Before starting with the measurement analysis, we first need to look at the difference between variables and attributes data.


  • Variable data: is a quantitative measurement of some physical characteristics like the height of something or the time needed for something and so on;
  • Attribute data is a quality characteristic or attribute described in terms of measurement. It can be measured in discrete numbers when you count something (a car with four wheels), or it can be a binary attribute like Yes or no, black or white, etc.

In each of these measurements, you can have two types of errors:

  • Type I Error: This is a false alarm, so you say that you have an error when you doesn’t have it;
  • Type II Error: This is a missed error, so you say something is good but has an error.

Make a measurement system analysis aims to find if the variability of the measurement system is low compared to the process variability.

Example: if you make an error of +/- 1 meter in a GPS, maybe is a low variability. Measuring the construction of parts of cars can be very big.

In addition, you can want to look if the measurement system can distinguish between different parts. 

Example: if you measure a person's height with a measurement tape, it's ok. If you use the measuring tape for the micron error on the electronic part, maybe you don't get the difference, and you don't get the error.  

The key difference in between Variable and Attribute analysis are:

  • Type of values: in Variable, you have continuous value instead of discrete;
  • What do you control in the analysis: In Variable, you control how the inspector measures the dimension and error. In the Attribute, you control how the inspector make the decision;
  • Find the type of error: The attribute data analysis help to identify the type (I or II) of error; instead, the Variable can’t do that.

But how much measurement error is acceptable? It depends on the process, but a euristics can be:

  • Less than 10% is ok
  • from 10% to 30% is a grey zone that depends on the process;
  • more than 30% is not ok;
Share on: