Measuring Performance

Counting defects is an important task. There are many ways we can use the data that we gather during the Six Sigma practice. We want to examine the outputs of our process very carefully, and capture data. We may be documenting elapsed time to provide a service. Or maybe the weight, or color of something, dimensions, or other attributes. When evaluating defects, we may be determining the percentage of rework required, the amount of scrap materials, or perhaps the number of complaints from our customers. We gather and we analyze all of this data to calculate the defects per unit, the overall defects per million opportunities, and the roll through yield. And how we are performing against the entire end to end process. So what are defects? Our customers are actually the ones who’ve primarily defined what defects are. And generally, a defect means we’re not meeting their specifications. Some of the things that will have specifications could be the life cycle. Does the product function for the correct amount of time before it needs to be repaired? Or for a product that requires routine maintenance, is it easy to maintain? Is the product the correct color and size? And does it function the way it’s intended to? In the service sector, meeting expectations could mean providing accurate responses within a reasonable amount of time.

At a fast food restaurant, was the food prepared quickly, did the sandwich contain the correct toppings, and was the customer treated in a friendly manner? The causes of the defects are typically found in the process. And they can include defective machinery, human error, and much more. In measuring process variability, we are comparing the attributes of our deliverable against customer requirements. Let’s say that the customer requirement is the mean or bull’s eye of the target. But there’s a range of acceptable values that are defined by upper and lower control limits. We determine, along this range of possible results, where our samples are falling. How close or how far from the target are they? Some of the things we might look at could include the weight of a food portion. The time it takes to respond to a support inquiry or time to repair a vehicle at a repair shop. It could also be a satisfaction rating that we’ve received from a customer. The values that we gather with this method can be placed and plotted along the continuous scale. It will generally form a normal distribution with data points tightly clustered around the mean. And also randomly distributed above and below the mean. These data points can be evaluated and converted directly into a sigma value. That will tell us what percentage of our samples are within the acceptable range and meet customer requirements, and what percentage would be considered defects.