Denver, Colorado November 16, 2004 D. R. Corpron Senior Manager & Master Black Belt

Transcription

1 Using Process Simulation in Quantitative Management Denver, Colorado November 16, 2004 D. R. Corpron Senior Manager & Master Black Belt

2 1 Preview What is the problem? Why process simulation? Steps to perform the simulation Data checks Mapping the process Animating the process Interpreting the results

3 2 What is the problem? Can we achieve our project s objectives for quality and process performance? Subject to: Required to add new features or fix defects according to Customer change requests Two levels of priority... A. Must Have (90%) B. Nice to Have (45%) Required to perform work on a fixed allocation of hours in a specific time frame

4 3 What s the practice? SP 1.4: Monitor the project to determine whether the project s objectives for quality and process performance will be satisfied, and identify corrective action as appropriate. [PA165.IG101.SP104]

5 4 Why process simulation? Build on existing process maps Can judge common cause risk because it is probabilistic rather than deterministic Even aperiodic processes can be executed many times to see variability in time compressed views Easy to perform what if analysis with instant results and no disruption of the real process Aggregate performance of the parts to predict the mean and variability of the overall process Compare predicted performance to Customer s critical-to-quality requirements

6 5 What steps are required? Sample data from process Test normality & stability Map Process Animate Process Run Scenarios Evaluate capability

7 6 Start by mapping the process Customer Start Submit Change Request * Process Customers, suppliers, and actors End Review Board Assign Priority Release to Customer System Integration Test * This is a simplification of the actual process Create Patch; Integrate & Test Priority A Create Patch; Integrate & Test Priority B

8 7 Sample data on incoming change requests * Incoming rate of CR s is stable * Shortrun Variation UCL= _ X= 6 No. of CR's * Special Cause Point Eliminated * Average LCL= Oct-2002 Dec-2002 Feb-2003 Apr-2003 Jun-2003 Aug-2003 Oct-2003 Dec-2003 Feb-2004 Apr-2004 Jun-2004 Date

9 8 Check normality of incoming CR s Probability Plot of Incoming Normal - 95% CI Percent * Anderson- Darling Test shows data is normal enough Mean StDev N 22 AD P-Value Incoming

10 Sample productivity Date 9 Dec-2003 Feb-2004 Apr-2004 Jun-2004 * Process is stable (and predictable) * Shortrun Variation * Average UCL= _ X= LCL= 6 People Hours per CR Oct-2002 Dec-2002 Feb-2003 Apr-2003 Jun-2003 Aug-2003 Oct-2003

11 Productivity is normally distributed 10 Probability Plot of Effort per CRs Normal - 95% CI Percent * Data is normal enough Mean StDev N 22 AD P-Value Effort per CRs

12 11 Input data via dialogue boxes Resources: Number of workers required to perform the task Task: Time to perform the task (in this case a normal distribution with the mean and standard deviation entered) Attributes: Any data calculation that needs to be performed (in this case the number of Priority A change requests to be implemented)

13 12 Animate the process * Model is executed many times * Generates a random number of CR s per month * Calculates effort * Assigns an A or B priority

14 13 Analyze the results * Simulated results show the P D ata process more LS L 0 than T arget capable * of U S L delivering S ample M ean required S ample N 25 S td ev (O v erall) performance Process Capability of Effort_A (using 95.0% confidence) LSL * Process Sigma = 3.1 ( ) USL O v erall C apability Z.B ench 1.59 Low er C L 0.75 U pper C L 2.70 Z.LS L 1.90 Z.U S L 1.93 P pk 0.63 Low er C L 0.41 U pper C L 0.85 C pm * Low er C L * O bserv ed P erformance % < LS L 0.00 % > U S L 0.00 % Total E xp. O v erall P erformance % < LS L 2.90 % > U S L 2.70 % Total 5.61 * First Pass Yield is 97.3%

15 14 Conclusion High confidence that Customer s requirements can be met Know how to adjust process variables to handle more or less change request demand

16 15