Gage R&R

Gage R&R (Repeatability & Reproducibility) is a statistical technique within the broader Measurement System Analysis (MSA) field used to quantify the precision of the measurement system. In a nutshell, it helps distinguish what portion of the variation observed in the data is due to the actual object or is due to the precision (or lack thereof) of the measurement system.

It may be used before trusting any data coming from a measurement system. Typical areas of application: method validation and method transfer, process improvement and process control. 

STEP BY STEPS BASICS

1. Take parts representative of the process (usually 10 parts), making sure that they represent the full variation of the process.
2. Let one appraiser measure all parts.
3. Repeat step 2 for the same appraiser (keeping the parts in random order to have the appraiser “blind”) at least once, better twice.
4. Let a second (and third) appraiser repeat steps 2 and 3:

• If the measurement process doesn’t depend on people, reproduce the measurements with other measurement equipment
• If there is only one person and set of equipment doing the measurements leave the reproducibility out of the study

5. Run the tests with a statistical software and analyze the results.

In Gage R&R studies it often matters how many tests are required to achieve reliable data.

WORK SYSTEMATICALLY WITH METHODS     

Our Courses 

TWO MAIN INDICES TO ANALYZE R&R

% R&R

Describes the variation of the measurement system in comparison to the total variation.

The acceptance criteria for this metric is the following:

• %R&R < 10%             : the system can distinguish the variation in the parts reliably and is considered a very precise measurement system.
• 10% < %R&R < 30%  : the system is considered precise enough.
• %R&R> 30%              : the measurement system is considered unacceptable and must be improved.

% P/T

Describes the variation of the measurement system in comparison to the part´s tolerances. This is important to make sure you do not send bad parts to customers nor to scrap good parts.

The acceptance criteria for this metric is the following:

• % P/T< 10%               : the system can distinguish the good from the bad parts reliably on a very good level.
• 10% < % P/T < 30%  : the system is considered to distinguish good and bad parts precisely enough.
• % P/T> 30%              : the measurement system is not able to detect bad parts and accept good parts consistently. It must be improved.

If your measurement system analysis should be done for discrete data, you should work with Attributive Agreement Analysis. For example, if your data is based on qualitative evaluations.