Variation Study on Mined Material - What do you call it (not a control chart)

shellig

Starting to get Involved
We have a customer who wants to have "control charts" on the trace metals content on our raw material that is mined from the ground and, other than grinding, has no treatment or processes done to affect the value.

I am trying to explain why a chart of the variable over time is not a "control chart" as we have no way of controlling the process.

Of course they would like specifications (based on "control limits") on those variables as well. We can set that up based on historical values BUT tomorrow, our raw materials supplier could mine a different area with completely different results.

Any thoughts?
 

Bev D

Heretical Statistician
Leader
Super Moderator
Well a control chart doesn't mean YOU have control. 'Control' applies to the state of the process...the control chart assesses how much of the variation over time is stable (the variation seen is simply random variation from a homogenous process stream) or not. Stable = "In-Control"

So the Customer is correct in calling the chart they are requesting a control chart...


However, specifications do not come from the control limits - they come from what your Customer needs and not from what the process was able to deliver at some random point in time.
 
M

Matt33

I completely agree with Bev.

If it helps you, call it a Behavior Chart as Donald Wheeler does.
 

shellig

Starting to get Involved
Thanks for your feedback, everyone! I wanted to differentiate it from other charts we provide for variables for which we DO have a controlled process. I agree with you "Bev D" but with no-one having control over that "process", there is no way it is stable or homogeneous and I do not feel comfortable with calling it a "control" chart :). And I also agree that specifications do not come from control limits but that is what the customer feels they "need." Whether that is the case is a sticky subject to discuss with this customer, unfortunately.
 

Bev D

Heretical Statistician
Leader
Super Moderator
That's unfortunate because it is a control chart - that is it's name, regardless of your (mis) interpretation or intent, no matter how well intended.

In fact if you think about it, using it properly would help you demonstrate to your Customer that their misguided request to use the process variation to set their 'specifications' is misguided.

A control chart is more than a vehicle to help you control your processes, it detects unstable processes. By properly plotting the process you can demonstrate that the trace elements are not homogenously distributed across a physical area and/or from area to area and that placing 'specifications' based on any area is futile since you cannot guarantee the results and that if the Customer really needs to keep the trace elements below some level they will have to pay more for it. If your Customer is reasonable this will work if they are not you can use this data to demonstrate to your leadership how futile such an approach would be...not all Customer's are worth having.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
How about:

A Waste of Time and Resources Chart

The technical term for this is a "Report Card Chart". It is used to verify process variation based on process noise, rather than process monitoring feedback to a control variable.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Thanks for your feedback, everyone! I wanted to differentiate it from other charts we provide for variables for which we DO have a controlled process.

It is a "Report Card Chart" that reports the variation from process noise. Don't lose sleep over the calling it a "control chart" - no Shewhart chart controls a process, they monitor a process to detect special causes that you may or may not have an opportunity to react to.

I agree with you "Bev D" but with no-one having control over that "process", there is no way it is stable or homogeneous and I do not feel comfortable with calling it a "control" chart.

If it shows a natural, random variation with no special causes and the samples show independent variation between them it is "statistically stable". If the variation falls within it historically typical variation (determined when developing control limits), it is "in control". Anything else is reading things into the terminology that is not really there. Is it capable? Shewhart charts do not relate to capability.

And I also agree that specifications do not come from control limits but that is what the customer feels they "need." Whether that is the case is a sticky subject to discuss with this customer, unfortunately.

That is generally the case, as most customers have not "bookended" their process to determine the true specifications to meet design functional requirements.
 
Top Bottom