Don't Say "Zero Defects" Unless You Really Mean It

[Bev D]

When you get to a very low defect rate (through robust product and process design, the things that are left are probably error induced as well as ‘random’ physics problems (like small debris that might flake off equipment). Inspection surely won’t help and error proofing for these rare (true ppm levels) becomes costly. This is not an excuse for not having zero defects as an aspirational target, but it is a cautionary note that too often we think of operator error in these cases.

 

[Mike S.]

When you get to a very low defect rate (through robust product and process design, the things that are left are probably error induced as well as ‘random’ physics problems (like small debris that might flake off equipment). Inspection surely won’t help and error proofing for these rare (true ppm levels) becomes costly. This is not an excuse for not having zero defects as an aspirational target, but it is a cautionary note that too often we think of operator error in these cases.

May we all have these kinds of problems....

 

[Jen Kirley]

True enough, except in the specific cases plaguing me at the moment, at least two are manual assembly jobs where the operator just forgot a step. One of the rejects went through 100% inspection prior to shipping, and the customer still found rejects...and so did we when we sorted our inventory again.

This is unfortunate on a couple of levels.
1) Being unaware of your particular type of assembly, I wonder if it's possible to kit the subcomponents (if there's a screw left over then the item is missing a screw) or address inattention by cycling tasks.
2) 100% inspection is notoriously unreliable. I wonder if it's practical for the next step to do a quick check-over, or again to cycle the inspector's tasks in order to address monotony that can affect even the best of inspectors.

Of course I don't know how costly these items are, or how much it cost you to sort your inventory again - in addition to fielding the customer complaint (was there one?) and replacing the item. The bath tub model of quality costs applies.

 

[toniriazor]

When you get to a very low defect rate (through robust product and process design, the things that are left are probably error induced as well as ‘random’ physics problems (like small debris that might flake off equipment). Inspection surely won’t help and error proofing for these rare (true ppm levels) becomes costly. This is not an excuse for not having zero defects as an aspirational target, but it is a cautionary note that too often we think of operator error in these cases.

Despite being physics problems , they still could be controlled to a certain extent I guess? Like, if the process owner notices that after a certain period of time from the machine debris start falling off , there should be a logical maintenance plan to prevent this and never get to this point. Recently I had a case where occasionally bolts (manual operation, not by a robot) did not get screwed in all the way to lock the component. One of the machines was not maintained at all and due to this a several small debris got inside the bed where the bolt goes. Or you mean there are operations where you could just not avoid the physics?

 

[Bev D]

All physics problems can be controlled just like all human errors can be controlled if yo are willing to spend the money (at the very low occurence rate). There is no such thing as a truly random event. Random means without ‘apparent’ cause, not without cause…if there is a cause it can be controlled.

 

[leftoverture]

This is unfortunate on a couple of levels.
1) Being unaware of your particular type of assembly, I wonder if it's possible to kit the subcomponents (if there's a screw left over then the item is missing a screw) or address inattention by cycling tasks.
2) 100% inspection is notoriously unreliable. I wonder if it's practical for the next step to do a quick check-over, or again to cycle the inspector's tasks in order to address monotony that can affect even the best of inspectors.

Of course I don't know how costly these items are, or how much it cost you to sort your inventory again - in addition to fielding the customer complaint (was there one?) and replacing the item. The bath tub model of quality costs applies.

We have several products for which we are kitting the subcomponents, and wouldn't you know it, we even found error occurrences in the kitting process! The types of errors we see are purely human errors; basically forgetting a step. We have tried many things, like one piece flow, improved workstation layouts, kitting subcomponents, training certified operators...all the common things people try to remedy these things. But, as we all know, any process that relies on human beings is subject to human error. Automation would be the key, but most of the time the economics don't support the business case.

 

[leftoverture]

So, one of my customers just notified me of a defect and asked if we need to sort all inventory (nice when they ask and don't demand). In checking our records, it was the first recorded defect for this process in 16 months. That puts our DPPM for this process at 3.5. Oh how I wish all our processes were that close to zero defects!