Understanding Data Variation in Six Sigma: Special Cause Variation & Common Cause Variation, & How to Reduce Variation

Understanding Data Variation in Six Sigma: Special Cause Variation & Common Cause Variation, & How to Reduce Variation
Page content

The Importance of Data Variation in Six Sigma

Six Sigma has its roots in the manufacturing sector, and a primary focus of the Six Sigma methodology is process and quality improvement via reduction of process variation. The term “sigma” is a concept from the statistics field that is a measure of the standard deviation of a set of data. The term “six sigma” reflects the ultimate goal of having a process that has only about three defects (failures to meet customer and manufacturing specs) out of a million opportunities. Such a process has a sigma level of 6.0. This level of performance is usually not realistic in service settings, but the name is used nonetheless. Most Six Sigma projects follow the DMAIC methodology to reduce variation and improve performance to meet customer expectations.

To understand the impact of variation on performance, consider the two graphs below. Suppose a pizza company is tracking their delivery times relative to the time that they tell customers to expect their pizza. Certainly customers do not want their pizza to arrive late. But there may even be customers who do not want it to arrive early, for instance if they are placing the order when leaving work to go home, or if they plan to take a quick shower first. The company has decided based on customer feedback that they want to be sure to arrive within 8 minutes of the planned arrival time at a customer’s house.

In the first graph, you see that this company is usually arriving within the window they consider acceptable, as shown by the large portion of the curve that is within the two dotted lines indicating the lower and upper limits. However a percentage of customers are receiving their pizza either more than 8 minutes before or more than 8 minutes after the time they were told to expect delivery.

Often to change process performance, a group will try to shift the average, for instance in this case the company might try to just achieve earlier delivery in all cases. But this won’t please customers who don’t want their pizza too quickly, and there will still be enough variation in the process that if something changes and the average increases again, performance will suffer.

Only by reducing variation will performance be substantially and sustainably improved. In the next graph, you can see the result of an effort to tighten up the pizza delivery process, so that all customers receive their pizza within 8 minutes of the expected delivery time. Furthermore, even if there is a short-term problem or a long-lasting change, the delivery time could increase or decrease somewhat without resulting in increased defects.

Process with low variation

Note that this improvement has occurred without even shifting the average time, or the upper and lower spec limits.

Types of Variation

So how does an organization go about reducing the variation in its process using Six Sigma? That will depend on the type of variation that is present in the process. Common cause variation is variation in process output that is essentially inherent in the process itself. So for instance, some variation in arrival time for pizza deliveries will occur simply due to variations in traffic. It would not be fruitful to try to understand why one delivery to a specific location took 13 minutes while another took 15 minutes, if 97% of the deliveries to that location are generally within 10-19 minutes. Both those values are well within the typical performance of the process. However it may be feasible to improve the overall process so that variation is reduced, perhaps by changing the way routes are determined, or using a bicycle which can maneuver easily through traffic jams instead of a car for deliveries.

If, however, the delivery to that location occasionally takes 32 or 34 minutes, management would be wise to seek information about the cause for each of those outliers. Was it a traffic accident or construction at that time on that route? Did a problem in the kitchen cause a delay in the delivery vehicle’s departure? Or is there one driver who consistently takes well over the typical amount of time for his deliveries?

When only common cause is present, process output is completely predictable, to the extent that performance can always be expected to fall within a certain range. On the other hand, when special cause exists as well, process performance does not always fall within that range and can not be reliably predicted. Control charts are often used to provide a visual representation of this concept.

The proper means of addressing process improvement and variation reduction depends on whether common cause, special cause, or both types of variation are involved. This is particularly important in settings where employee performance is being evaluated: it is neither fruitful nor motivating to interrogate or discipline employees for doing something more or less quickly than another employee if the variation is all inherent in the process.

Techniques for Reducing Variation

Reacting to common cause variation requires a different mindset and set of techniques than reacting to special cause variation. When special cause variation is present, it means that specific factors exist that can impact the process performance in a specific instance, as compared to performance in other situations. For example, weather problems, power outages, and traffic accidents represent special cause that can impact whether pizza is delivered to a customer within the appropriate amount of time.

In order to reduce variation due to special causes, Six Sigma project teams can put in place a procedure for detecting the presence of those factors in a timely fashion, and using contingency plans to minimize their impact. For example, when a delivery route is blocked due to accident or unexpected construction, a notification can go out to all dispatchers and driver so that another route is employed.

When only common cause variation is present in a process, it does not make sense to try to understand why performance was as it was in one specific instance. It is necessary to look at the process overall and determine how variation can be reduced and performance improved. While special cause may be a sign that a process is fundamentally sound but prone to impact by outside forces, common cause is a sign that substantial variation exists in the process itself and that it needs fundamental changes to meet the performance goals and to reduce variation.

The most common way to understand and impact common cause variation is to determine major factors that vary, such as department performing a function, time of day that the process is performed, or equipment used in the process. Graphing and data analysis techniques are used to separate out these different parts of the process, such as by evaluating performance separately for the day shift and the night shift or based on which type of product is being processed.

Once the factors affecting performance are identified, the DMAIC process can be followed to uncover root causes for defects and variation and to implement sustainable improvements.