Understanding Variation

Analyzing cause to implement corrective action. Organizations with quality management systems compliant to ISO 9001:2008 are required to take action to eliminate the causes of nonconformities.

Clause 8.5.2 defines steps required for corrective action (CA), including determination of the nonconformity cause, along with determination and implementation of necessary action to prevent recurrence. Similar wording is also present in clause 8.5.3 regarding preventive action (PA).

A cause of a nonconformance should coincide with a variation occurrence. With regard to this variation, is its source common cause or special cause? This question must be answered for proper cause identification because the answers may yield different resulting paths for CA. Determination of variability type is often absent from problem-solving methods, leading to ineffective actions.

An example explains the two types of causes of variation:

Common cause: It normally takes me 25-35 minutes to commute to a neighboring town. Note that it does not take me exactly 31.5 minutes each time because there is attribution of common cause variation. The value in the range could be affected by factors such as the number of red lights I hit, traffic volume or weather conditions, such as rain or sun. These are a normal part of the drive. Expected common cause variation may be predicted by a control chart, often with limits of the mean +/-3 standard deviations. Common cause variation is present in every process.
Special cause: One day, I arrive at the town in two hours, which is statistically peculiar. There is a special cause associated with this incident that is outside the normal system: On that day, a blizzard contributed to the delay.

Address the cause
To solve a problem with a special cause, the team should be looking for what changed or is different, whereas solving problems attributed to common cause will require reducing the variance, increasing the spec range or shifting the process mean. All of these relate to not what is different, but rather what is the same (intrinsic) in the process.

It is difficult to create an effective solution if the problem and its cause are not accurately understood. CA teams can spend a lot of time solving a problem, only to have it recur because the team treated a common cause as a special cause.

To demonstrate the effects of common cause, consider a pizza shop in my hometown that includes the neighboring town in the earlier example in its delivery area. This pizza company guarantees delivery within 45 minutes or the pizza is free. It takes the shop 15 minutes to prepare the pizza. Then, the driver takes 25-35 minutes to drive to the neighboring town due to common cause variation. This results in a total delivery time of 40-50 minutes. At this rate, the driver is on track to give away 50% of deliveries to that town.

Should the manager be giving the driver grief for a system that is incapable due to common cause variation? Perhaps the manager could focus more on the 15 minutes of preparation time. He or she may have more influence over the sources of variation in the store to reduce the process mean. Or, the manager could change the scope of the spec by removing the neighboring town from the delivery area.

Fix the system
Most causes are system problems that should have system solutions. If an operator makes a mistake, something in the system allowed the mistake to occur or go undetected. Instead of engaging the system, a common response by CA teams in this situation may be to retrain the operator.

The operator’s performance may be outside the range of other employees, implying an outlier, but what if the training itself is ineffective? Why retrain with faulty training? Why was the training ineffective the first time? Is there an aspect of the training that needs to be changed?

I wouldn’t consider "employee was retrained" to be a CA. If this is not a management issue—such as a blatantly defiant employee—something in the system needs modification. Correspondingly, each CA or PA should have some system documentation change as a part of the record.

As all of these examples illustrate, determining the variability type—special cause or common cause—in root cause methods will lead to more effective CA and PA approaches within an organization.

Reference: QP