From: Stan Hilliard
Date: 26 Nov 2003
Time: 10:49:54
Greetings Khanna,
I assume that each individual call is either defective or nondefective. In sampling Customer Service Calls it is important to give attention to the "subgroup" or "lot" that a sample represents. For example, do you want the sample to represent a day, an operator-day, etc.
Here are two approaches to developing a plan, depending on your goal for the plan:
www.samplingplans.com/software.htm
1) Estimation only: Choose a margin of error. Calculate the sample size from the margin of error and the approximate expected failure rate. Use the software program: "Audit Sample Planner" (ASP). For any actual sample result, ASP will calculate a confidence limit report.
2) Estimation plus decision rule: Choose the producer's point and the consumer's point. Calculate both the sample size and decision rule (Ac). Use the software program: "Sampling Plans for Attributes" (TP105) to do this calculation. The methodology is described at:
www.samplingplans.com/modern3.htm
The simplest kind of sampling plan would be a fixed sample size attribute plan. This is also the kind of plan that requires the largest sample size.
A sequential type attribute plan is more efficient in that the sample size is smaller. There are examples of each type of plan using TP105 -- follow the links.