Dear Folks,

I'm the QM of a manufacturing organization.
I was asked to write a procedure in order to control non conforming PROCESSES (every process defined in our organization process map).
I have three basic requirements:

1. Identify and record the process variation that generate the process NC
2. Evaluate if this NC has generated a PRODUCT NC
3. Identify and Control the NC product according to existing procedure

Do you have sample procedures covering points 1 and 2 in order to give me some inspiration to start?

Thank you.
Giuseppe:thanx:

You want the SPC forum, SPC Monitoring and Statistical Analysis Techniques (http://elsmar.com/Forums/forumdisplay.php?f=19)

Depending on what and how your company makes stuff is the basis for the program you will need and use to find the NC in your process. Once you have a metrics in place to find the NC of the process then you must identify how and why the NC are made in the process, once you have done that eliminate the cause of the defects. That is the basics of quality, since you do not know them I suggest you hire someone who does and take credit for the success. :yes:

Since SPC is based on events and not all processes would have measurables that would be served well with SPC, I suggest a different approach.

Here is a thread titled Preventive Action (PA) and Corrective Action (CA) - One or Two Procedures? (http://elsmar.com/Forums/showthread.php?t=5985&highlight=procedures) Within that thread is a flow chart procedure for Preventive Action that I would like you to look at.

It's important to recognize that not all variation results in NC. Thus, preventive action can be taken by noting something in a process that should be improved, even if it has not yet produced flawed product or service.

The evaluation of needs and steps for preventive and corrective actions are not profoundly different. You could read and consider the steps and decide what would work for your organization.

I hope this helps!

Since SPC is based on events and not all processes would have measurables that would be served well with SPC, I suggest a different approach.




I can not think of anything without a process and some way to apply some form of SPC to study it.

If it is a go/no-go (attributes) process you can still run a SPC program for it.

If he tells us what his process is, I bet someone here will be able to lay out a general SPC plan, based on the process. What he describes sounds to me they want a SPC program.

Thank you for the answers.

In principle I agree with both of you.

Jennifer, the FC is great and I think it would fit my needs.
What I need to do first is to define when a process variation (a KPI variation) means that my process in NC.

I have metrics for every QMS process.

For example, what I would do for "On Time Delivery" metric is:

1. set a monthly minimum acceptable level.
2. Set a monthly target for optimal performance
3. Check performance against these two levels

I would consider the Planning process NC if it does not meet the minimum Acceptable level (CAR Issued).
I would consider the planning process ok if it would be above the minimum level.
If the process consistently meet the target (i.e. 6 consecutive months), minimum and target levels are reviewed.

Any thoughts?
Thank you.

I can not think of anything without a process and some way to apply some form of SPC to study it.

If it is a go/no-go (attributes) process you can still run a SPC program for it.

If he tells us what his process is, I bet someone here will be able to lay out a general SPC plan, based on the process. What he describes sounds to me they want a SPC program.

I think the misunderstanding perhaps comes from an overly narrow view or definition of "SPC." Every repetitive process has characteristics that can be measured, and those measurements may be evaluated through statistical analysis. This doesn't necessarily mean Shewhart, n or p charts; it can be as simple as plotting data and observing patterns or trends. For example, if a process is supposed to be yielding n conformingparts per hour and the actual output is n-x, plotting the output over several shifts, hours or days might reveal patterns that will help identify the problem(s). In its broadest sense, this is SPC.

I'm glad the link helped you. Yes, I have questions.

1. How are your targets set?
2. How often do you look at the process's performance and decide if something needs improving? Does anything prevent you from targeting a process for a different reason? If so, what sort of reasons might that mean?

Deming cautioned against Management by Objective for good reason. If a numerical target becomes the ideal, what role does the process user play in deciding an opportunity for improvement is apparent? Any?

Have you tried putting this in practice? Let's consider what it might look like.

A billing system is examined after a month to see if cycle time is satisfactory. The manager decides it is not, because the cycle time has missed targets--it takes too long to send out the invoices. The manager initiates action to improve the process.

This leads me to a number of questions:

a) Why wait a month?
b) Who gets to determine an action is needed because a cycle time target is missed? Is there a limit to this authority?
c) Did the billing clerk know there was a problem and say nothing because a limit such as in (b) existed, or was assumed?
d) What could have been gained if the billing clerk had been allowed to alert the manager--or better yet, take some initiative if he/she has the authority, to solve a problem that lengthens cycle time?

I can appreciate the wish to make a well defined improvement plan, but I want to stress that some flexibility can be very helpful. Avoid stipulations (they can be guidelines) that lock you into action, or inaction, or unncecessary delay.

I hope this makes sense!

I think the misunderstanding perhaps comes from an overly narrow view or definition of "SPC." Every repetitive process has characteristics that can be measured, and those measurements may be evaluated through statistical analysis. This doesn't necessarily mean Shewhart, n or p charts; it can be as simple as plotting data and observing patterns or trends. For example, if a process is supposed to be yielding n conformingparts per hour and the actual output is n-x, plotting the output over several shifts, hours or days might reveal patterns that will help identify the problem(s). In its broadest sense, this is SPC. I agree that I do not easily envision statistical management of nonproduction processes such as materials procurement. While outcomes could indeed be trended, they are often very hard to control like production processes are. So, I would tend to grade them in qualitative fashion instead of using the kind of data that would be charted.

Jim, Jennifer I definitely agree with both of you.

I set targets on my KPI once per year during annual management review.
Then I check performance each month during quality management review.

Coming back to the examples (my OTD index on monthly orders and Jennifer's Billing process) I think that a monthly performance evaluation is fairly reasonable.
Unfortunately the reaction time in terms of corrective actions is not that fast but I would find it rather difficult to increase the process reviews in terms of cost/benefit ratio.

Of course real life may be full of surprises... so maybe the right choice is to start with a definite criteria and then check if it works or further adjustments are necessary...

What do you think?
Thnak you again for your valuable help on this subject!

Check out a performance index weighted, then use the values you get to chart aganst your goals. You can even tie in a pareto for causes of failures.

There's no question of the value in using performance data to recognize need and urgency for process improvements. As long as customer satisfaction is not damaged by a delay in this recognition process (time needed to collect data) then agility might not be very important.

However, I want to stress that all process users should be invited to be sensitive to how well their work is going, and to think up what could help. Not all process improvements need to be formal affairs, and the top performing organizations share attentive employees as a common success trait. I would hate to sacrifice that for pragmatism. I would prefer to somehow manage a blend of the two. We don't want to be like weathervanes, tweedling a process with every notion; nor do we want to become automatons and suffer what's called "death by a thousand cuts".

So I would build into my process corrective action process a selection of means for identifying need for improvement without waiting for study of performance against KPI.