DMAIC Define and Measure 6 DMAIC Define and Measure GEOP 4316
Define The initial phase of the DMAIC process Understand the problem(s), reasons, measures, individuals involved, the effects of it Create plans and determine needed resources Justify the need for the project Projects may not progress beyond this phase
Define Project Definition Projects must Have a financial impact or significant strategic value (e.g. solidify relationships with key customers) Results significantly exceed effort and investment Addresses a complex problem Aims at an improvement change of at least 70% Project must solve a business problem that hurts a key performance related to: Costs, Cycle time Customer satisfaction Revenue Potential Process Capability/ Capacity
Define Use of case writing tools for project justification As a company, our _____________ performance for the ____________ area is not meeting _______. Overall this is causing _______ problems., which are costing as much as $_____ per ________. Example As a company, our on time delivery performance for the health product area is not meeting the goal of 95% on time delivery. Overall this is causing customer satisfaction problems, which are costing as much as $2Milion in lost orders and revenue per year.
Define 1.1 Determine what needs to be improved. What is the problem ( a high level effect); e.g. high inventory levels consuming space and assets. Where is the problem occurring; e.g. the Westland Warehouse Timeframe of the problem; e.g. Since Product Family ds was introduced in 2017. Customers or business are affected by this problem; Production planning, sales given higher costs.
Define 1.2 Determine the outputs (CTQs = y’s) Characteristics or process outputs affected (the specific y’s); e.g. inventory levels. Identify the primary metric for each y; e.g. inventory days Identify tradeoff metrics to monitor for negative impact; e.g. orders filled.
Define 2. Determine associated processes and location a. List the major process steps and location (top level processes); e.g. PO generation, order replenishment, inventory reconciliation, production planning. b. Develop a top level flowchart 3. Determine the baseline performance of the y’s a. Estimate the magnitude of the problem using the primary metric; e.g. between 31.2 to 40.5 days of inventory with a mean of 37.4 days.
Define 4. ID the cost and impact of the problem a. Identify the cost centers affected by problem b. Based on a baseline improvement level, what is the financial impact of the project. Soft versus Hard Savings Soft: increased customer satisfaction > this increases loyalty and customer orders. Estimates included but hard to trace Hard: reduced inventory in warehouse = reduced assets and increased cash available.
Define 5. The problem Statement A description of the problem and the metrics Where its occurring and in what processes Timeframe and Magnitude Should not simplify it: audience will not know the problem and the processes. Inventory levels are too high. Inventory levels at the Westland warehouse are consuming space, taking up management time, and creating cash flow issues. Inventory levels are averaging 31.2 days with a high of 45 days. These levels exceed the 20 day target on 99% of the days since October 2016. We could save $400,000 if at the target level.
Define The level of improvement Based on current performance, entitled performance, and the team leader goals Entitled performance: the best performance that a process as currently designed has demonstrated in operation. Long term performance of a line: 90% on time delivery Mar 4 – Mar 23, 98% on time performance. 98% is the entitled performance So what x’s caused the 98% performance level. Breakthrough improvement: 70% change
Define 6. The Objective Statement 7. ID and recruit the team Based on the Problem Statement Improve some metric from baseline in t amount of time with an impact towards a corporate goal or strategy. Reduce Inventory levels Reduce inventory levels from an average of 31.2 days to an average of 20 days with a maximum of 25 days. This will be completed by August 2019. This project will save $400,000 per year in interest, space, and management effort. These savings are in support of our corporate goal of improving asset management and ROI. 7. ID and recruit the team 8. Obtain approvals and launch
Measure During the Define phase the critical process variables were identified targets were identified The Measure phase captures and analyzes process/ system data What is the gap between target and actual performance? Analysis of the gap is based on statistical analysis. Statistics is the key to understand variability. Data may not be available, and this phase includes data collection.
Measure Process/ Product Measurement Identification of error opportunities (for the product) Possible missing or wrong parts, information, delays in completing a task, … Defects per Opportunity Opportunities are the critical dimensions / characteristics of the product (unit) Directly related to the complexity Issue: overestimating the ops. with trivial elements. This will lead to over assessing the process capabilities
Measure Process capability Based on data and on specifications Specifications: performance values of a process or product characteristic that are acceptable to the customer. Specifications should not be arbitrary: unless they have a real reason (separate the good from the bad) they are a stumbling block to progress Too tight: spend more resources/ effort than necessary, too loose: customer will not be satisfied
Measure DC1 data: xave = 21 hours; s = 1.1 hours Six Sigma assumption: variables are normally distributed.
Measure DC1: xave (m) = 21 hours; s (s) = 1.1 hours Thus there is a 68.3% chance that the time to fulfill the order is between 19.9 and 22.1 hours … A 95.4% that is between 18.8 and 23.2 hours 17.7 19.9 22.1 24.3 18.8 21 23.2
Measure DC2: xave (m) = 19 hours; s (s) = 3.4 hours Thus there is a 68.3% chance that the time to fulfill the order is between 15.6 and 22.4 hours … A 95.4% that is between 12.2 and 25.8 hours 8.8 15.6 22.4 29.2 12.2 19 25.8
Measure DC1: xave (m) = 21 hours; s (s) = 1.1 hours What is the probability the fulfill time would be less than 22.1 hours? = 0.5 + .3413 = 84.13% What about less than 24.3 hours? = 0.5 + 0.3413 + 0.1359 + 0.0214 = 99.86% 21 22.1 23.2 24.3
Measure Sigma score (level): relates the process performance to the CTC limit Sigma score is how many standard deviations can you fit within the mean and the specification level Z = ( SL – xave ) / s s xave SL
Measure SL = 24 hours. DC1 data: xave = 21 hours; s = 1.1 hours Z = ( SL – xave ) / s = ( 24 – 21 ) / 1.1 = 2.72 2.72 s s 21 24
Measure SL = 24 hours. DC2 data: xave = 19 hours; s = 3.4 hours Z = ( SL – xave ) / s = ( 24 – 19 ) / 3.4 = 1.47 1.47 s s 19 24
Measure Key question. What is the probability of not meeting the CTX? In other words, what is the failure level for each DC? 19 21 24
Measure Relating Z to the long term defective level Get the Z value from the data. Find P(X ≤ Z) from the cumulative normal table Percent Defective = 1 –P(X ≤ Z) Defects Per Million Opp = (1 – P(X ≤ Z)) x 1M The discussion in this presentation will assume no process shift (the process mean does not move). For a discussion about the 1.5 shift see http://www.isixsigma.com/new-to-six-sigma/dmaic/15-sigma-process-shift/
Measure Relating Z to defects Another version is to assume the mean will shift (move) Get the Z value from the data. Get Zshift = Z – 1.5 Find P(Zshift) from the cumulative normal table Percent Defective = 1 –P(Zshift) DPMO = (1 – P(Zshift)) x 1M http://www.isixsigma.com/new-to-six-sigma/dmaic/15-sigma-process-shift/
Measure No shift DC1. Z = 2.72 P(X ≤ Z) = 99.67 Defective = 0.33% DPMO = 3,300 If they fill 1,000 orders per week, about 4 will take more than 24 hours 0.33% No shift 19 21 24
Measure No shift DC2. Z = 1.47 P(X ≤ Z) = 92.92 Defective = 7.08% DPMO = 70,800 If they fill 1,000 orders per week, about 71 will take more than 24 hours 7.08% No shift 19 21 24
Measure Processes can have USL, LSL or both. Sigma Score (Z) ZU = (USL – xave ) / s ZL = (xave – LSL ) / s Z = min. (ZU, ZL) Capability index (Cpk) = Z score/ 3. A capability index of 1 or more indicates a capable process. A Cpk = 1 is equivalent to a 3 Sigma process
Measure Example 1 Customers are complaining about pedal problems. CTQ = Part strength in psi of a bike pedal Data from 956 bike pedals resulted in an average of 255 with a standard deviation of 6 To meet the bike design specification (based on customer expectations), the LSL = 240.
Measure Z score (Sigma score) = (255 – 240) / 6 Z = 2.5 No shift LSL = 240 psi xave = 255 psi s = 6 psi Z score (Sigma score) = (255 – 240) / 6 Z = 2.5 No shift P(Z) = 0.9938 or about 0.62% defective 6,200 DPMO 15 6 6 xave = 255 LSL = 240
Measure Z score (Sigma score) 2.5 LSL = 240 psi xave = 255 psi s = 6 psi Z score (Sigma score) If the mean shifts towards the LSL by 1.5 Zshift = 1 P(Zshift) = 0.84134 or about 15.865% defective 158,650 DPMO 15 6 6 xave = 255 SL = 6 LSL = 240
Measure Example 1. Conclusions Process is not capable as Cpk = 2.5/3 < 1. Assuming not mean shift, process would produce less than 1% defective. Assuming a mean shift, process would produce about 15% defective.
Measure Example 2 A bag of aguacate dip must weigh between 3.5 and 3.95 ounces. CT Quality = No fewer than 3.5 CT Cost = No more than 3.95 Data from 300 containers resulted in an average of 3.79 ounces with a standard deviation of 0.07 ounces LSL = 3.5; USL = 3.95.
Measure Z = min. (ZU, ZL) = 2.29 With no shift; P(Z) = 0.989 sigma score Zu = (3.95 – 3.79) / 0.07 = 2.29 Zl = (3.79 – 3.5) / 0.07 = 3.85 Z = min. (ZU, ZL) = 2.29 With no shift; P(Z) = 0.989 Defective = 1.1 % 11,000 DPMO 0.29 0.16 0.07 0.07 0.07 LSL = 3.5 xave = 3.8 USL = 3.95
Measure Z = min. (ZU, ZL) = 2.29 Zshift = 2.29 – 1.5 = 0.79 sigma score Zu = (3.95 – 3.79) / 0.07 = 2.29 Zl = (3.79 – 3.5) / 0.07 = 3.85 Z = min. (ZU, ZL) = 2.29 Zshift = 2.29 – 1.5 = 0.79 With shift; P(Zshift) = 0.7852 21.48% defective 214,800 DPMO 0.29 0.16 0.07 0.07 0.07 LSL = 3.5 xave = 3.8 USL = 3.95
Measure 2.29
Measure A Sigma scores changes if: The mean shifts The standard deviation changes The specs change
Measure sigma score = (260 – 240) / 6.5 = 260 LSL = 240 = 255 = 3.07. Cpk > 1 No shift P(Z) = 0.9989 Defective = 0.1% DPMO = 1,000 With shift Zshift = 1.57 P(Z) = 0.9425 Defective = 5.75% DPMO = 57,500 Improvement as previous values were no shift: 6,200 DPMO w shift 158,650 DPMO After some changes to the equipment, the average psi is increased to 260. However, the standard deviation increased a bit to 6.5 psi. 20 6.5 6.5 6.5 xave = 260 LSL = 240 xave = 255
Measure New equipment to pack the dip is purchased and installed. Data from 400 containers resulted in an average of 3.7 ounces with a standard deviation of 0.02 ounces sigma score Zu = (3.95 – 3.7) / 0.02 = 12.5 Zl = (3.7 – 3.5) / 0.02 = 10 Z = min. (ZU, ZL) = 10 With no shift; P(Z) ∼ 1 Assuming sigma shift, P(Zshift) ∼ 1. This is a 10 sigma process. Basically, DPMO = 0. 0.2 0.25 LSL = 3.5 xave = 3.7 USL = 3.95 xave = 3.8
Measure Measurement System must be precise Measurement System Audits Repeated measurements of the same thing should result in the same value, regardless of other conditions. Measurement System Audits Used to determine how the MS is working Select several items, run them through the process several times. Measurement should be the same. Audits are very important in human measurement systems