyright 3 by Søren Bisgaard. All rights reserved..7 SB&A Organization Leadership: Chapions Master Black Belts/consultant (MBB) Black Belts (BB) Green Belts (GB) Key to Success: Select successful people, Not just war bodies!!! Copyright 4 by Søren Bisgaard. All rights reserved..8 4
The Guiding Coalition Chapion Master Black Belt Project Definition / Charter Project Manageent Results Manageent Black Belt Copyright 4 by Søren Bisgaard. All rights reserved..9 Organizational Structure: Chapions Master Black Belts BB's BB's BB's GB's GB's GB's GB's GB's GB's GB's GB's GB's GB's GB's GB's Building capability to drive breakthrough at all levels Copyright 4 by Søren Bisgaard. All rights reserved..3 5
Design for Six Siga (DFSS)? DFSS: The design of new processes, products and services fro scratch: - Where a process, product or service does not previously exist - Where the existing process, product or service is not capable of being iproved Focuses on systeatically gathering the voice of the custoer, prioritizing requireents, and building those requireents into new, processes, products or service - Identifies targeted custoer requireents Focuses on growth of the business Not just for engineering for all business areas Copyright 4 by Søren Bisgaard. All rights reserved..3 The DMADV Road Map Define Identify the new (or odified) product Measure Plan and conduct research to understand custoer needs Analyze Develop alternative design concepts, select a concept for high level design and predict the capability Design Develop the detailed design, evaluate its capability and plan a pilot test Verify Conduct the pilot test, analyze the results and ake design changes as needed Copyright 4 by Søren Bisgaard. All rights reserved..3 6
Strategy: Three Horizons* Horizon 3 Create viable options Profit Horizon Build eerging businesses DFSS Horizon Extend and defend core businesses Traditional Six Siga (DMAIC) Tie (years) *Adapted fro Badhai, Coley and White (999) Copyright 4 by Søren Bisgaard. All rights reserved..33 Why Organizational Transforations Often Efforts Fail! Research show that eight coon errors in anaging change, two of which are:. Not establishing a sense of urgency. Not systeatically planning for and creating short ter wins Copyright 4 by Søren Bisgaard. All rights reserved..34 7
Research Shows: An Effective Strategy. Map out how the organization should look after a ajor change. Use a series of rapid cycle projects that provide the eployees opportunities to develop their skills in anaging change 3. As they learn to change, ount increasingly larger scale, ore strategic efforts and ove upstrea 4. Periodically review and odify the overall strategic plan Benefit: The projects pay for the progra Copyright 4 by Søren Bisgaard. All rights reserved..35 How to Get Started? Establish A Sense of Urgency Create a Guiding Coalition Evaluate strategic goals and low hanging fruit opportunities Develop tea charters and select teas Just-in-tie GB/BB training Manage projects for success: First wave ust be a success Prepare for second wave Copyright 4 by Søren Bisgaard. All rights reserved..36 8
Schupeter Again Joseph Schupeter: The fundaental ipulse that sets and keeps the econoic engine in otion coes fro the innovation of new products, new ethods of production or transportation, new arkets,new fors of industrial organization The priary reason for profits is as a preiu for the risk of innovation Copyright 4 by Søren Bisgaard. All rights reserved..37 Statistical Challenges Statistical Research Copyright 3 by Søren Bisgaard. All rights reserved..38 SB&A 9
The Iportance of Practice for the Developent of the Theory of Statistics Brewer/cheist/statistician W. S. Gosset working for Guinness Brewery in Dublin: Student s t-test R. A. Fisher at Rothasted Experiental Station Henry Daniels working for the Wool Industries on variance coponents George Barnard working during WWII George Box at ICI: Response Surface Methods Gwily Jenkins working on airplane design and tie series analysis collaborating with George Box to develop process control ethods resulting in Box & Jenkins tie series ethod used in econoetrics John Tukey working for Bell Labs: Signal processing and exploratory data analysis Etc. Copyright 4 by Søren Bisgaard. All rights reserved..39 Soe Recent Inspirations for Research in Multivariate Process Control Copyright 3 by Søren Bisgaard. All rights reserved..4 SB&A
Three Tie Series of Teperature Readings fro a Cheical Process k4_c 54.33 54. k4_c 589.89 588.78 k4_c3 573. 57.34 Tie Copyright 4 by Søren Bisgaard. All rights reserved..4 54.33 54. k4_c 589.89 588.78 k4_c 573. 57.34 k4_c3 54. 54.33 588.78 589.89 57.34 573. Copyright 4 by Søren Bisgaard. All rights reserved..4
3D Plot of 3 Dependent Tie Series: MC, MC MC3 546 545 544 543 k4_c 54 54 54 539 538 537 588 k4_c 589 59 59 57 573 574 k4_c3 Copyright 4 by Søren Bisgaard. All rights reserved..43 Hotelling s T T p - Diensions (x x)s (x x) UCL LCL p( n + )( n ) F( n( n p) p, n p) ( α) Copyright 4 by Søren Bisgaard. All rights reserved..44
MTB > %tsquared c-c3 c7 5 Tsquared Chart of k4_c-k4_c3 Tsquared 5 Saple 3 4 5 Copyright 4 by Søren Bisgaard. All rights reserved..45 StackTep 6 5 4 Profiles 3 4 5 6 3 4 5 6 3 4 5 6 7 8 9 MCLocation Data July : MCTEMPPROFILE.MPJ Copyright 4 by Søren Bisgaard. All rights reserved..46 3
6 StackTep 5 4 MCLocation 3 4 5 6 7 8 9 3 4 5 6 CTie Copyright 4 by Søren Bisgaard. All rights reserved..47 Contour Plot of StackTe MCLocation 9 8 7 6 5 4 3 375 4 45 45 475 5 55 55 575 6 65 65 3 4 5 6 CTie Copyright 4 by Søren Bisgaard. All rights reserved..48 4
Principal Coponent Analysis: PCA Data atrix: k diensions, n observations X x X M X M Xk x n L L D (X) Σk k x x k M kn P (p, K,p k ) Copyright 4 by Søren Bisgaard. All rights reserved..49 Y or Y y y M k X P L y n Y M M L y Y kn k P X Copyright 4 by Søren Bisgaard. All rights reserved..5 5
6 The Econoic and Scientific Context of Quality Iproveent and Six Siga Copyright 4 by Søren Bisgaard. All rights reserved..5 Spectral Decoposition k k n p p p p Σ + + λ λ L X p,y X, p Y k k K P X Y PY X P X Y Copyright 4 by Søren Bisgaard. All rights reserved..5 Reconstruction of X k k D D then Let p p p p p p p p P Y P X PY X Y Y Y * * * + + λ λ λ λ K M O O K M M * ),, ( ) ( ) (
7 The Econoic and Scientific Context of Quality Iproveent and Six Siga Copyright 4 by Søren Bisgaard. All rights reserved..53 Building up X k then Let,,, K M M * * PY X Y Y Y Copyright 4 by Søren Bisgaard. All rights reserved..54 Building up X k,,, * K PY PY X X * * * Y P Y Y Y Y Y P Y Y P X X X * * * M M M M M M ) ( ) ( * δ
Building up X Let X X k δ then : X δ X Copyright 4 by Søren Bisgaard. All rights reserved..55 The Original Surface 5. 4.5 4. 3.5 Di 3 5 5 5 Tie Copyright 4 by Søren Bisgaard. All rights reserved..56 8
Surface Coponent 5.5 StackX 4.5 3.5 Di 3 5 5 5 Tie Copyright 4 by Søren Bisgaard. All rights reserved..57 Coponent.3.. StackX. -. -. -.3 -.4 Di 3 5 5 5 Tie Copyright 4 by Søren Bisgaard. All rights reserved..58 9
Surface and 5.5 XRecand 4.5 3.5 Di 3 5 5 5 Tie Copyright 4 by Søren Bisgaard. All rights reserved..59 Coponent 3.5 StackX3. -.5 Di 3 5 5 5 Tie Copyright 4 by Søren Bisgaard. All rights reserved..6 3
The reconstructed surface 5. 4.5 XRec 4. 3.5 Di 3 5 5 5 Tie Copyright 4 by Søren Bisgaard. All rights reserved..6 Autocorrelation in Process Monitoring Copyright 3 by Søren Bisgaard. All rights reserved..6 SB&A 3
Exaple: Out of Control? I Chart for MC 546 UCL546 Individual Value 545 544 543 54 Mean543 54 54 5 Observation Nuber LCL54 Descriptive Statistics: MC Variable N Mean StDev MC 543.. Copyright 4 by Søren Bisgaard. All rights reserved..63 Effect of Autocorrelation A siple autoregressive process y t ϕyt ε ~ N(, σ ) t k ρ φ k σ y + ε σ ε φ ρ autocorrelation t Copyright 4 by Søren Bisgaard. All rights reserved..64 3
A Siple Approach to Dealing with Positive Autocorrelation Positive autocorrelation eans essentially that we don t have as uch data as we think and that the estiated variability in sall saples is saller that it really is To copensate for this we can siply increase our estiate of siga or widen the control liits to ore than ±3σ Alternatively we can onitor the residuals fro the tie series fit with regular control charts Copyright 4 by Søren Bisgaard. All rights reserved..65 I Chart for MC 548 UCL547 Individual Value 546 544 54 Mean543 54 LCL539 5 Observation Nuber Copyright 4 by Søren Bisgaard. All rights reserved..66 33
Another Tie Series Tie Series Plot for _c 58 _c 58 579 3 4 5 6 7 8 Tie Copyright 4 by Søren Bisgaard. All rights reserved..67 I and MR Chart for _c Individual Value 58.5 58.5 579.5 578.5 UCL58 Mean58 LCL579 Subgroup 3 4 5 6 7 8 Moving Range..5 UCL.3 R.348. LCL Copyright 4 by Søren Bisgaard. All rights reserved..68 34
EWMA Chart for _c 58.5 EWMA 58. UCL58 Mean58 579.5 LCL579 3 4 5 6 7 8 Saple Nuber Copyright 4 by Søren Bisgaard. All rights reserved..69 Autocorrelation Function for _c Autocorrelation..8.6.4.. -. -.4 -.6 -.8 -. 5 5 Lag Corr T LBQ Lag Corr T LBQ Lag Corr T LBQ.69.3 6..95 39.94 47.79 8 9. -..3 -.5 49. 49. 5 6 -..5 -..3 59.77 6.7 3.6.35 48.8 -.8 -.46 49.65 7..6 6.8 4 -.6 -.39 48.43 -.4 -.83 5.43 8 -.3 -.9 6.9 5 6 -.8 -. -.5 -.6 49. 49.3 3 -.9 -.9 -.4 -. 54.9 58.44 9 -. -.5 -.4 -.3 6.5 6.56 7..3 49.7 4 -. -.67 59.77 Copyright 4 by Søren Bisgaard. All rights reserved..7 35
Partial Autocorrelation Function for _c Partial Autocorrelation..8.6.4.. -. -.4 -.6 -.8 -. 5 5 Lag PAC T Lag PAC T Lag PAC T 3.69 -.34. 6. -3..7 8 9. -.5 -.9. -.43 -.77 5 6 7.6 -.4 -..5 -.38 -.93 4 5 6 7 -.5.3.9 -.8 -.45.8.8 -.75 3 4 -.4 -...4 -.33 -.5.3.34 8 9.6.5 -.3.5.4 -.5 Copyright 4 by Søren Bisgaard. All rights reserved..7 ARIMA (,,) Model: _c Final Estiates of Paraeters Type Coef SE Coef T P AR.984.6 9.5. AR -.37.66-3.49. Constant 65.836.4 473.84. Mean 579.79. Copyright 4 by Søren Bisgaard. All rights reserved..7 36
~ y ~ φ ~ + t AR() Process φ yt + yt a t Stationary! Variance: σ φ σ φ + {( + φ) a y φ } Copyright 4 by Søren Bisgaard. All rights reserved..73 Using Expanded Liits I and MR Chart for _c Individual Value 58 58 58 579 578 UCL58 Mean58 LCL578 Subgroup 3 4 5 6 7 8 Moving Range UCL. R.6768 LCL Copyright 4 by Søren Bisgaard. All rights reserved..74 37
Using Expanded Liits EWMA Chart for _c 58.3 58. UCL58 EWMA 58. 58. 579.9 579.8 579.7 579.6 579.5 579.4 Mean58 LCL579 3 4 5 6 7 8 Saple Nuber Copyright 4 by Søren Bisgaard. All rights reserved..75 Conclusion Schupeter: The fundaental ipulse that sets and keeps the econoic engine in otion coes fro innovation Quality is about innovation product and process innovation service and anufacturing Hayek: Need local inforation and local decision aking Statistics and scientific ethod play a key role Copyright 4 by Søren Bisgaard. All rights reserved..76 38