An interesting dilemma to ponder over…
I have recently been employed to look at a component facility and their approach towards risk assessments and validations. The volume going out every day is quite high and a batch size of 10 to 20,000 is very normal, with 2/3 batches a day coming off a machine. Inspection is done to an agreed AQL with the customer. However, due to the high risk nature of some of the components, a lot of customers are now 100% inspecting the components before use (this is a gap in itself, we cannot 100% inspect given the volume, it is not practical). This is resulting in customer complaints which on first glance seem to be quite high (depending on the customer and the their inspection criteria), however it may be one or more in 20,000. This according to our procedure means we have a low occurrence (1 in 25,000) however we are still getting a complaint in every batch sent out (which to me indicates there is a problem in the process), but the approach here is given the volume of product we produce, we have a low occurrence, and therefore does not need additional controls. However I am beginning to think that we need to look from a different angle and instead of units, maybe look at batches instead of individual units e.g. we had the same defect over 5 different batches, even if only 2 pouches affected from each batch this is a high occurrence. So instead of saying we had an occurrence rate of 10 in 100,000 units (very low occurrence rate), we had the defect in every batch (higher occurrence rate) and requires action. This discrepancy between our inspections and the customers is leading to a sharp increase in customer complaints, leading to irate customers, however given the volume of individual units manufactured, is it really a drop in the ocean??
Am I been overcautious? Would people accept the defective product given the volume that is flying out on a daily basis? Or do we need to increase our controls??? We are still (probably) within our 32 defects per million (CPK 1.33)
The first question I would ask is: What is the severity of the defect? If it is low and customers are only inconvenienced then you may be overly cautious (unless it drastically drives down marketability). If it is a more severe defect that could harm a customer, or worse, then more in-process inspection may be needed. Perform an assessment of failure modes to determine if the product of severity, occurrence, and detectability (RPN) is at an acceptable level.
In order to determine the correct occurrence rate to use, root cause analysis should be used to determine if the process variation affects the product units individually or the entire batch. You may need to perform studies on intra-lot variation as well as lot to lot variation in order to determine the impact of the variation.
I think that another option of this problem may be as a result of non- suitable testing method.
I suggest to conduct test method validation to all tests that used during production.
I propose to use “gauge r&r” technic for quantitive testing and “attribute agreement analyses” technic for pass/fail testing.
I think it is time to train your core team to identify various defects quickly. As i read thru your blog, you described it leaving no loop holes.
Identification of a defect on a packing line looks very easy but very critical to quality.
Every one there on the visual observation and verification team should be trained based on their distance from product observation, speed of packing line, speed of conveyor, light intensities in that particular room and last but not least his eye sight.
Training and evaluation is very critical. I think it will create a lot of effect on your CPK values.
Thanks for those great replies. Interesting, I have just added an additional inspection method to one of our lines, which is a full automated 100% inspection step, and so far this has noted a defect with a high severity on approx. 5% of our product (a batch of 2000 had approx. fall out rate of 90), which is spread across the batch.We are currently running gauge R&R’s at the moment, but I have noted from our gauge R&R’s that operator training is an issue again, and again. I am noting though that people sometimes don’t understand/are afraid of risk assessments, and believe by identifying an area as “high risk” it is a failure of sorts, instead of understanding that risks can be mitigated and the process as a whole improved. And as such resistance to identifying an area as “high risk” is very strong. We even had a risk assessment, where we were introducing a major new fully customized software control system, and people tried to say (before it was validated!) that it was low risk, to avoid high RPN’s, and therefore associated actions.