Presentation by vignesh swamidurai

Mantra for Process Excellence
Vignesh Swamidurai
Assistant Consultant
TATA Consultancy Services

Taming RTB using
Central Composite
Design for Process
Improvement
PMI National Conference – Mantra for Process Excellence

Table of Contents
1. THE BUSINESS CHALLENGE ................................................................................................4
2. CENTRAL COMPOSITE DESIGN ...........................................................................................5
3. INITIATION ...............................................................................................................................6
4. APPLYING CCD TO TICKET OPTIMIZATION ........................................................................7
5. APPLYING CCD TO COST OPTIMISATION ..........................................................................9
6. SETTING UP THE GOALS FOR 2013 .................................................................................. 10
7. THE PATH TO OPTIMISATION ............................................................................................. 11
7.1 The Ticket Optimization Exercise ................................................................................... 11
7.1.1 Ticket Reduction using Causal Analysis and Resolution Process Area ........................... 11
7.1.2 The Results –Ticket Optimization................................................................................. 12
7.2 The Cost Optimization Exercise ..................................................................................... 13
7.2.1 Cost Optimization using Organizational Performance Management Process Area ........... 13
7.2.2 The Results ................................................................................................................ 14
8. CONCLUSION ........................................................................................................................ 15
9. ACKNOWLEDGEMENTS....................................................................................................... 15
10. REFERENCES.................................................................................................................... 15

Taming RTB using Central Composite Design
ABSTRACT
The application of Central Composite Design (CCD) in Software estimation is an innovative approach to
determine the Optimum goals for a project. This combination of Statistical estimation and highly matured
process deployment can result in substantial benefits and reduction in operating cost for an Organization
and can be applied to any RTB project across any industry. CCD is a statistical methodology of identifying
the independent variables or factors that affect a product or process, and then studies thei r effects on a
dependent variable or response in order to find the optimum setting of factors. In this paper, the Authors
have evaluated the usage of CCD for Software Optimization Estimation to determine the optimum goals
for a project. As a case study for this analysis, the Authors have chosen an RTB project for a Retail
Banking Customer. The different project parameters required for optimal performance have been
assessed and these parameters have been applied to CCD model to arrive at optimized goals. The two
primary focus areas were Ticket Optimization and Cost Optimization. The optimum goals for the project
have been derived through research of historical data and applying them through the CCD model which is
explained in the paper. The following goals were set for the project for 2013 as an outcome of CCD and
these determined goals were achieved successfully by the project team.
• 50% reduction in the number of Production Failures / Resolution Days
• 0.5 Million USD total cost of application savings / Soft Dollar savings
Keywords: Process Improvement, Cost Savings, Ticket Management, Central Composite Design, CMMI,
Pareto, RTB
1. THE BUSINESS CHALLENGE
Businesses in the Industry vary daily due to market dynamics in the global economy. Organizations are
striving hard to cope up with the ever-growing business needs, in addition to facing challenges such as
slim operating margins, rapidly changing technologies, decreasing time to market and strong competition
with counterparts. Coping up with these increased dynamism, uncertainty and complexities are a huge
challenge faced by the organizations of the current era.
Organizations are looking for every possible opportunity to reduce expenses, conserve resources and
increase their stock prices. An ideal scenario expects the Run theBusiness (RTB) cost to remain constant
and at a minimum, thereby allowing organizations to concentrate more on Change the Business(CTB)
projects.

However in the current scenario, due to the market dynamics, the cost incurred by RTB is increasing
exponentially. Therefore the CTB cost has to be compromised, to keep the business running. This causes
a huge pressure to reduce the day-to-day operating cost to grow more efficiently. The ideal scenario and
current scenario sample is depicted in Fig.1 and Fig.2.
Figure 1: RTB Vs CTB - Ideal Scenario** Figure 2: RTB vs CTB–Current Scenario**
** - Graphs plotted with mocked data to show the distribution, does not include actual spread
To reduce the RTB cost, there is a need for a verified Optimization technique and a proven Management
methodology to achieve good results. This paper discusses about the modernistic approach of using a
well-established Statistical Optimization model – Central Composite Design (CCD) for Software Process
Optimization. The paper also describes the use of Management Art to achieve the set of determined
goals using the Scientific Optimization model in an RTB Project.
This innovative combination of statistical study and management principles has facilitated in reducing the
operating cost of RTB, thereby helping to invest more in CTB, which in turn benefits the organization to
stay competitive and Customer-focused.
2. CENTRAL COMPOSITE DESIGN
Central Composite Design (CCD) is an established design approach for Optimization, which is
predominantly used in Metallurgical and Pharmaceutical industries. It is a technique that revolves around
the study of the influence of different variables, based on the outcome of a process. It involves identifying
the independent variables or factors that affect a product or process, and then studies their effects on a
dependent variable or response. CCD can also be treated as an enabler, to find optimum setting of
factors.

The implementation of CCD involves identifying the following three key factors:
1) A Factorial Design in parameters, each having two different coding levels.
2) A set of Centre Point for each parameter whose values is a median of levels identified in the
Factorial design
3) A set of Axial Points, which considers values that are below and above the median of Factorial
Design Points.
To use CCD, a controlled environment is required. A Production environment is an ideal fit for such a
controlled environment, because the Production environment is a setting where the reliable software and
robust hardware configuration are available for commercial daily operations of applications. Hence, it
becomes easier to vary the required set of parameters and realize the results. In addition, RTB projects
have historical data and can be easily tapped to be used for CCD analysis.
In a conventional CCD, a set of controlled experiments is performed with identified factors. However, in
this case, some re-computation was performed with the historical data, to arrive at the results of a
process. For this purpose, Data Mining was performed for each category of Optimization and the factors
were determined. Using the factors determined, the Quadratic expression of CCD is used to obtain the
Optimum goal point.
3. INITIATION
CIOs find it challenging to mandate the reduction of the RTB cost, as much as possible and invest the
savings benefit into strategic CTB projects. Increased complexities with RTB cannot be managed easily
using legacy management principles and skills. As a result, the Authors of this paper have chosen the
approach of CCD, which is a proven statistical model for Optimization.
For CCD analysis, the Authors chose an RTB project for a Retail Banking Customer. To start with, the
focus areas have to be set for the project. The basic requisite for an RTB project is the stability of
applications and reduction of tickets. Based on the applications’ performance in the past, complexity of
applications and the nature of work, the following two primary focus areas were determined:
 Ticket Optimization
 Cost Optimization
The following are the limitations of the current setting in the project:
 Cost budget is limited
 Number of FTE is constrained

The two Optimization areas were considered for investigation with CCD technique to obtain optimum
levels that can be set as a goal for the year 2013.
4. APPLYING CCD TO TICKET OPTIMIZATION
Being an RTB project, the priority of the project primarily lies on ticket reduction. Tickets are created,
whenever there are any Production failures (known as Abends or Abnormal End in Mainframe).
Historical data from 2012 was collected for tickets that occurred during 2012, and they were scrutinized
for factors contributing to them. There are three primary factors that affect the numbers of tickets (from
the year 2012):
 Number of Production failures(Abends) in a month
 Recovery time for each Production failure1
 Resolution time for each Production failure2
These factors were fed to the CCD technique, to get an optimum value of tickets that could be fixed with
current base resources in the year 2013. Number of resources available to provide break -fix installs and
baseline hours were almost the same for 2012 and 2013 and hence this factor was a constant. The
Authors used a standard CCD with five center points and alpha value of 1. The corresponding factor
properties were updated within the CCD model and each independent variable had three levels assigned
to it (– 1, 0 and +1), based on a predetermined range of tickets or time and categories.
Table 1: CCD Variable – Coded Level Matrix
Independent variables
Coded levels
-1 0 1
Top 5 categories 80 100 120
Time/abend (mins) 35 50 65
Time/resolution (days) 80 100 120
The study was carried out, according to the CCD, and the experimental points were used, based on data
points from 2012. The regression coefficients for the second order polynomial equations and results for
the linear, quadratic and interaction term are presented in Table 2.
1 – Recovery time of a ticket is the time taken to restore service after an incident has occurred. This may be a temporary solution
or work-around to restore the Service
2- Resolution time of a ticket time taken to permanently fix the ticket, so that reoccurrence will not happen again.

Table 2: Variation of Parameters for Optimum Point
S.NO
Top 5
categories
Time/abend
(mins)
Time/resolution
(days)
No of tickets
that can be
resolved
1 80 38 88 51.11
2 83 45 80 53.3
3 120 56 93 50.09
4 105 35 99 52.2
5 119 55 119 48.33
6 114 60 120 52.11
7 103 65 111 49.6
8 109 37 86 48
9 94 44 94 38.88
10 117 35 90 49.3
11 96 39 86 49.09
12 112 49 90 49.7
GRAND
TOTAL 1252
1156 591.71
The second order polynomial formula for CCD was used, to arrive at the number of tickets that can be
resolved.
y = β0 + β1x1 + β2 x2 + β12x1x2 + β11x12 + β22x2
2
The contour and quadratic formula result indicates that the maximum benefit occurs when an average
(591.71/1252) tickets can be reduced. As the average is 47.26, 50% ticket reduction is considered, as a
result from the Optimization. The predicted maximum achievable result will be 50% ticket reduction in
2013, out of the total tickets from 2012.

5. APPLYING CCD TO COST OPTIMISATION
The major factor for the increasing RTB spend is the total cost of applications. By effectively managing
this cost, a good savings can be achieved, which can be turned into a profit for the organization.
The four major factors that impact the Cost Optimization are as follows:
 Number of Production failures(Abends) in a month
 Recovery time for each Production failure
 Resolution time for each Production failure
 Effort per cost saving task
These factors were introduced to the CCD technique to obtain an optimum Cost savings that can be
obtained in the year 2013, with current resources being the same during 2012 and 2013. The
corresponding factor properties were updated within the CCD model and each independent variable had
3 levels assigned which were – 1, 0 and +1 based on a predetermined range and categories.
Table 3: CCD Variable – Coded Level Matrix
Independent variables Coded levels
-1 0 1
Number of Production Failures in
a month 80 100 120
Recovery Time per failure (hours) 35 50 65
Resolution Time per failure (days) 80 100 120
Effort per cost saving task 100 200 300
The study was carried out according to the central composite design and the experimental points were
used from project reports for 2012.The regression coefficients for the second order polynomial equations
and results for the linear, quadratic and interaction term are presented in below table.

Table 4: Variation of Parameters for Optimum Point
S. No
# of Prod
Failures /
month
Recovery
time / failure
(mins)
Resolution
Time / failure
(days)
Effort per
cost
saving
task
(hours)
Cost Savings (in
MM USD)
1 80 38 88 256 0.8
2 83 45 80 210.03 0.63
3 120 56 93 190 0.45
4 105 35 99 188 0.41
5 119 55 119 189.11 0.37
6 114 60 120 188.78 0.309
7 103 65 111 280 0.965
8 109 37 86 210 0.699
9 94 44 94 194 0.501
10 117 35 90 193.67 0.58
11 96 39 86 189 0.49
12 112 49 90 190.9 0.51
Grand
Total
6.714
Applying the second order polynomial equation to this data, the Cost Savings in MM USD was obtained:
y = β0 + β1x1 + β2 x2 + β12x1x2 + β11x12 + β22x2
2
The contour and quadratic formula result that the maximum Cost Savings obtained by the team could be
near 6.714 MM /12 = 0.56 MM. The determined achievable Savings Goal would be in this case as 0.5
Million USD in 2013.
6. SETTING UP THE GOALS FOR 2013
Based on the results from CCD Approach and after internal brainstorming within the teams, the following
goals were set for 2013:
 50% reduction in number of tickets from Year 2012
 0.5 Million USD Total Cost of Application Savings / Hard Dollar Savings

7. THE PATH TO OPTIMISATION
After the goals were identified, the next step was to identify a good management approach to achieve
them. Based on constraints on the problem, various structured frameworks are available for solving
business problems and effective process execution. The Authors of the paper have chosen the well-established
and accepted model for implementation of goals – CMMi framework. TCS being at CMMI
Level 5, both the goals were targeted to be implemented with CMMI Level 5 areas for Support and
Process Management.
7.1 The Ticket Optimization Exercise
For the Ticket Optimization exercise, the authors chose the Causal Analysis and Resolution (CAR)
Process area of CMMI Level 5 model. As per this process area, the root causes of historical tickets from
2012 wereanalyzed systematically using Pareto Chart. Once the Causal Analysis for the occurrence of
the tickets was determined, they were systematically addressed to prevent their reoccurrence.
7.1.1 Ticket Reduction using Causal Analysis and Resolution Process Area
Based on the CCD Model described in previous sections, the goal for 2013 was set as 50% Reduction in
the number of tickets from previous year. The Ticket Reduction Exercise was done in 2 phases as per the
Specific goals outlined in CMMI Level 5 CAR Process area as below:
Phase I : Measure and Perform Causal Analysis
In this phase, the historical data of 2012 tickets were analyzed and causal analysis was done for each of
them. After this, similar causes were grouped together and Pareto principle was used to prioritize the
deployment actions.
Figure 3: The Pareto Analysis for Root Cause for 2012 Tickets

7.1.2 The Results –Ticket Optimization
Once the prioritized causes were identified, the next step was to implement the action proposals to avoid
the occurrence. For this, the second specific goal statement of CMMI Level 5 CAR Process area was
followed:
Phase II :Prioritize and Deploy action proposals
For causal analysis and resolution tracking, a new tracking approach, Ticket Tracking Sheet, was
introduced to keep a track of all tickets and causal analysis for each ticket. On analyzing the tickets, it
was seen that Top 3 causes attribute to 80% of the tickets. By fixing the Top 3 Causes, more than 50%
reduction of tickets could be achieved.
Proactive approach was followed to mitigate all the top 3 root causes. Automated tools were introduced to
track frequently occurring problems. Supplementary automated monitoring systems were introduced to all
critical interfaces to avoid data issues and thereby reduce mass incidents. False alert Optimization was
done to make sure that tickets are created only when there is really an issue in the Applications. This also
reduced the manual intervention on delays and false notifications. Bottom line of this exercise was "Every
ticket should have a follow up or action item" thereby to nail down the root cause at the first occurrence of
an issue. Meticulous reviews were conducted when a change was installed in Production from the
Development team, to reduce the issues arising due to Integrated Releases.
Figure 4: Ticket Reduction – Actual Vs Target
Due to quicker Root Cause Analysis, number of tickets reduced leading to quicker resolution for the
tickets and thereby leading to tasks getting completed at a faster pace. All of this contributed to more time
available for working on Value Additions. This leads to a Cyclic Effect, due to which the applications were
optimized to the maximum.

Figure 6: The CYCLIC Effect
7.2 The Cost Optimization Exercise
7.2.1 Cost Optimization using Organizational Performance Management Process Area
The purpose of Organization Performance Management (OPM) Process Area of CMMI Level 5 is to
manage the organization’s performance and meet the business objectives in a proactive manner. This
process area follows a very well defined sequence of procedural steps that can be used to realize a goal
to improve the Organization’s performance. The specific goals of this process area and how they were
realized is given below:
Phase I : Identify the Scope and areas for Process Improvement
In this phase, the scope of the problem is defined and the SME’s from the teams defined which
applications have opportunities to start with for Cost Optimization.
Phase II : Review and Validation of Selected improvements
In this phase, the current state of the problem is reviewed, all the relevant information is studied and to
identify if there are any bottle-necks for implementation. During this phase, the Proposals are submitted to
the Performance Management team and reviewed with them to identify the cost saving opportunity. The
estimated savings for the task is identified in this stage.
Some of the opportunities identified and proposed to the Performance Management team are as follows:
 Deletion of unused datasets in Production regions
 Purging of Obsolete data in tables

 Replace Full updates with Delta Updates in Tables
 Usage of new Utilities and newer version of software to reduce the cost
 Program changes which will bring about Cost savings
Cost-Benefit analysis is done based on the benefits yielded by the improvement against the effort spent
on deployment. After this, the improvement is taken forward to the next phase.
Phase III : Verify, Execute and Evaluate the benefits
In this phase, the identified opportunities are prioritized and a draft roadmap is outlined. The required
changes to enable and sustain the improvements are done in this phase. The implementations of
identified opportunities are done and the actual cost savings are calculated and submitted to the
Customer for feedback. The feedback is collected and the cost savings are approved by the Customer.
7.2.2 The Results
As a result of the steps outlined by the Organization Performance Management Process Area of CMMI
Level 5, teams worked on different Process Improvement tasks in the year 2013. These tasks have
yielded good results and yielded substantial savings to the Customer.
Figure 8: The Cost Savings Achieved in 2013

8. CONCLUSION
Thus the CCD Optimization Technique is a good fit for Software Optimization Estimation also. This
technique can be used for any RTB project which has historic data. Using the CCD, realistic goals can be
determined. Skillful planning and meticulous approach to the determined goals can result in substantial
benefits and can reduce the Operational cost. This will balance the RTB and CTB cost and the
Organization can spend more in Strategic CTB projects.
9. ACKNOWLEDGEMENTS
We would like to sincerely thank Kannan Balamurugan (Delivery Head) for providing valuable inputs
about Central Composite Design and for reviewing the paper and providing constructive feedback which
helped us to keep progressing. Our thanks to Sivashankar Hariharan (Delivery Manager), for providing
his inputs on Management techniques in Process Improvement and also initiating the idea for the paper.
We would also like to thank Rejo Reghunadh(Senior Business Analyst) for his contribution towards the
project data analytics.
10. REFERENCES
[1] Myers, Raymond H. Response Surface Methodology. Boston: Allyn and Bacon, Inc., 1971
[2] Koch, Richard – The 80/20 Principle – The Secret to Success by achieving more with less, 1998
[3] N.S. Sreenivasan, V.Narayana – Continual Improvement Process, 2005
[4] Carnegie Mellon Software Engineering Institute, “The CMMI Version 1.2 Overview presentation 2007.
http://www.sei.cmu.edu/library/assets/cmmi-overview071.pdf
[5] CMMI for Development, “CMMI-DEV V1.3,” Technical Report, Software Engineering Institute,
Pittsburgh, 2010.

Presentation by vignesh swamidurai

More Related Content

What's hot (20)

Similar to Presentation by vignesh swamidurai (20)

More from PMI_IREP_TP (9)

Presentation by vignesh swamidurai