Once your data is collected, what should you do with it? For starters, graph it. Put it into a histogram, run chart, control chart, pareto chart, i.e., anything that will help you get information concerning the metric you’re trying to improve. Let’s discuss a few specifics.
If you have variable or continuous data, such as a product dimension, start by putting the data into a histogram to determine if the data is skewed or symmetrical. Next, put the data into a run chart to determine if there are any patterns, trends, etc. Next, control chart the data to see if it is in statistical control. Are there any special or assignable causes present or is there only common cause variation present? If you can stratify your data into categories, such as, machine, person, customer, etc. a useful tool to compare them is a box plot. All of this is important information that can help you proceed with the analysis phase.
If your data is attribute in nature, i.e., number of errors, defects, etc., you can first use a pareto chart to separate the data into categories and then follow up with an attribute control chart to see if your process is in statistical control.
I was investigating a process once where a team had conducted a capability study and found the process capable based on a sample size of 30. After collecting a larger sample of 100 pieces and plotting the data in a run chart, we notice a pattern that wasn’t evident in the smaller sample and lead us to solve the issue.
As the famous statistician, Ellis Ott, said, “Always, always plot your data.” And remember, a picture is worth a thousand words!