Using Histograms for effective data interpretation in Six Sigma

Last Updated on 16 September 2023

A histogram is a chart (in effect, a bar chart), where instead of the usual 2 or more variables, there is only one variable that you are measuring. You will often in Six Sigma come across a large amount of data for one variable (usually the variable you are trying to control), and it can be hard to interpret this large amount of data. A histogram makes this easy to see by showing the data in an easily to interpret format.

This will easily show:

• Where certain values of the data make up unexpectedly high or low proportions of the data
• Where there is a skew (e.g. low values occur more often than high values)
• The spread – is there high or low variation in the results
• How many data points there are far away from the others, that could otherwise skew the findings
• The shape of the peak – is it tight over a small area, large over a lot of ‘buckets’ (high variation), or even more than one peak?

This can give you a lot of information about your data that would be otherwise difficult to receive.

How to make a histogram

1. Measure your variable as many times as is practical. You want at very least 25 measurements, but you will get a better result if you have more than this.
2. Find the range of your data (largest to smallest value)
3. Split the range into equal buckets (range of the variable) (for my example below I have 20 ‘buckets’ of 5mm each, covering the range 0 to 200mm – you can change the number to get an easily interpretable chart)
4. Plot the data in a bar chart with a bar for each ‘bucket’, with the variable on the x (horizontal) axis and the frequency on the y (vertical) axis.

Example

If you have cut lengths of a metal, you need to know usual length and consistency. You can create a list of all the values, even put them in order, and find an average, but that still doesn’t give you that much information. Putting them in a graph would also be meaningless (they all have different lengths, so all the frequencies would be 1) and wouldn’t add value. You therefore create a histogram – in my example I’ve put them into 10mm buckets:

From this you can easily see that a large number are just over 100mm long, and it clearly shows the spread and variance much quicker than other methods of showing the data.