Natural process limits for quality metrics. Do you use them?

[qawebguy]

I've been giving this topic some thought ever since I read Wheeler's Understanding Variation book in which he talks about all processes having natural variation and knowing when to act on special causes and not on common causes.

My question is, do any of you use process limits with your quality metrics? For instance, I'm contemplating using limits for scrap indicators. I realize this is going to be a big hurdle for management to jump over when the numbers go up from a previous month, but like all processes it's just part of the natural variation built into the process.

It's a waste of time trying to figure out why a number went up or down while it's still within the limits when it's just part of the process.

I'm not saying that you just sit on your duff and do nothing about it, but when your numbers are within the limits your only choice is to improve the process to gain improvement.

I'd like to hear from anyone using limits with their indicators and how you've overcome what I think will be an issue dealing with management when numbers increase and there's no corrective action to respond to the change.

[Marc]

Mostly what I have seen has been a consideration of process limits with regard to processing specific items. I have been in situations where a job was bid and then equipment assigned which couldn't keep within specification because the process wasn't capable of holding the tolerance. Is this what you're talking about?

[e006823]

qawebguy,


I worked for an ASIC (Mil-PRF-38535) manufacturer. The devices were tested over at room/hot/cold temps. We knew from past experience (control chart of lot yield) that our average yield was xx% and our LCL was yy%. In our nonconforming material procedure we based our actions on these limits:

Lot fails PDA or lot yield < YY%:
production control system comment/hold; MRB review and approval on an NCR.

Lot yield < XX% and > YY%:
production control system comment/hold; Engineer review and approval on the affected production traveler.

In addition on a quarterly basis we reported the following metrics to corporate and the managment review:

Supplier Quality: PPM

Mfg. Quality: DPMO

Design Quality: DPMO

Delivery: % On Time

Delivery Quality (Escapes): PPM

Custumers Evaluation: % rated as preferred

These metrics also were plotted on a control chart and action taken when an out of control point occurred.

“I'd like to hear from anyone using limits with their indicators and how you've overcome what I think will be an issue dealing with management when numbers increase and there's no corrective action to respond to the change.”

When corporate rolled out 6 Sigma managers at every site were given 6 Sigma lite. Wheeler's Understanding Variation book was part of that training. So we had very little resistance to the idea of not taking action at the a point below target as long as it was within the control limits. That is not to say we did not try to improve the processes that were in control in order to limit variation.

[Steve Prevette]

Originally Posted by qawebguy

My question is, do any of you use process limits with your quality metrics? For instance, I'm contemplating using limits for scrap indicators. I realize this is going to be a big hurdle for management to jump over when the numbers go up from a previous month, but like all processes it's just part of the natural variation built into the process.

It's a waste of time trying to figure out why a number went up or down while it's still within the limits when it's just part of the process.

I'd like to hear from anyone using limits with their indicators and how you've overcome what I think will be an issue dealing with management when numbers increase and there's no corrective action to respond to the change.

It ain't easy. But yes, after 12 years at it, I can say that Fluor Hanford uses the process limits (ie UCL and LCL in a SPC chart) on its quality and safety metrics. It has helped management to focus its questions. The key becomes in identifying which processes are:

1. Trending in the wrong direction (per SPC, "out of control" in the non-improving direction). Corrective Action is needed.

2. Stable at a level that is an unacceptable risk or outside of customer specifications (common cause variation, work on the system to improve).

3. Trending in the right direction (again per SPC, "out of control" in the improving direction). Reinforce whatever is happening.

4. Stable at an acceptable level, within customer specifications. If no processes fall into category 2, perhaps some of these need to be looked at again for continual improvement. Otherwise, no action required.