Finding Process Center in the Eye of the Data Storm
If you work in manufacturing, you probably see data every day. You may see it so often that you’ve become numb to it. For example, how often do you see product quality data? Is it always sitting there, right in front of you every time that you take a walk on the production floor? I know that in my 40 years of working in manufacturing, data existed in every corner of the plant. I would see so much quality and process data in the form of individual data points that it just blended together in my head like a big data tornado.
So, knowing that this data exists everywhere in your factory, how can it be summarized and characterized into more usable formats. Well, there are multiple ways to do this. Some are simple and some are more complex, and each methodology provides us with different insights into our products and processes. Six-Sigma methodologies envelop an entire umbrella of process statistical tools. However, in this video, we’re just going to take a look at the simplest way of characterizing process data and discuss measures of central tendency.
The measure of central tendency is simply a measurement of a data set's center. If a data set meanders back and forth like the body of a snake, the snake’s head represents the center of the body. In the midst of a jungle of data and total information overload, a measure of central tendency can really help to simplify your view of any data set.
There are 3 ways to characterize a data set’s center. Median, Mode and Mean. These are all measures of central tendance. When you look at any data set, the median is the center most data point and the mode will have the highest frequency of occurrence. However, these measures don’t typically have a place in manufacturing process management. So, let’s just set these aside.
Instead let’s talk about statistical mean. The mean of a data set is simply the data set’s average, and you can easily calculate mean by adding all the data points together and dividing by the total number of data points. In the world of process management, a data set’s mean is great information to know. But why is it good information for us? Well, all by itself, it can be meaningless, but when you provide it with some good context, then it comes to statistical life.
If for example your manufacturing process is producing safety harnesses for powerline workers, you may want to know the mean tensile strength of harness stitching. Let’s say that you have access to this data, and you calculate the mean stitch strength to be 917 lbs. Is that result a good result or a bad result? The truth is that you just don’t know without some level of context.
To get some context, let’s start by looking at your products specifications. You sell this harnesses and market them under the premise that it safely supports up to 500 lbs. of weight. However, your factory specification calls for a stitch strength of 750 lbs. This buffer is in place such that harness is actually over designed versus your label claim for safety purposes. So, the fact that your actual stitch strength is 917 lbs. tells you that you are in good shape vs. your spec.
However, if your factory specifications called for 1000 lbs. minimum tensile strength, then you would be in trouble. Product may need to be placed on quality hold and maybe even a recall would need to be initiated. In this example, context in the form of a product specification matters.
When it comes to statistical mean, what about trends? Trends can also provide context. If your recent data reflects a mean result of 917 lbs. of stitch strength, what was the mean stitch strength last week? What was it last month? By comparing your current process mean to previously calculated means, you can start to piece together process trends. And by analyzing these trends, you can garner valuable information about process controls and unknown process inputs.
So, in summary, mean is a measure of central tendency and reflects a data sets centering. Mean can also be called average, and it is easily calculated by adding up all of the data points and dividing them by the total number of points. The mean of a data set can be meaningless without context in the form of process specifications or trend data.
Lastly, a mean calculation is limited in usefulness without additional process information. As such, it is critically important to enhance our process understanding beyond that of just central tendency. When looking at process data we need to expand our toolbox and explore measurements that characterize process variation too.
Quantifying process mean and process variation will provide us with the needed information to control our manufacturing processes. If you are interested in the dynamic field of manufacturing, then don’t forget to enroll in our CML100 training!
Stay connected with news and updates!
If you want some weekly T4T wisdom coming straight to your inbox for your reading pleasure - look no further! Join our mailing list to receive the latest blogs and updates.
Don't worry, your information will not be shared.
We hate SPAM. We will never sell your information, for any reason.