Hi,

I am a IT professional and from my experience and by looking at various case studies, it looks like Six Sigma tools like QFD, FMEA, Pareto Chart, Ishikawa diagram are the few important tools, which are mainly used across all six sigma initiatives in software industry

I have not seen any case study in software industry, where advanced Six Sigma Tools like DOE or Regression analysis etc has been used or is really required..

Am I missing something here ? Is there any such example where advanced Six Sigma tools also come into focus.

Most of the time, we have abundant data (incident log, failure log, duration log etc), which is analysed via Pareto and FMEA tools and finally Causal analysis is done and maybe this will conclude into some Process Guidelines or Automation as the end result.

Please share your thoughts


Thanks

SG

Dear Sandeep

Your knowledge of Six Sigma is undoubtedly accurate, but somewhat incomplete. While tools like FMEA, DoE and Pareto Analysis are used to treat with pain areas or unknown problems, there are methods used to identify them also. VoC and VoP are two methods which are first used to identify areas demanding Six Sigma application. The data used for Six Sigma has to be real-time and credible because the ultimate aim of Six Sigma is to bring real improvements and not merely showing it on papers. To know more about application of Six Sigma in IT, click here.

All the methods you and I have mentioned are covered in the Green Belt Training. If you are interested to know more, kindly call at our toll free number: 1800-102-3003

Warm regards

AbhishekM

u can use gage R& R also,

c&E matrices, run chart ,scatter chart ( as type of data  attribute or continous). regration anaylises & so many tool as per problem we can use them

Hi Sandeep, 

 

You might find an interesting usage of Experimental Design while creating test cases. The objective is to create a combination of test cases that tests all functionalities of a module (with the aim of reducing testing time without losing comprehensiveness.)  

Here are a couple of use cases for Design of Experiments.

 

1. Suppose you have a software development team that execute projects of varying complexity using different software design approaches, with different team sizes, and the primary measure of project success is the project cost. The response is, in this case, project cost. You can use DOE to run pilot projects of varying complexity, design approach and team size (three factors). If you have two levels for each factor (more complex, less complex; two competing approaches, and a team size of say 2 versus another team size of 4), you could conduct 2^3 = 8 experiments for a full factorial DOE. Such an approach would be valuable when setting up the optimal project execution methodology for different projects from a client, or from different clients.
   
2. Let's say you are building a cluster for large scale computing, and want to understand how many processors to use, how much RAM to buy, how many CPUs per motherboard, etc. If you have a simulator which helps compete Gigaflops (computational power), you could plug in these factors at different levels, and conduct full / fractional factorial experiments. You can use response surface analysis (RSA) to determine the optimal settings in terms of cost, and computational power.
   

Essentially, wherever the performance of a system has to be optimized - and this could be a system of networked computers, or a system of people executing a project - you can use DOE.

 

Here are a couple of examples of Regression Analysis:

 

1. Suppose you want to understand the relationship between project duration and project cost in IT projects. These are two continuous variables, and you can collect past data for project duration (or manhours estimated) and project cost, and understand how this varies from client to client, or from region to region.
   
2. Suppose you want to understand revenue per employee and its correlation with cost per employee. You could plot the revenues for various employees by project on one axis, and the costs from respective employee GL codes on another axis. This would allow you to assess which employees are adding more value to your organization, as a trend line.
   

Such studies can be done in IT specific projects for tools like ANOVA, Hypothesis tests and other core statistical analyses. 

 

The key thing to remember in all these cases, is the fact that you require enough reliable data from each business situation, and that the costs of collecting additional data should be understood and accounted for in different business situations... since all data collection comes at a cost to the organization.