How to Master Data Evaluation in Medical Device Packaging

Let’s talk about PGa, and I don’t mean the Professional Golfers Association.

When I was going through Process Excellence Black Belt Training at Johnson & Johnson, our instructors introduced a way to evaluate data that has stuck with me through the years and is still beneficial. The instructors specifically listed each of the initials with a different size to emphasize where to prioritize most of your effort/time.

First, Practical.

The large “P” stands for Practical. We are encouraged to look at data from a Practical perspective. Is there anything unusual about the data set when just looking at the numbers? Are there clear outliers? Could there be data entry errors? Is the standard deviation reasonable compared to other studies from this process and or material type? Is there anything from a practical perspective that triggers further investigation?

If there are obvious characteristics of unusual data, one could start going through root cause analysis using tools such as the Ishakawa “Fish bone” diagram or 5 Why’s.

Then, Graphical.

The medium-sized “G” stands for Graphical. Only after we’ve determined there are no practical reasons to question the data, the next logical step is to look at the data graphically. Looking at graphs/charts gives additional insight into any potential watchouts with the data set from the subject process.

Graphically is also a good way to depict a comparison of two or more data sets. Edward R. Tufte, the “da Vinci of data,” said it well: “Graphical excellence consists of complex ideas communicated with clarity, precision, and efficiency … (it) is that which gives to the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space.”

Finally, Analytical.

The tiny “a” stands for Analytical. Only after we looked at the data Practically and Graphically should we spend time looking at it Analytically.

If one goes straight to Analytical, as many do, we miss obvious things that need investigation. We’ve seen lots of time spent on analyzing normality and debating which transformation is acceptable to perform to get the p-value to be ≤ 0.05. Some companies limit which transformations are allowed and documented in statistical Standard Operating Procedures (SOPs); while some are less sophisticated. But much of that time could have been saved by spending a little more time in the Practical phase.

This piece is light on statistics but is intended for teams to pause and look at the data Practically before spending lots of time in the Analytical phase.