LSSG Green Belt Training Measure: Finding and Measuring Potential Root Causes
DMAIC Six Sigma - Measure Objectives Identify Inputs and Outputs Determine key inputs and outputs for the process and measures to be analyzed Measure Process Capability Collect data and compare customer requirements to process variation Revise Charter Validate project opportunity and perform charter revision Measure Control Analyze Improve Define
Agenda for Measure 1. 1.Types of Measures/Setting Targets 2. 2.Data Collection and Prioritization, MSA 3. 3.SPC, Control Charts 4. 4.Process Capability
Measures Purpose of measurement: Performance of a process vs. Expectations Select Measures “SMART” Objectives Clear operational definitions E.g. Losing Weight ObjectiveLose 13 Pounds in 3 months Secondary Objective Lose 1 Pound per week Driver(s)Calories consumed less Calories burned Critical Success Factors (Drivers Run 4 miles/day and consume less than 1500 calories/day Must measure both the result (Y) and the drivers (Xs). Measure daily – to determine if CSFs are met, and to make adjustments to plan.
LSS Measurement Measurement is not control! So, what is it? Causes/ Effects Measurement System Control System Historical data Current data Measurement vs. Control Measurement Plan DataOperational Definitions and Procedures What data type? How measured? What conditions? By who?Where measured? What sample size? How to ensure consistency of measurement? What is the data collection plan?
Setting Targets Set Targets Objective/Meaningful Management-employees collaboration Team goal compatible with value stream objective Balanced Score Card Perspectives Financial Customer Internal Process Learning & Growth
Agenda for Measure 1. 1.Types of Measures/Setting Targets 2. 2.Data Collection and Prioritization, MSA 3. 3.SPC, Control Charts 4. 4.Process Capability
Data Collection and Prioritization Some Collection Tools Customer Survey Work / Time Measurement Check Sheet Some Prioritization Tools Pareto Analysis Fishbone Diagram Cause and Effect Matrix
Work Measurement Goals of Work Measurement Scheduling work and allocating capacity Motivating workers / measuring performance Evaluating processes / creating a baseline Determining requirements of new processes
Time Studies Typically using stop watches For infrequent information - estimates OK Measure person, machine, and delays independently Medium Duration - not too short; not too long Eliminate Bias - Compute Standard times from observed times
Time Study: Calculations Step 1: Collect Data (Observed Time) Step 2: Calculate Normal Time from Observed Time, where: Step 3: Calculate Standard Time from Normal Time, where:
Time Study: Numerical Example A worker was observed and produced 40 units of product in 8 hours. The supervisor estimated the employee worked about 15 percent faster than normal during the observation. Allowances for the job represent 20 percent of the normal time for breaks, lunch and 5S. Determine the Standard Time per unit.
Data Analysis Tools Scatter DiagramRun Chart Can be used to identify when equipment or processes means are drifting away from specs Can be used to illustrate the relationships between factors such us quality and training Frequency Data Ranges Histogram Use to identify if the process is predictable (in control) Can be used to display the shape of variation in a set of data Control Chart
Cause and Effect Diagram Material Method Environmental ManMachine Effect
Pareto Charts Root Cause Analysis 80% of the problems may be attributed to 20% of the causes
Continuous Improvement Process Orlando Remanufacturing And Distribution Center
Phase 1: Internal Kickbacks Five Most Common Reasons For Returns From QA Missing/ Wrong Part Dirt/RustDefective Part LeaksPoor Insulation Impact of Reasons for Returns from QA - Weighted Average Weighted Avg. = % Occurring X Defect Cost (0-10, Based on Time to Repair) LeaksDirt/RustStainless Steel Missing/ Wrong Part Defective Part Poor Insulation Equipment To Be Remanufactured Tear Down And Wash Remanu- facture Reassembly Final Clean-up QA Unit Not OK To Customer
Why Dirt? (Fishbone) Environment Dust/Humidity Poor Lighting Space Limitations Methods Reworking Steel after Valves are Installed Need to Rinse Parts off after Sandblasting Lack of Communication QA to IT Rework Rinse Training Attention to Detail Poor Lighting Dust/Humidity Space Limitations Tools for $$ Cleansing Compounds Larger Wire Brushes Environment Dirt MachineryMaterials MethodsManMeasurement Materials Cleansing Compounds Need Larger WireBrushes People Need More Training More Attention to Detail – Do it Right First Time Machines Best Tools for $$? Measurement QA Manager Fixes Some Things Without Informing the Technicians
Why Leaks? (Fishbone) Environment High Temperatures Poor Lighting Methods Check Units for Ways They Could Leak Does Testing Create Leaks? Materials Bad Tubing “O” Rings Too Old (Dry) People Use Wrong Clamps Don’t Crimp Properly Forget to Connect Machines Need Rims That Make it Easier to Install Tubing Measurement No Testing for Leaks Prior to QA Which Mfr./Model Leak the Most? No Leak Testing Prior to QA Quality Check Don’t Crimp Properly Use Wrong Clamps Poor Lighting High Temperature Reengineer Rims “O” Rings Old Bad Tubing Environment Leaks MachineryMaterials MethodsManMeasurement Forget to Connect Mishandle Units Identify Most Occurrences
Variation Analysis Most variation without “special” causes will be normally distributed Variation is typically classifiable into the 6 M’s Variation is additive Variation in the process inputs will generate more variation in the process output Methods Environment Machinery Materials Man Measurement Output Variation is Present in All Processes!
Measurement System Analysis (MSA) Goal - To identify if the measurement system can distinguish between product variation and measurement variation Some key dimensions Accuracy Precision Bias Tools: Gage R&R, DOE, Control Charts
Agenda for Measure 1. 1.Types of Measures/Setting Targets 2. 2.Data Collection and Prioritization, MSA 3. 3.SPC, Control Charts 4. 4.Process Capability
SPC vs. Acceptance Sampling Acceptance Sampling: Used to inspect a batch prior to, or after the process Take Sample Receive Lot Meet Criteria? Accept Reject Rework /Waste Send to Customer Yes No Statistical Process Control (SPC): Used to determine if process is within process control limits during the process and to take corrective action when out of control LCL UCL
Statistical Process Control A Process is not in control when one or more points is/are outside the control limits Special Causes UCL LCL Process in Statistical Control Process not in Statistical Control UCL LCL UCL LCL A Process is in control when all points are inside the control limits Statistical process control is the use of statistics to measure the quality of an ongoing process
When to Investigate Even if in control the process should be investigated if any non random patterns are observed OVER TIME UCL LCL In Control UCL LCL Close to Control Limit UCL LCL Consecutive Points Below/Above Mean UC L LCL Cycles UCL LCL Trend - Constant Increase/Decrease
Control Chart Development Steps INPUTS OUTPUT X’s Y’s Identify Measurement 1 Collect Data 2 Determine Control Limits 3 Improve Process 4 ABCD Defects Start Eliminate Special Causes Reduce Common Cause Variation Improve Average
Frequently Used Control Charts Attribute : Go/no-go Information, sample size of 50 to 100 Defectives p-chart, np-chart Defects c-chart, u-chart Variable : Continuous data, usually measured by the mean and standard deviation, sample size of 2 to 10 X-charts for individuals (X-MR or I-MR) X-bar and R-charts X-bar and s-charts
SPC Attribute Measurements p-Chart Control Limits percentage defects (mean) Standard deviation of p Z Number of standard deviations n Number of observation per sample (i.e., sample size) UCL Upper control limit LCL Lower control limit Z- VALUE is the number of Standard Deviations from the mean of the Normal Curve Normal Distribution: Z-Value Z
p-Chart Example 1. Calculate the sample proportion, p, for each sample 2.Calculate the average of the sample proportions 3.Calculate the sample standard deviation 4.Calculate the control limits (where Z=3) 5.Plot the individual sample proportions, the average of the proportions, and the control limits
SPC Continuous Measurements Chart Limits R Chart Limits X-bar, R Chart Control Limits Shewhart Table of Control Chart Constants
SPC Continuous Measurements R Chart LCL UCL Sample Range Sample LCL UCL X-bar Chart
Proper Assessment of Control Charts Find special causes and eliminate If special causes treated like common causes, opportunity to eliminate specific cause of variation is lost. Leave common causes alone in the short term If common causes treated like special causes, you will most likely end up increasing variation (called “tampering”) Taking the right action improves the situation
Quarterly Audit Scores Quarter Score Did something unusual happen?
Quarterly Audit Scores What do these lines represent? Quarter Score
Quarterly Audit Scores Now what do you think? Quarter Score
Agenda for Measure 1. 1.Types of Measures/Setting Targets 2. 2.Data Collection and Prioritization 3. 3.SPC, Control Charts 4. 4.Process Capability
Process Capability Introduction “Voice of the Process” (The “Voice of the Data”) Based on natural (common cause) variation Tolerance limits (The “Voice of the Customer”) Customer requirements/Specs Process Capability A measure of how “capable” the process is to meet customer requirements Compares process limits to tolerance limits
Process Capability Scenarios natural variation specification A natural variation C specification natural variation B specification natural variation D
Process Capability Index, C pk Capability Index shows if the process is capable of meeting customer specifications Find the C pk for the following: A process has a mean of and a variance of The product has a specification of ± ± 4.00 Mean = Stdev = 1.5
Interpreting the Cpk Cpk > or = 0.33Capable at 1 * Cpk > or = 0.67Capable at 2 * Cpk > or = 1.00Capable at 3 Cpk > or = 1.33Capable at 4 Cpk > or = 1.67Capable at 5 Cpk > or = 2.00Capable at 6 * * Processes with Cpk < 1 are traditionally called “not capable”.. However, improving from 1 to 2 , for example, is extremely valuable.
Calculating Yield Task 1 Task 2 Task 3 Task 4 Task 5 96 units 4 rwk 98 units 2 rwk 95 units 5 rwk 90 units 10 rwk 96 units 100 units Traditional Yield (TY) Rolled Throughput Yield (RTY): another way to get “Sigma” level The Hidden Factory = TY - RTY The Hidden Factory = =0.19 Traditional Yield assessments ignore the hidden factory!
Six Sigma Quality Level Six Sigma results in at most 3.4 DPMO - defects per million opportunities (allowing for up to 1.5 sigma shift) ,000,000 10, ,000 1, DPMO IRS Tax Advice Doctor Prescription Writing Airline Baggage Handling Domestic Airline Flight Fatality Rate (0.43PMM) 93% good 99.4% good 99.98% good Restaurant Bills Payroll Processing SIGMA Is Six Sigma Quality Possible? Source: Motorola Inc.
Six Sigma Quality Six Sigma Shift The drift away from target mean over time 3.4 defects/million assumes an average shift of 1.5 standard deviations With the 1.5 sigma shift, DPMO is the sum of and , or 3.4. Instead of plus or minus 6 standard deviations, you must calculate defects based on 4.5 and 7.5 standard deviations from the mean! Without the shift, the number of defects is.00099*2 =.002 DPMO. Z P(<Z) P(<Z) * 1,000,
Quality Levels and DPMO Defects per million opportunities Assumes 1.5 sigma shift of the mean Sigma Level DPMO (Defects per million opportunities) Reduction from previous sigma level % % % % % Regardless of the current process sigma level, a very significant improvement in quality will be realized by a 1-sigma improvement!
Is Six Sigma Quality Desirable? 99% Quality means that 10,000 babies out of 1,000,000 will be given to the wrong parents! One out of 100 flights would result in fatalities. Would you fly? What is the quality level for Andruw Jones?