**Special or Common Causes**

From: M. D. Anderson Department of Performance Improvement

Date: January 15, 2008

**Kathy Price: **Hi, I'm Kathy Price and I'm new to the department of Performance Improvement. And I wanted to share with you today about special or just common causes.

So lets go back - what is a good measure. Well a good measure should reflect the voice of the customer, tie to business goals, identify opportunities to improve the process and the measure should be consistent. If we focus on consistency, we can establish a valid estimate of process performance.

So there are two types of variations. We call them common cause, which is always present to some degree. We call this normal variation. Its random, its expected. Then there's special cause. That's something different happening at a certain time. Its unexpected or a fleeting event. That's when things are out of control. So, if we measure over time, we can use control charts to visualize the process.

So how to read a control chart? Remember the normal distribution curve, or the bell curve, that's around an average or a mean of normally distributed points? We look at the central line to see if its relative to specifications, our needs, or our objectives.

Then we analyze the data. Is it within the control limits? Then we look at it and say, okay, is it a special cause or is this common, a normal variation? The process is in statistical control when its not being effected by special causes. All the points must fall between the control limits. And they are randomly dispersed. But know that in control does not necessarily mean that the product or service will meet your needs. It only means that the process is consistent. When we look at a control chart, we have the upper and lower control limits and outside of those upper and lower control limits we call that the region of special cause. So something unusual is happening if a data point is there.

But we also look in the region of common cause, in between the upper and lower control limits, and we look for runs, trends or patterns that are indicators of special cause. There are several guidelines or rules that we use, to look for the special causes. These guidelines are based on probabilities so, you want to use your process... your knowledge of the process and these rules to determine if it's a special cause. Because you know the process and you ought to be able to identify something unusual as soon as it occurs or pretty close after it occurs. So of these we've got the action point. You want to action something if it goes out of the upper control limit. We've got the four out of five. When you see four out of five points in zone B or out in the two sigma level, we've got statistical trend. That's when there are 6 or more successive points in a row, either going up or going down. And we have 2 of 3 which would be 2 of 3 successive points beyond the 2 sigma level in the zone A. And then we have process shift, where 8 or more successive points are on the same side of the mean.

There's a couple of other guidelines or rules that we didn't put on this slide, but there's 14 alternative consecutive points within zone C, which we call saw-toothing the mean or the central line. And the other one is 14 consecutive points in zone 1 which is hugging the mean.

So if you see any of these trends or special causes, what are you going to do? Our goal is to eliminate these special causes. We want to make an unstable process, stable. So we want to get timely data. We want to take immediate action. Find out what was different on that occasion. And then develop a remedy for that obvious problem.

And then what if we don't have special causes? We just have common causes, everything is in between the upper and lower control limit? Well, we would say, the process is stable, but does it need to be improved? It could be in control, statistical control, and be statistically stable, but it doesn't meet our customer's needs.

So are the control limits within the specifications of our customer's needs? Does the mean meet our expectations? Leaving the process alone is not improvement. So we would have to go further to improve the process.

Well thank you for your time and attention and now we'll go back to the studio.