2 replies to this topic

[earthbornstew]

Hi Everyone,

 

This will seem like a simple question, but, should micro results on a COA be reported as an average of results, or the max. value found within the batch / section of batch received? I ask as I've seen both be used by different suppliers.

 

My own thought is that micro results on a batch COA should be the max. found, so that if a customer retests the product then they should find results either below or equal to those stated on the COA. If part of the product is found to exceed specification, then that can be separated into a different grade, and the max value of the remaining is valid for the initial grade.

 

E.g. if 100Mts of stock had results of 100, 300, 400, 600, 1000, and the spec was max 800. Then the stock covered by the first 4 results would go to grade 1, and the remaining volume goes to grade 2. The COA for the grade 1 product will report a max 600 result, and the Grade 2 will report 1000.

 

Plus, if average is used, its very easy to hide a spike in results.

 

 

If there are any guidance notes or text which refer to this that would be great to know.

[Irishlass105]

Hiya,

 

We work on a value found for that batch system. Each batch tested gets its own results and when the lab sends our results back to us we just confirm them against our target or acceptable range. As you say, if it is averaged it is hard to see any issues as it may not show. This can work depending on your product or what you are testing for. If you have an issue with results averaged how would you know where the issue lies? This is basically standard for the lab we use, who are UKAS accredited and are registered for testing for the like of M&S and Tesco.

 

IL

[Charles.C]

HI earthbornstew,

 

Yr query essentially relates to sampling/analysis of product lots and deciding whether the lot is acceptable or not. Meaningful answers rely on Statistics as far as possible.

 

Unfortunately any hope that micro.(count) data has substantial accuracy is usually unfounded. I have never encountered the use of a grading system such as you mention but this may also relate to yr particular situation, what you are doing with the unknown material and the latter's own microbial characteristics.

 

Tthe official answer for determining the overall micro. acceptability of a lot is usually via the nmMc system (with n typically 5). This methodology implicitly acknowledges the well known lack of accuracy in microbiological measurements. I expect  you are already familiar with this approach.

 

In contrast a typical statistical answer would be for a supplier to provide an average figure with a confidence interval. The latter will probably be "wide". Acceptability would require both purchaser/seller using similar procedures/criteria.

 

In practice, methods used by officials, suppliers and receivers are often  "ad hoc" so the "result" for a given lot may be nearer to a lottery.

 

This topic has generated a multitude of textbooks/publications but you may require some familiarity with sampling/analytical procedures to understand the details. The theory behind the above nmMc system is not so simple but its implementation is.

 

If you would like to see a thread here on this topic, can try the following and particularly the sub-links in Post 2 -

 

http://www.ifsqn.com...-sampling-plan/

 

PS - welcome to the Forum !