Sampling protocol/frequency
Hello all, I'm hoping to get some ideas on a sampling program for finished products in our facility.
Some background: we are a small R&D facility that also runs production. We specialize in emulsions for cosmetics, food, and dietary supplement ingredients. Our production is dependent on customer orders and can vary from 40 lbs to 18,000 lbs. (We have small scale and large scale processing equipment). We primarily package in 1 of 3 packaging mediums/sizes: 275 gal tote (2300 lbs), 55 gal drum (450 lbs) or a 40 lb bag. All packaging are liners within cardboard box/drum/tote.
My dilemma is determining a robust sampling plan to ensure that we are evaluating representative samples of the production. Our batches are approx 2000 lbs, so a tote may contain two separate batches, while a batch may fill several drums and 40lb bags. Do we sample per batch, per container, beginning/middle/end? We currently sample the larger productions by package...tote sample, drum sample. The smaller productions that are packaged in 40 lb bags are sampled at beginning, middle, end.
Do you all think that this is sufficient, or do you have recommendations?
Also - our analyses are as follows: TPC, YM, Coliforms are performed in house, pathogen testing sent to an outside lab. Internal analytical performed.
Thank you!
That's a tough call because there are so many potential variables, but I think if I were in your shoes, I would sample at least beginning, end, and at some set, documented frequency between (Middle? Every hour? Two hours?) for each batch. That way, you have representative samples of the whole batch, regardless of the packaging it goes in to. Of course, all of this is assuming that you are tracing everything by batch...correct? If not (for example, you have multiple batches in one lot, and trace by lot), that might affect how you sample.
I was previously the quality supervisor in an SQF certified company where we made agglomerated powder products (beverage, granulated sweeteners, etc.) - we would produce batches, which we would sample B, M and E. Most of the time there were multiple batches per lot, so we would create lot composites from those batch samples. Depending on customer and product specifications, we would create lot composites for micros, organoleptic, retained samples, etc. Depending on how you trace, maybe that will help? I'm not saying it's the best way for everyone, and it may not work for some people, but it worked for us.
Hi svnh.bell,
I assume the immediate objective is to comply with a customer's requirements. The situation may be different if the results are for internal use only.
Similarly to the previous post IMO you must initially decide on certain items to enablle a statistically meaningful sampling plan such as -
(1) How do you define yr finished LOT ? For example how do you determine a "LOT" for labelling/coding purposes ?
(2) what is the (hopefully mutually agreed) product specification for which a "lot" will be subsequently acccepted/rejected ? eg nmMc format ?
Knowing 1, the details of a sampling procedure to evaluate compliance with (2) can then be examined.
Note that "zero/non-zero tolerant" micro. criteria may require rather different numerical approaches. eg see this similar current thread -