Sampling Plan - The process involves verification of employment

[Rakhi]

Hi All,

This is the first time am posting a message. I seek guidance on Sampling plan for one of our process.

The process involves verification of employment through ‘work number’ site. If the information is not available on the site then we call up employers via phone and confirm employment details.

I need to develop sampling plan for Monitoring. The team volumes per day are roughly around 1000. Team size is 6 and we do not have recording facility for the calls made. The auditor has to barge in using Y cord to monitor a call. Client has specified 10% monitoring of the total volumes every day.

In such a scenario what kind of sampling method is more apt. How do we ensure that random sample is selected?

Hope to get some assistance.

Rgds,
Rakhi

 

[gjhaddock]

Hi Rakhi,

The second part of your question may be easy. If each team consists of six members, then number them from one to six and roll a die to determine who gets audited next. If you have several teams of six members each, just make it a two stage randomization.

As to the sampling plan, assuming that you have a set of objective criteria to classify the call as either pass or fail, then you can use an attributes sampling plan, such as ANSI/AQC Z1.4.

For a lot size of 1,000, the normal (level II) inspection level calls for a sample size of 80, which is less than your customer requirement. You can get around that by using tightened inspection (level III) which calls for a sample size of 125 (but now you would be doing 25% more sampling than your contract requires). Maybe you could go back to your customer and get him to accept a known risk plan vs. an arbitrary 10% sample with unquantified risk)

To continue, here are a few Accept/Reject numbers for the level III plan for what may be reasonable AQLs. Ultimately it is you and your customer who need to agree on producer and consumer risk. Recall that the AQL is defined as the maximum percentage of failing calls that can be considered acceptable as a process average for the purpose of sampling inspection

(Note that these numbers are based on the entire lot size of 1,000 calls.)

AQL Accept Reject
10% 21 22
6.5% 14 15
4.0% 10 11
2.5% 7 8

I hope this helps

Jerry Haddock, P.E.

 

[Wes Bucey]

The process involves verification of employment through ‘work number’ site. If the information is not available on the site then we call up employers via phone and confirm employment details.

I need to develop sampling plan for Monitoring. The team volumes per day are roughly around 1000. Team size is 6 and we do not have recording facility for the calls made.
. . .
Client has specified 10% monitoring of the total volumes every day.
Rgds,
Rakhi

Welcome to the world of posting, Rakhi!

I'm stunned by the numbers you cite.
What kind of quality can anyone expect when each team member makes an average of 3 minutes per call including all the folderol of getting a human[rounded]?

1000 calls per day divided by 6 team members
= 166 calls / day / team member.
8 hours (no breaks) times 60 minutes
= 480 minutes divided by 166 calls [= 2.89, round to 3]
= 3 minutes/call [rounded] including dialing/transfers/ringing
ergo, if you let it ring 7 times, you used one of your minutes

If you call most companies I deal with, getting through to a human versus machine is a 1 in 3 bet. If you get the human, it's another 1 in 3 bet he/she can give an answer in that call. If you catch ME with an inquiry about someone's employment, I answer, "I'm sorry. Privacy considerations prevent us from acknowledging or denying whether someone is or has been employed by us to ANYONE without a signed authorization from the subject person."

Thanks to folks smarter than I, I have revamped the calculation - very embarrassing for me!

 

[gjhaddock]

Wes,

In your calculation, in the 5th line of the quote, it should be 3 minutes/call.

Jerry Haddock P.E.

 

[Wes Bucey]

Wes,

In your calculation, in the 5th line of the quote, it should be 3 minutes/call.

Jerry Haddock P.E.

Absolutely! That's why I buck all my hard questions to Steve Prevette - I'll change it. Still not much time to do a meaningful Q&A.

 

[Sam]

Good points Wes. I was "assuming" Rakhi only wanted a plan to sample the 1000 calls utilizing the 6 members of the team.

 

[Rakhi]

Hi Everyone!
Thanks for the responses. Let me explain.
Although arriving with a sampling plan for only outbound calls is O.K, My major problem is how to address randomisation of Data (WORK NO) and calls simultaneously when a data verification may or might not result in a call being made.
Hope this makes sense...incase anyone needs more explanation plz let me know,
wating for response,
Best Regards,
Rakhi

 

[gjhaddock]

Sampling Plan

Hello Rahki,

I suggest that it might be a good idea to describe your present process flow, including sampling audit, in a step-by-step itemization. That will minimize ambiguity and hopefully lead to some actionable recommendations for you.

Regards,

Jerry Haddock