Accuracy OR Precision

“Be precise. A lack of precision is dangerous when the margin of error is small.” – Donald Rumsfeld

As the title implies, I want you to think about which is more important: accuracy OR precision. Most quality professionals know that you need both accuracy AND precision to aid in process control, but if someone made you pick one, which would you choose and why?

Here are some simple definitions of the two:

ac·cu·ra·cy: the quality or state of being correct; the degree to which the result of a measurement conforms to the correct value

pre·ci·sion: the quality, condition, or fact of being exact; refinement in a measurement represented by the number of digits given

A somewhat famous way to visualize accuracy and precision is in a bulls-eye target:

To summarize the difference,

• Accuracy is hitting the intended target (think bulls-eye).
• Precision is repeating the same thing multiple times (think repeatability).

Now to answer the question of whether you prefer accuracy OR precision. My answer is easy: I prefer precision. In a process, it is much harder to change precision (variation) than it is to change the accuracy (mean). Let’s consider process capability similar to the bulls-eye picture above. The lower spec limit is 5.0, the upper spec limit is 10.0, and the targeted nominal is 7.5.

If my target is the middle point between the red vertical lines, then the 2 charts on the left have a mean that meets the target. This means they are accurate. If my data has a small standard deviation, then it is precise (such as the top 2 charts). 

The reason I would prefer the top right chart over the bottom left chart is that the top right chart is showing a process that is providing consistent output. If I’m able to shift the mean closer to the nominal (middle), then I end up with the top left chart which is both accurate and precise (and a very capable process).

The bottom left chart shows a process that is hitting the target an average number of times, but also varies wildly. Finding all the potential causes of that variation can be a daunting task. Furthermore, reducing the variation is often not easy and can be costly.

In summary, if I must pick between accuracy and precision, I choose precision because, generally speaking, accuracy (mean) is easier to change in my process than precision (variation).