Hi Maturity in the CMMI Services Context
2 likesâ˘409 views
The document discusses the application of CMMI services in enhancing service delivery through strategic perspectives and effective modeling of service systems. Key insights include the dynamics of request handling, resource utilization, and the impact of shift timings on SLA compliance, with a focus on production support for banking applications. Recommendations include optimizing team resources and structuring shifts to improve service efficiency and reduce queue loads.
Hi Maturity in the CMMI Services Context
- 1. - 1 - âHi Maturity in the CMMI Services Context" Chinmay Pradhan QAI Global
- 2. CMMI For Services - 2 - ⢠Helps Build Better Services ⢠Looks at Services from Strategic perspective ⢠Applicable across broad spectrum of work done ⢠Aligned and Leverages existing service management frameworks eg ITIL, CoBIT etc
- 3. Snapshot of Services Where CMMI SVC has been applied - 3 - ⢠Majority is for Application Management â˘Production Support â˘Change Requests â˘Bug Fixes ⢠Other Non Conventional work types are also catching up Work Types WIth SVC Implemented Application Management and Support Training Services Risk Consutling ERP Configuration ERP Support Staff Augumentation BPO
- 4. Service System ⢠Services are useful intangible and non-storable results delivered through the operation of a service system. Âť CMMI SVC ⢠Services are characterized by âSimultaneity âHeterogeneity ⢠A Service System is a combination of âPeople using â˘Tools and Resources to execute âProcess Steps to complete an service operation - 4 -
- 5. Service Components: Elements that help deliver a Service - 5 - Work products Work Processes Tools Infrastructure People Streamlined Service Delivery
- 6. Modeling in an Application Support Service - 6 -
- 7. Typical Expectations - 7 - Year-on-Year Savings Technical Complexity
- 8. Assumed system - 8 - No of Requests per Day ⢠Based on actual average TAT per step and priority â˘Helped analyze the capacity required per day and hence plan for the month ⢠Was simple to use and understand But!!!
- 9. In RealityâŚ. - 9 - 19 4 ⢠Events are dynamic with multiple factors affecting the performance ⢠The inflow of tickets changes constantly ⢠The performance for each of the steps is not fixed, but variable An animated view of the service system
- 10. In RealityâŚ. ⢠There were far too many dynamic factors to be considered in the entire resolution process like âIf a high Priority ticket came in all others have to be dropped till this is fixed âIf resources are busy then tickets will be in queue âThe Process Performance is a function of the complexity and not just the priority, ⢠Additional complexity: can have high priority simple tickets and also complex tickets. - 10 -
- 11. - 11 - ⌠to control the Key Process Outputs Key Process Inputs: Control these ⌠Output Variability âOutputs inherit variability from inputs.â Sub Process and its Impact
- 12. One approach for Modeling Such a System - 12 -
- 13. Letâs Take an example The Service System we have picked up is âService to resolve Production support requests for large banking application âThere are 2 different kinds of requests that come in to the queue, Type 1 and Type 2. Each having its own priorities and SLAâs âThe Service provides a 6X18 hr support. Some of the challenges faced by the service system are âThe resource allocation and utilization vis a vis the SLA performance is sub optimal i.e if the SLA compliance is comfortable then there is low utilization âTeam composition and shift allocation - 13 -
- 14. Creating the Model The service system was aptly described by a queue based system. To model such a system we required: ⢠Standard Simulation Modeling Tool. E.g Process Model ⢠Resources with the knowledge of such modeling and the skill to use the tools ⢠Process maps of actual events ⢠Map of all possible occurring conditions i.e high priority gets picked first, there are shifts with breaks, ⢠Actual data with respect to the âRequest classifications i.e no of type 1 and type 2, breakup of their priorities. âThe Turnaround time for each process steps âWaiting time if any - 14 -
- 15. The data required â Not very different from what is typically collected ⢠The data collected were âA sample of the actual turnaround time for each process steps by their priorities and types âRequest incoming patterns; quite an eye opener. âOperating condition were ⢠Shifts of 9am to 6pm, 6pm to 3am ⢠Till now separate teams work on separate types ⢠If high priority request comes it is picked up immediately - 15 -
- 16. The Model Type 1 Type 2 Storage Decision Study and Response Study and Response 2 Analysis Analysis 2 Debugg Debugg 2 Close Request Engineer 1 Engineer 2 Engineer 3 Engineer 4 - 16 - â˘Daily incoming pattern described as statistical distributions for each type of request â˘Data was entered for the entire time period for support i.e 9 am to 6pm and then 6pm to 3 am â˘Requests had priority assigned using probabilities â˘Requests arrive at a system before they are assigned and allocated across
- 17. The Model Type 1 Type 2 Storage Decision Study and Response Study and Response 2 Analysis Analysis 2 Debugg Debugg 2 Close Request Engineer 1 Engineer 2 Engineer 3 Engineer 4 - 17 - â˘Current team assignment was depicted using the shift timings â˘One Resource was to complete the whole request before he or she was free ⢠Before the shift started or during breaks the queues will build up â˘If a high priority ticket enters the system, it will get addressed immediately â˘SLA are established for all ticket categories and priorities
- 18. Using the Model with current team settings - 18 - â˘Shift 1 â˘Engineer Group 1 (4 Resources) â˘Engineer Group 3(4 Resources) â˘Shift 2 â˘Engineer Group 2(4 Resources) â˘Engineer Group 4( 4 Resources) â˘Average idle time of 15% for Eng 1 and 14% for Eng 3
- 19. Using the Model with current team settings - 19 - â˘Average SLA at 95% CI is predicted to be â˘Type 1 â˘P1 92% â˘P2 91% â˘P3 89% â˘Type 2 â˘P1 89% â˘P2 74% â˘P3 98% With a target of 90% compliance there could be quite some misses
- 20. Evaluating Strategies : ⢠The Model was used to ⢠Simulate various scenarios and Identify âEvaluate available strategies to improve Utilization and SLA compliance âIdentify the potential areas of improvement in the process steps. - 20 -
- 21. Evaluating Strategies ⢠One of the things noticed in the setup was the queue that was getting built up in the non working period of 3 am to 9 am. ⢠Since the data for the inflow had not been analyzed before it was not realized that the engineers used to start work with the queue that was leading to SLAâs being breached ⢠What would be the impact if the team did not have the queue? ⢠Also it was noticed that there was typically some free resources in the day in the first shift but the second shift was tight. ⢠What if we overlap shift timings? ⢠What if we cross skill the people in the night shift so that anyone free can take in the other queue? - 21 -
- 22. Changes Made in the rules ⢠The inflow was modified to evaluate the impact of yanking away the requests logged during the non working hours. ⢠One of the resources was cross trained and assigned to both the queue so that requests can be resolved from both. ⢠The timings of this resource were made on an overlapping slot of 2pm to 11 pm so that there is an availability in both the shifts. - 22 -
- 23. Using the Model with changes - 23 - â˘Shift 1 â˘Engineer Group 1 (4 Resources) â˘Engineer Group 3(2 Resources) â˘Shift 2 â˘Engineer Group 2(2 Resources) â˘Engineer Group 4( 1 Resources) â˘Mid Shift â˘Engineer 5 ( 1 Resource) â˘A reduction of 6 resources from the team â˘Average un-utilization was the highest for Eng 4 at 53% for the only person there Type 1 Type 2 Storage Decision Study and Response Study and Response 2 Analysis Analysis 2 Debugg Debugg 2 Close Request Engineer 1 Engineer 2 Engineer 3 Engineer 4 Engineer 5
- 24. - 24 - â˘Average SLA at 95% CI Predicted â˘Type 1 â˘P1 93% â˘P2 88% â˘P3 90% â˘Type 2 â˘P1 93% â˘P2 88% â˘P3 100% SLA Compliance much healthier Using the Model with changes
- 25. - 25 - 0.00 10.00 20.00 30.00 40.00 50.00 60.00 StudyandResponse StudyandResponse2 Analysis Debugg Analysis2 Debugg2 Storage PercentofTotalMinutes Top 7 Hot SpotsTop 7 Hot Spots Percentage of NVA Minutes Percentage of BVA Minutes Percentage of VA Minutes â˘The Study and Response Process Step was seen to the most variable process step and identified for further analysis and improvement Using the Model with changes
- 26. Inferences ⢠âYankingâ the queue for the non working hours can reduce the load on the system ⢠Cross training can be a significant leverage ⢠Overlapping shift timings can greatly impact the SLAâs as well as the effective team utilization. ⢠The initial step of study and response is also identified as the potential area for improvement - 26 -
- 27. Key Takeaways - 27 - ⢠Involve subject matter experience âMeaningless (statistical) relationships should be discarded âExpected (statistical) relationships can be verified ⢠GIGO âReview and feedback cycle ⢠Be clear on objectives and expectations from simulation ⢠Ensure adequate interactions between âModel builder, Management, Practitioners ⢠Train personnel operating the model âTo know how to use it â make appropriate inferences âTo know when it is not working â seek help ⢠Check against known results
- 28. Tips and Tricks on Model usage in Project Management - 28 -
- 29. Service Transition - 29 - â˘Right turnaround time to promise â˘Right SLA â˘Right Volume
- 30. Capacity Management - 30 - 16 People handle 71 work requests Is this Ok If volume increases by 20% then how many resources?? By when?? Capacity during Service Continuity?? SLA during SCON
- 31. Setting Up Service Delivery - 31 - Type 1 Type 2 Storage Decision Study and Response Study and Response 2 Analysis Analysis 2 Debugg Debugg 2 Close Request Engineer 1 Engineer 2 Engineer 3 Engineer 4 Engineer 5 Resource Overlapping VS Type 1 Type 2 Storage Decision Study and Response Study and Response 2 Analysis Analysis 2 Debugg Debugg 2 Close Request Engineer 1 Engineer 2 Engineer 3 Engineer 4 No Overlapping
- 32. Dependency Management - 32 - Dependency on customer answers
- 33. Common Examples of Controlled Factors: Enough????? - 33 - Sl No Y Parameter (Outputs) X ( Controllable factors) Impacted Sub Processes Monitored 1 % SLA Met by Priority Skill of Resources in a team/Shift Optimal No of Resources in a shift/team Usage of Knowledge Database TAT of Resolution for each Incidents TAT of Response for each of the incidents Assignment Time for each incident No of backlog incidents per day 2 Utilization of Resources Skill of Resources in a team/Shift Optimal No of Resources in a shift/team Usage of Knowledge Database TAT of Resolution for each Incidents TAT of Response for each of the incidents Assignment Time for each incident No of backlog incidents per day
- 34. What is Critical!!!!:2013 Malayasian GP Pitstop2.05Secs Time Difference between Winner and 2nd :4.2secs - 34 - Pit stops not mandatory Critical Sub Process can be outside the chain of direct delivery process step
- 35. Watch Out for - 35 -
- 36. Misaligned Goals: Improvement in Productivity in T&M - 36 -
- 37. Swamped by information - 37 -
- 38. Measurements at in-appropriate granularity - 38 - Measurement is a function of work complexity and Need for information not convenience Measurement does not always mean a permanent measurement system
- 39. Propagate beyond Delivery and Interweave - 39 - 24X7 support is driven by transport: Do they do capacity management???
- 40. Questions - 40 -
- 41. References ⢠Using Process Performance Models to enhance CAM in CMMI for SVC- Mukul Madan and Chinmay Pradhan; Presented at SEPG NA 2011 ⢠Service Management: Operations, Strategy, and Information Technology- - James A. Fitzsimmons, Mona J. Fitzsimmons ⢠Introduction To Operations Research- Billy Gilett ⢠Process Model User Guide and Tutorial. http://www.processmodel.com/ ⢠CMMIÂŽ for Services, Version 1.3 â CMMI Product Team (CMU/SEI- 2010-TR-034) ⢠Improving Organizational Alignment Leveraging High Maturity Principles: Sankararaman D: HMBP 2012 - 41 -
- 42. Please feel free to write in. Chinmay Pradhan QAI Global [email protected] - 42 -