How Six Sigma Organizations Implement CMMI Level 5 CMMI Technology Conference & User Group 17-20 November 2008 Rick Hefner, Don Corpron, Dave Miner Northrop Grumman Corporation Alice Parry, Raytheon Corporation 1
Background CMMI and Six Sigma are two well-known process improvement approaches with strong synergies When an organization knows the tools and methods of Six Sigma, organizational and project implementations take a more customer-focused perspective, and often yield greater value than traditional Level 5 implementations In this presentation, two leading Six Sigma and CMMI Level 5 organizations will share the ways in which Six Sigma has shaped their high maturity practices 2
Agenda A Tale of Two Organizations Six Sigma Approach for Quantitative Management Example 3
A Tale of Two Organizations Organization A Organization B 200 people, one building 10 projects for 3 clients; all fixed-price; all developing banking SW Deployed ML5 practices as a competitive discriminator during organization stand-up 18,000 people, offices in all 50 states 200+ projects for 20+ clients; fixed-price, cost-plus, LOE; SE, SW, HW, services CMMI ML5, ISO 9000, AS9100, etc.; continuously reorganizing and acquiring new pieces of the organization 4 How might their high maturity practices vary?
High Maturity Implementations Organization A Organizational goals make a profit (productivity, low fielded defects) Project goals same as the organizational goals Organization builds baselines and models around productivity and defects Projects select peer review and testing subprocesses for quantitative management Projects follow CMMI practices Organization B Organizational goals satisfy shareholders (growth, stability) Project goals all different because of different domains, different customer needs Organization builds baselines and models around productivity and defects Projects select a wide variety of subprocesses (e.g., training delivery, action item closure, estimation, field support, etc.) Projects follow Six Sigma approach 5
Focus in a Six Sigma Organization Are you measuring the right things? How do you know what s right? Stay focused on getting your product to your customer as promised! 6 Budget and monitor the value-producing processes. The ones that transform inventory into finished product
Things flow through a process 7 The flow of requirements through the processes is our chief concern Things are what customers pay for In manufacturing, materials are the things In design and development, requirements are the things In services, external Customer needs are the things In administration, internal Customer needs are the things
Identify the Project s Value Stream The transformation of requirements into product features and functions for which Customers pay money Project Planning Customer Customer Requirements are the inventory that transform as they go through the process Requirements Design Code/ Build Integrate Test 8
Identify the Measures For each value-added or value-producing process Identify what constitutes the inventory Identify how the inventory is measured Establish the measure for the rate of transforming the inventory into product 9
Collect and Analyze 10 Collect data for each valueadded, value-producing subprocess Collect at regular intervals Use voice of the process Analyze the data Establish the statistical understanding This is the Process Performance Baseline Compare against the allocated budget (subprocess capability)
Assess the Overall Ability to Achieve Incorporate the data from the statistical understanding into the process map The Process Performance Model Run simulations to assess the project s ability to get there from here Identify needed improvements Use the simulation to make decisions 11
Example Discrepancy Reports Data required: Submitted DR s per time period (arrivals) Open DR s (backlog) Resolved DR s REJECTED SUBMIT OPENED RESOLVED 12
Categorize DR s 1. Complete Failure: System crashes 2. Partial Failure: Required functionality does not work, and no workaround 3. Partial Failure: Required functionality does not work, but a workaround exists 4. Cosmetic: Defect does not materially affect any functionality 13
Collect a time series of measurement data about DR submittals Date Submitta ls Resolved Open 4/6/2007 4 0 4 4/13/2007 7 2 9 4/20/2007 6 3 12 4/27/2007 12 6 18 5/4/2007 13 : : 5/11/2007 19 : : 14
15 Analyze the Submittal data with a control chart I Chart of Submittals (Transformed) 6 1 5 2 4 3 2 Individual Value 4/6/2007 5/11/2007 6/15/2007 7/20/2007 8/24/2007 9/28/2007 Note: Transform 11/2/2007 12/7/2007 1/18/2008 2/22/2008 1 counting data with (c+.5) 0 2 2 2 2 Date Transformed with SQRT(c+.5) 1 1 1 UCL=5.689 _ X=3.209 LCL=0.728
16 A shift took place about October 26; use current performance I Chart of Submittals (Transformed) 7 6 Stage 1 Stage 2 Actual number is (2.392)² -.5 = 5 5 4 3 2 Individual Value 1 0 4/6/2007 5/11/2007 6/15/2007 7/20/2007 8/24/2007 9/28/2007 11/2/2007 12/7/2007 1/18/2008 2/22/2008 Date Transformed with SQRT(c+.5) UCL=4.586 _ X=2.392 LCL=0.197
3/25/2008 17 Likewise, analyze Resolved with a control chart I Chart of Resolved (Transformed) 14 12 10 8 6 4 Stage 1 Stage 2 Actual number is (3.37)² -.5 = 11 A process performance baseline Individual Value 2 0 8/7/2007 8/31/2007 9/18/2007 10/9/2007 10/30/2007 11/20/2007 12/11/2007 1/1/2008 1/22/2008 2/12/2008 3/4/2008 Date Transformed with SQRT(c+.5) UCL=9.97 _ X=3.37 LB=0
Simulate the process using the data Stakeholders 1 3 Start Submit DR Developers 5 Opened DR's In other words, Create a Process Performance Model using the Process Performance Baselines 4 Assign DR's Model adjustments may include: Reject Accept Inventory arrival rates Transformation rates 6 End 2 Resolve DR Staff levels and attrition Standard work schedule 18
10/14/2008 19 Analyze the open DR s with a time series chart Time Series Plot of Total SVT/IST Open 500 400 300 200 100 0 DR Backlog on 3/25 was 477 Total SVT/IST Open Nominal prediction is for a 11/25 finish 8/7/2007 9/18/2007 11/6/2007 12/25/2007 2/12/2008 4/1/2008 5/20/2008 7/8/2008 8/26/2008 Date This came from a simulation
Simulate the DR work-off 50 Weeks to Finish 40 30 Weeks 20 10 0 Sim #1 Sim #3 Sim #5 Sim #7 Sim #9 Sim #11 Sim #13 Sim #15 Sim #17 Sim #19 Sim #2 Sim #4 Sim #6 Sim #8 Sim #10 Sim #12 Sim #14 Sim #16 Sim #18 Sim #20 Count Sim #1 Sim #2 Sim #3 Sim #4 Sim #5 Sim #6 Sim #7 Sim #8 Sim #9 Sim #10 Sim #11 Sim #12 Sim #13 Sim #14 Sim #15 Sim #16 Sim #17 Sim #18 Sim #19 Sim #20 24 20 28 31 35 35 35 28 32 31 29 41 23 39 29 34 28 25 31 29 20
Analyze the results with a CDF Empirical CDF of Time to Finish Normal 100 Mean 30.35 StDev 5.244 N 20 80 75 Percent 60 40 20 25 Tied directly to remaining schedule 0 26.81 33.89 20 25 30 35 Time to Finish 40 45 21
Predict Outcomes If defect arrivals and resolution stay as is ; defects will not add risk to the end date Complete similar analysis for the other valueproducing processes 22
Summary Are you Getting there from here? Understand what produces value for the customers Set performance budgets Measure the valueproducing processes Model and analyze performance 23