# Sampling Strategy – Practicality Matters

Six Sigma – iSixSigma › Forums › General Forums › Methodology › Sampling Strategy – Practicality Matters

This topic contains 1 reply, has 1 voice, and was last updated by Strayer 8 months ago.

- AuthorPosts
- April 15, 2018 at 12:50 am #55977

SaraDear Six Sigma Experts,

I am in the look out of real solution for solving our real problem.

Background:

We are peanut shellers and processors and supply them to snackfood factories. One of the defect that we are to look for is aflatoxin, which is a microbiological poisonous content in peanut. We do have a laser machine to identify but it is not 100% foolproof. Therefore, after processing the peanut thru a series of machine, we do sampling and then a composite of the samples to take them tot he lab to check the aflatoxin results.Our current sampling method:

1. 1 Lot = 20 metric tons and that is = 400 bags of 50 kilo each.

2. our sampling is 100%, therefore we draw 500 grams of peanut from each 50 kilo bags.

3. we will have a total of 200 kilo of peanut from the 400 bag sampled.

4. we then take randomly 20 kilo from this 200 kilo of peanut because we can only crush 20 kilo to test for aflatoxin. So we crush 20 kilo to make a batter and then take a small quantity of the batter to test aflatoxin results

5. If we hire outside laboratory, they only check 100 bags (25% sampling) and draw 20 kilo only from a total of 100 bags.Pain area: Sampling is probably not catching enough representation and when the peanuts reaches their destinations, they tend to fail. We prefer it to fail in our factory itself so that we can fix them.

Questions:

1. is the sampling size of 500 grams for every 50 kg bag is okay?

2. after making a single composite of 200 kilo from the 400 bags, we taking random 20 kilo for crushing is right> is there a better method?

3. is there any other process or strategy we must follow to truly represent the population and sample?April 15, 2018 at 7:26 pm #202473Have you done GR&R on your laser machine? How close is it to 100%? Do you get the same result when you retest the same sample. Does it lean toward false positives or false negatives? Can it be recalibrated?

- AuthorPosts

You must be logged in to reply to this topic.