Thursday, September 18, 2014

The Humble Histogram

Histograms.  The most underused, yet potentially helpful form of visual representation of data. They show you where most of the measurements are located and how spread out they are. This is helpful when you need to summarize large data sets graphically, compare process results with a spec limit, or communicate test results visually.  They are a product testing tool that can be very helpful in decision making! 
It’s a natural reaction to dislike things we don’t understand, and many people aren’t fond of histogramsmuscle_confusionsimply because they don’t understand them. So, let’s start with the definition: A bar graph of a frequency distribution in which the widths of the bars are proportional to the classes into which the variable has been divided and the heights of the bars are proportional to the class frequencies. 
In other words, a histogram is a plot that lets you see the frequency distribution of a data set, a graphical representation of a set of measurements. The histogram creates a shape that allows you to see the number, or the percentage, of defects or intolerances produced during the time you collected the data. You could also add product spec limits to your histogram, so you can see with just a glance if the current process is able produce products that meet the spec. The histogram can also show which values occur most often. Having all this information in a handy chart makes it simple to communicate data to other team members and then, deciding your next step becomes much easier. I suppose you could say that the histogram tells you whether the process will produce goods or services that are within the specification limits. 
Well, in and of itself, it cannot tell the future. However, it will help you predict, with a simple glance, whether future products will be within the tolerances allowed, or whether or not your process is effective at producing goods or services that are within their assigned specifications. Notice we did not say it WOULD predict, only that it will help you predict the outcome for your product or service. But, that’s not all you can learn from a histogram. Depending on the SHAPE of the histogram, you can learn much more than just product or process performance!

Well, there are several identifiable shapes in a histogram that have a certain ‘meaning’.  For example, a bimodal shape as shown below could signify that something has changed in the middle of your production run.  
Maybe it is caused by a substandard batch of material.  Maybe a part on a machine broke and it hasn’t been discovered, yet.  Maybe the production rate changed.  The bimodal histogram indicates that further investigation is warranted.  What gets discovered can improve product consistency in the future.

The Bell Shaped histogram generally represents a normal distribution, where the values are concentrated in the center, but if the values are more frequent around the high or low ends, you will have what is called a ‘Skewed’ shape. It could be skewed to the left (the shorter side, or tail, points to the left) and is said to be negatively skewed.  The opposite is true, if the tail is to the right, it is skewed to the right and is positively skewed.  A skewed histogram is related to the mean (average), median (middle of the data), and mode (the value that occurs most often) of the data.
A plateau shaped histogram often means the process is not clear to those performing the work. Everyone is accomplishing the end goal with a different set of steps. This means your process is not as clearly defined or efficient as it could be. 

Yes, it is! Histograms are very useful in identifying quality issues, as well as product performance. NTA uses histograms in their test reports, which allows you to really SEE how your product behaves, and how efficient your manufacturing is! If you would like more information on how our test lab can use a histogram to help you understand your product and increase efficiency, contact us!

No comments:

Post a Comment