How to determine the right sample size for transactional quality in BPO/KPO industry. It should be different from the way we do for our GB/BB projects for improvement.

There is no set rule to pick the sample size in a call centre environment. You can choose the sample size as per the number of the calls, however 10% sampling of a particular data set is always good.

Thanks,

Ajeet 

Sample size depends on 3 things:

1- Size of population, 2- Confidence level and 3- Margin of Error.

In addition to what Inderpal just mentioned above, sample size also depends on the kind of hypothesis test you plan to carry out. In case you are working with discrete data - sat vs dissat, you need larger sample sizes. If you are working with continuous data like handling time or TAT, you can manage with a smaller sample sizes. 

Some of you may be wondering what sample size has to do with hypothesis test. Reply to this post if you wish to know more. 

Hi Vishwadeep,

 

Question 1)

 

Question on Sampling. 

There is a client with us whose Monthly volume (Chats) are around 40000 and we have a practice of analyzing the DSAT's and categorizing them as per the guidelines. Because the volume is high, it is apparent that we get more number of DSAT's considering the factor that multiple agents handle the chats. 

Now the challenge is that we cannot analyze all the DSAT's (since the number will be high) and was thinking of doing the sampling.

 

I agree to the fact that we need the following for Sampling.

1- Size of population, 2- Confidence level and 3- Margin of Error.

 

How to go about if my 

1)Size of population - 2000 (DSAT's)

2) Confidence level - 95%

3) Margin of Error - 5%

 

I downloaded the Sampling size calculator attached and and could not understand "p" though the above 3 points were very clear. Please help me as to how to proceed. Also, in some of the calculators I downloaded, it was asking for known Standard deviation as well. 

 

Question 2) A brief explanation or an article as to why the sampling size would be different for Continuous and discrete date will be different and what sample size has to do with hypothesis test?

 

Thanks in advance,

 

Regards,

Pavan Samala

Dear Pavan, 

 

Here are some answers. 

 

1. Continuous data requires less number of samples as compared to discrete data in a similar sitiuation. 

 

2. Sample size depends on the hypothesis that you are trying to test.

 

Best Regards, 

VK

What must be a right precision in TQ while sampling at HPD of less than 5%...