Confidence Intervals and Random Sampling with >99.95% Accuracy

[simms]

Every week, there is an order for about 100,000 boxes of product.
We currently audit every box to ensure it matches what is ordered. We check each label and match it to the order.

I am hoping to replace this with random sampling to achieve with 99.95% confidence that the 100,000 boxes of product are accurate.

For example, it might be that we need to check X% of the order each week. If there are Y% defects, then the entire lot must be checked manually. Otherwise, if there are 0 mismatches in the X%, the lot passes with 99.95% confidence.

How can I determine this?

 

[Hacp]

Hire me and I'll help you! Or ask any one of your engineers. You're probably going to need to do something with binomial distribution. You'll need to analyze the probability of falsely labeling a box from previous data.

 

[Hacp]

I asked around and it seems excel can give you a function to estimate the stuff.
BINOMDIST.

 

[Hacp]

This is an easy tool that can help you out.
http://src.alionscience.com/toolbox/oneshotcalc.htm

 

[simms]

Thanks very much Hacp, I'll check this out. Thanks for pointing me in the right direction.

 

[warlicrox1394]

whoa whoa, why binomial distribution i suggest u use Poisson distribution as all binomial distributions tend to Poisson for large numbers (100,000)

and if u want 99.95% statistical quality control its best to opt for normal distribution (check out 6 sigma limits)