Importance of Performing Gage Repeatability and Reproducibility (GR&R) before actual quality data collection Part 2

In measurement system analysis (MSA), we would like to accomplish accuracy and precision of the ,measurement data.   Accuracy can be achieved through performing calibration on the measuring device for variable GR&R and clear operational definition for attribute GR&R.  Precision can only be achieved through GR&R studies followed by improvement of the measuring process if the GR&R does meet the required standard.   Calibration and GR&R studies are completely different.  Some manufacturers tend to  mix up both calibration and GR&R studies.  There was once an established supplier show me a calibration record when I ask for GR&R studies report????!!!!

GR&R studies need to have a lot of data collection,  very often most suppliers or manufacturers choose to ignore conducting GR&R studies unless it is requested by customer.  After working with over 100 electronic parts suppliers,  most of them will jump straight  into actual data collection without bothering about accuracy and precision of their measurement system.  They did not realize that without GR&R studies there is no way to know if the data they are collecting is reflecting the actual process variation.  GR&R study is a must for new measuring device or after the device had undergo major repair or move to another location.   In some industries GR&R study is required in new product qualification where GR&R report should be included in First Article Inspection (FAI) Report.

Although measurement process is one of 6 process input (5Ms and 1E),  however it is NOT the actual variation from process input. The variation in measure comes from measurement process and not the product itself.  We would like to reduce variation contributed from measurement process to be a small as possible so that our observed data will show more actual process variation.  Ideally in variable GR&R we only want < 20% of total process variation come from measurement input per item 5.  Therefore if GR&R value is above the acceptable level then we must improve the measurement process. 

The correct method to conduct GR&R is also very important right from operational definition to selecting sample to interpreting the results.  For those who had already attended my training class before would know the proper steps to obtain accurate GR&R value.

Add caption

Without GR&R study we will never know if the data collected show is the actual process variation or measurement process variation.  There are cases where measurement process variation contributed more than 50% of the total process variation. In this case we will only need to improve the measurement process rather than the actual process improvement!  If the measurement data is not accurate and/or precise, it will lead us to the wrong action and conclusion.  We could do a lot process improvement to man, machine, method or material and yet the data does not reflect the improvement we made due poor measurement process. 

In summary we have to conduct GR&R studies with the purpose
  • Quantify the precision of the Measurement System
  • Check for measurement system integrity
  • Ensure our data from measurement data does NOT have error which cause us to make wrong decision
Do refer to my course for more information on setting up effective inspection system:  Click on image to get link

Therefore I hope that after reading this article,  my reader will make some effort to conduct GR&R studies before any process improvement initiatives to ensure that process variation is NOT contributed from measurement process. 

Importance of Performing Gage Repeatability and Reproducibility (GR&R) before actual quality data collection Part 1

From six process inputs (5Ms & 1Es),  only 5 inputs (man, machine, method, material and environment) are contributing to actual process variations and it is inherent to the product.  Measure is actually a process that assess the product quality characteristic. Since measure is process by itself, there is process input variation which will contribute to the process output,  the measurement data.  The observed variation of a process is consist actual process variations and measurement gage variation.

Measurement gage variation comprise of 2 elements ,  repeatability and reproducibility.

Repeatability: closeness of the agreement between the results of successive measurements of the same object carried out under the same conditions of measurement
Can I measure the same thing more than once and get the same answer?

Reproducibility: closeness of the agreement between the results of the measurements of the same object carried out under changed conditions of measurement
Can I change the method of measurement, the observer, the location, the time (next day), and get the same answer?

In order to quantify the variation contributed by measurement gage repeatability and reproducibility we will need to conduct Gage Repeatability and Reproducibility,  GR&R studies.  Per my earlier article on Continuous Improvement Program CIP – 6 sigma  Methodology,  I have mentioned that there are 2 types of quality data, attribute obtain by counting and variable obtain by measuring.  Therefore,  we will also have 2 major types of GR&R studies base on the measurement process of attribute and variable data. 

There are many literature in the internet or textbook which document the method to conduct attribute and variable GR&R studies and I will summarize the methods of conducting GR&R study in this article :-

Before we start any GR&R studies we must ensure that :-
  • Variable GR&R study  :-  measurement device must be already calibrated to give accurate data
  • Attribute GR&R :-  clear operational definition such as accept or reject criteria must already established before counting or inspection
  1. Need to plan for sample size and prepare parts to measure or inspect.  Sample for attribute GR&R should be 3 times more than variable GR&R sample size. 
  2. Same operator inspecting measuring the same part multiple times for repeatability studies and record the measurement or inspection data.  Multiple operator who will be conducting the actual assessment (either measurement for variable data or inspection for attribute data) need to measure same part for reproducibility studies.
  3. You can either build GR&R calculation template base formula given books in excel spreadsheet or you can use statistical software such as JMP or minitab or Statistica which have this function to find the variation contributed by measurement through GR&R study
  4. Variable GR&R calculation is actually base on the percentage how much variation contributed to the total process variation (P/TV) or the total product spec tolerance. (P/PT)

According to Automotive Industry Action Group (AIAG) guideline the acceptance percentage for variable GR&R are per below :-

%  P/TV or P/PT
Resulted from
< 10%
Automated inspection or measurement process which do not operator judgement
Automated solder paste height measurement
Inspection or measurement which need operator judgement
Measurement of length using Vernier caliper
Not acceptable
No standardize measurement method Operator not trained
Product change over time very fast

  • Attribute GR&R is based on agreement of the measurement value between and within operators and operator effectiveness as compare  to standards.  Agreement value is report using a calculated Kappa index between value of -1 to +1,  the higher the value the better. Effectiveness is calculated base on percentage of correct decision over opportunities for decision,  for example I can inspect the part for 100 times and I get it correct only 80 times.  Therefore effectiveness percentage is 80%.Typically the acceptance >80% and best is >90%.

In part 2 of this article,  I will discuss a bit on measurement system analysis (MSA) two important component, precision and accuracy and what is the consequences of not performing GR&R.

Continuous Quality Improvement Program CIP - 6 Sigma Methodology

Common cause variations always exist in process inputs and outputs  therefore it is in the best interest of manufacturer to reduce these variations and optimize the output performance as much as possible through various methods. This is also known as a continuous quality improvement program (CIP) which focus on reducing defects which contributed by common cause variations and/or bring the process to optimum level to produce consistent good quality product.   Six sigma methodology is a popular method use for this purpose. This method warrants for systematic phase by phase approach of deploying advance statistic tools to analyze the data collected in a manufacturing process. 

The phases are :

Common tools used
1.    Define
Identify and understand the problem, establish problem statement. Define SMART goal
Scope the project
process flow
2.   Measure
Identify what parameter to measure, collect current state data to understand what is the performance of the  key parameter (Y). Calculate cost of quality
Critical to quality
Gage Repeatability and reproducibility
GR&R study
Statistical control chart
Understand data distribution
Process capability study
3.   Analyze
Understand what the root cause that is contributing to the problem, root cause analysis,  what are the X
Fish bone diagram
Fault tree analysis
5 whys
X and Y correlation
Design of experiment  to identify  and quantified process input (root cause) which contribute to Y
4.   Improve
Establish improvement plan to address multiple process input x  which contributes to poor performance in Y.  Pilot run the action plan and track if there is any significant improvement
Brainstorming  tools
Error proofing
Design of experiment to optimize Y through control for X
Anova analysis
5.   Control
Standardized and sustain
After improvement action. Document action plan in work instruction and ensure all follow the same work instruction -  Standardization Track the performance of Y output for a time line to if the improvement is sustainable, Calculate cost savings
After improvement control chart
After process capability analysis
Cost benefit calculation
6.   Report (optional)
Report phase 1-5 to management
Lesson learned and look for next project

Based on the input and output diagram,  process input  is termed as X and process output performance is termed as Y per below :-

Therefore,  process input (X) will always have an influence on process output (Y).  Most of process inputs can be controlled such as machine parameter setting and it could have an influence on process output Y. 

The simplify relationship of X to Y is Y= f(x)

In six sigma approach,  we will need to understand what is the process output, Y which we would like to improve.  We will need to collect data to understand which specific process input x  have an effect on Y using various statistical tools.   Once we have identified process input x which has an impact on Y, we can create action plan to optimize and control x.   For example if we would like to improve yield rate for a product, yield rate is Y, and process input which influence yield rate is x.  After we have optimize process input X,  then we will need to control X  and monitor Y performance.

There are 2 types of process output data (Y) being collected in a process :-
  1. Attribute data  (discrete data)
  2. Variable data (continuous data)
Also known as
Cannot be divided have decimal point ,  whole round number
Clear operational definition of
Go, no go,
reject, accept
pass, fail

Generated from a calibrated measuring device and can be divided infinitely
Distance, Length, weight
Counting of good and defect parts after going thru inspection such as visual or using go no go jig
Measure using measurement device
Yield rate
Defect rate
Count of reject
Dimension Length, width thickness,  strength, weight etc

The two types of  data have very different traits and use different statistical model use to analyze. During the define and measurement phase we must understand what type of process output data we will be using for the six sigma project.  The table below shows the different statistic models which are used the 2 types of data in various six sigma phase.

Statistical analysis
Attribute Data
Variable Data
Attribute GR&R
Variable GR&R
Measure Improve
Statistical control charts
P chart
nP chart
U chart
C chart
X bar R chart
X mR chart
X bar sigma chart
Measure Improve
Process performance index 
Process capability study
Analyze Improve
Design of Experiment
Binary logistic regression
Simple/Multiple Regression
Comparative methods
ANOVA methods
T test or Z test
Categorical analysis
Chi Square

Attribute data is a count data,  therefore a bigger sample size is needed for all the statistical analysis above compare to variable data.  In six sigma, variable data is prefer and it always encourage to find a variable metrics from attribute data.  An example would be printed circuit board assembly PCBA in circuit test (ICT) which test several electrical metrics at the same time,  it would be  more beneficial if we can study selected variable electrical metrics rather than just the yield rate of ICT. 

In real life, sometimes it is not possible to study every parts that had been produce especially in high volume manufacturing environment,  hence we will need to use statistical analysis of  sample to made conclusion on entire population of the part produce through 6 sigma approach.  We must maintain the rigour of data collection and statistical analysis to make a correct conclusion on the improvement made if it is effective to achieve consistent good quality product at optimum level.

To learn with case study to help you understand how to apply six to improve quality, click on the image or the link below👇:- 

Paying Tribute to a Century of Modern Quality Practices

Have you ever wonder when does the  quality practices begin.   In ancient days,  where a craftsman would be able to manage  all aspects of  transforming raw material to finished goods which will eventually be sold and he will also be the inspector of his own product.   An example would be a carpenter would choose a tree which would yield good wood to chop and he will transform  the wood into furniture.  He had a control on each step including inspecting his own work in the process making the part.  In this scenario  time to reach consumer is not a primary concern to finish the product.

Frederick Taylor (1856- 1915)  enabled mass production with his  principles of Industrial engineering   to building  products with higher  efficiency  and more systematic approach where product time to reach consumer become a primary concern.  Therefore mass inspection conducted by  trained inspectors is formulated in tandem with mass production to enable operator to focus on just transforming  the parts from raw materials or semi finished parts.  This could be the beginning of modern quality practices per American Society of Quality (ASQ).   Since then,  there are many quality philosophies being conceived over the last centuries by great quality gurus from America and Japan.  After world war II,  leading quality engineering experts Juran and  Deming had went to Japan to introduce quality management tools to help Japan to build up their manufacturing industry.  This had created a ripple effect where Japanese quality philosophies had been discovered.  Together all great gurus had contributed to the founding of modern quality management concepts in manufacturing industry.   

Modern quality practices adopted a lot of tools such as FMEA, 8D, advance product  quality planning (APQP)  from military and automobile  industry  as both  was the most advance industry hiring the best expert before the digital age.  With the invention of computer in digital age and the growth of semiconductor had created a need for more systematic quality improvement methods.  Thus systematic problem solving approach such as 6 sigma was conceived by Motorola in late 1980s can be used in continuous quality improvement.

We shall trace back the development of the modern quality techniques and philosophies using the table below. 

Quality culture contributors
Quality principles and techniques
Mass quality inspection by Frederick Taylor
Walter Shewart
Statistical Process control chart
Bell Telephone
1930s -1980s
William Deming
Sampling, PDCA, help Japan industry to grow through his quality philosophy,  14 key principles to transformation which is foundation of total quality management
America to Japan to global
1930s -1980s
Joseph Juran
Adopt Pareto principle in quality analysis, cost of poor quality,  Juran trilogy Quality control, improvement and planning, quality from top management, quality cannot be achieved thru inspection,   Juran Quality handbook
America , Japan to global
1960s -1990s
Philip Crosby
Quality management and maturity, Zero Defects,
1960s -1980s
Kaoru Ishikawa
Concepts of quality circle, Fish bone cause and effect diagram
1960s -1970s
Genichi Taguchi
The Taguchi loss function,  Taguichi method in Design of experiment

Advance product quality planning (APQP)
Semiconductor, Motorola
Quality improvement Six sigma methodology 
International Organization for Standardization
Standardized quality management systems to ensure product meet customer requirement

The biggest quality principles contributions which lead to today quality engineering strategy and techniques happen after  World War II with aggressive movement to build up nation economy at both sides of Pacific. 

With so many quality tools and principles being developed in last century,  it is sad that today  there are not many organizations are able to execute quality tools well.  Perhaps we should focus on how to apply all the quality tools properly to improve manufacturing process and shifting the quality management paradigm from inspecting quality in parts to build in quality to parts in a manufacturing process.

Sharing is caring

Continuous Improvement Program CIP - 6sigma Methodology