Hi All,

I work for an ITES in a BPO environment for one of the Tech processes.

I plan to run a project on C-SAT. following are the challenges:

1. I have historical data of 12 months.

2. C-SAT is sent to customers on random basis, so its just 3-5% of the

total volume that we support.

3. My understanding is that this should be treated as "Discrete data".

Please correct me if am wrong.

4. Customer support process involves support from other outtask teams

as well.

Thanks all, B Panday

The above discussion was started by B Panday on YAHOO GROUPS

 

 

Past responses to the above discussion is posted below.

PASTED BELOW FROM YAHOO GROUPS

Re: [benchmarksixsigma] Suggestions invited hey, Nice to see you work on C-SAT . Would just wana let you know my experience and understanding C-SAT is a very huge ( vast area ) for a green belt project as its not only one department involved so it is a BB project . The sample size you are talking about is very less 3 -4 % so a true RCA cannot be done on it . I would suggest you to pick up a factor that impacts the most on C-SAT and start a project on that . Others please share your thoughts.

Regards

rohit bhardwaj

Re: Suggestions invited

Hi

You can calculate the sigma level for C-Sat's in the following manner:

Defects (D): # of dissats received

Unit (U): # of total surveys replied to by the customer (Sats plus

dissats)

Opportunity for error: 1 (Either the survey will be a sat or a dissat)

Total Opportunity: Opportunity x Unit

Defects per opportunity(DPO): Defects / Total Opportunity

DPMO: DPO x 1 million

Sigma Level (z): Find corresponding sigma level from the z tables

Let me know if you require any further explanation in regards to

running a six sigma project for CSI.

Regards

Shahzad

Re: [benchmarksixsigma] Suggestions invited

if 3-5% sample is gppd enough( more than 30 data points) but make sure that customers are selected on ramdom basisto avoid skewness in the data. you can treat this data as cotinuous and move forward.

Thanks Manoj Singh

Re: Suggestions invited

You can determine the number of samples you need to do your analysis.

In order to do this, you will need to determine the following:

a) What is your overall Population Size (N).

What level of confidence do you want in your analysis. Typically,

this is 95%. This corresponds to Z = 1.96

c) What is your margin of error in your analysis. For example, if

your margin of error (ME) = 4% and your reported CSAT = 80%, then you

are 95% sure that the true value of CSAT lies between 76% and 84%.

Based on the above inputs, the number of samples you would need is:

SS = [(Z^2*0.5*0.5) + ME^2]/[ME^2 + Z^2*0.5*0.5/N]

If N is very large, then the last term in the bracket will be close

to zero. For the above example, the required sample size would be:

SS = [(1.96^2*0.5*0.5) + 0.04^2]/[0.04^2] = 601.

So, you would need at least 600 samples. As noted earlier, make sure

that the samples are random and truly unbiased.

Hope this helps,

SJ.

Hi,

This is further to Shahazad's reply. We can also consider each parameter in the survey.

Unit - Each survey replied

Opportunity: 1 - Each parameter answered in each survey

DPU - No. of Defective units/Total no. of Surveys

Total Opportunities: No.of parameters in each survey * Total No.of Surveys

DPO - No.of parameters scored less than agreed rating/Total opportunites

This approach will give more specific area to concentrate for improvements.

Hope this helps...