Measurement Systems Analysis - How To

Measurement System Analysis
How-to Guide
Version 6.1
August 2013
MSA How-to Guide
2013 Rolls-Royce plc

The information in this document is the property of Rolls-Royce plc and may not be
copied or communicated to a third party, or used for any purpose other than that for
which it is supplied without the express written consent of Rolls-Royce plc.
This information is given in good faith based upon the latest information available to
Rolls-Royce plc, no warranty or representation is given concerning such information,
which must not be taken as establishing any contractual or other commitment binding
upon Rolls-Royce plc or any of its subsidiary or associated companies.
2 | 2013 Rolls-Royce plc
Contents
INTRODUCTION
Introduction to Measurement System Analysis

Page 6
What is Measurement System Analysis?

Page 7
Different Types of Measurement System Analysis

Page 8
How does Measurement System Analysis work?

Pages 9 11
Choosing the appropriate type of study: Continuous or Attribute

Page 12
CONDUCTING A MEASUREMENT SYSTEM ANALYSIS
Step 1: Be Prepared
Pages 16 21
Ensuring the prerequisites are in place

Step 2: Planning the Study
Pages 22 27
Selecting appropriate parts for the study
Identifying appropriate people to involve in the study
Deciding when to run the study
Planning the data collection

Step 3: Conducting the Study
Pages 28 29
Collecting the data
Role of the observer

Step 4: Interpreting the Results:
Gauge R&R for Continuous Data

Pages 30 47
Analysing the data in Minitab
Interpretation of graphical output
Interpretation of numerical output
Communicating the results
Attribute Agreement Analysis for Attribute Data

Pages 48 60
Analysing the data in Minitab
Interpretation of numerical output
Interpretation of graphical output
Communicating the results

Step 5: Taking Action if the Results are Unacceptable
Pages 61 65
What to do if results of MSA are unacceptable
Common reasons for unacceptable measurement systems
Identifying improvement opportunities
Confirming the improvement

Step 6: Maintaining the Improvement
Pages 66 71
Ensuring the gains are sustained

APPENDICES
Continuous Data
1. Setting-up and Randomising the Spreadsheet in Minitab
2. Entering the Data in Minitab
3. Carrying out Gauge R&R in Minitab
4. Supplementary Information on interpreting the output
5. FAQ for Gauge RR
Attribute Data
6. Setting-up and Randomising the Spreadsheet in Minitab
7. Entering the Data in Minitab
8. Carrying out Attribute Agreement Analysis in Minitab
9. Supplementary Information on how to interpret the output
2013 Rolls-Royce plc | 3
MSA How-to Guide

How to Use This Guide
This How to Guide is designed as a complete training package that you can
work through individually at your own pace (or in small teams as part of a
facilitated training exercise).
By carefully reading the text, and practising the tools in the associated
Workbook, you will become competent and confident in using these process
control tools in your work area.
The How to Guide is designed to be applicable for use primarily by
Manufacturing Engineers and Lean Sigma Practitioners from any area of the
business. For this reason, the technical explanations are based on general
business application examples to ensure everyone can relate to them.
Throughout the guide, there are case study examples which show how the
theory is applied at the different stages of the process control sequence.
Before you start, make sure you also have the Workbook available. It is
essential that you work through this in parallel with the How to Guide, and that
you complete the practise questions, plus case study exercise before you start
to use MSA techniques in the business.
Icons are used throughout to highlight key elements, and to
signpost supplementary information where appropriate.
The technical
explanation of the
core terminology
F.A.Q
Indicating where you

must seek help from
practitioner experts
such as Black or
Green Belts
Where you can find
additional
information, and the
next phase of the
improvement journey
Indicates a key
learning point
The approved
Rolls-Royce answers
to main queries asked
by users
Tips on the commonly

observed pitfalls - &
how to avoid them
WORKBOOK
EXERCISE
The separate
Workbook which you
must use in parallel
with your learning
Guide Structure
The flowchart illustrates the structure of this Guide which is designed to provide step by
step guidance for conducting and interpreting the Measurement System Analysis. Steps
1, 2, 3, 5 & 6 below are common for all types of MSA study. Step 4 differs depending on
whether the study is for the collection of continuous or attribute measurements (see
page 14 for full guidelines on how to select the appropriate type of study).
Supplementary information on how to carry out the analysis using Minitab statistical
software together with some of the more detailed analysis of the statistical output is
provided in the appendices as outlined below.
Step 11
Be Prepared
Step 22
Plan the Study
Step 33
Conduct the Study
Appendix 1:
Setting up and
randomising the
spreadsheet in Minitab
Appendix 2:
Entering the data in
Minitab
Step 44
Type of Study
Appendix 3:
Carrying out
Gauge R&R in
Minitab
Appendix 4:
Supplementary
Information on
Interpreting the
Graphical
Output from
Gauge R&R in
Minitab
Continuous Data:
Gauge R&R for
continuous data
Attribute Data:
Attribute agreement
analysis for
attribute data
Step 55
Taking action if the results
are unacceptable
Step 66
Maintaining the improvement
Appendix 8:
Carrying out
Attribute
Agreement
Analysis in
Minitab
Appendix 9:
Supplementary
Information on
Interpreting the
Output from Attribute
Agreement Analysis
MSA How-to Guide

Section 1
Introduction
.................................................
In this section:
1.1: The need for Measurement Systems Analysis
1.2: What is Measurement System Analysis?
1.3: Different types of Measurement System Analysis
1.4: How does Measurement Systems Analysis Work?
1.5: Choosing the appropriate type of study: Variable or Attribute
.................................................
1.1: The Need for Measurement System Analysis
In our day to day work we often collect and use data to make important decisions about
our processes. For example we measure the critical dimensions of the parts we produce
to check that they are correctly manufactured; we inspect documentation and drawings
to check that they have been correctly completed; we test our engines to confirm that
they are functioning to specification.
Think for a moment about the data that is collected in your own work area. Regardless
of whether you work in a manufacturing, design or transactional function you will
undoubtedly be able to think of examples of data which are regularly collected to confirm
the quality of work, to monitor performance against targets or to allow in-process
decisions to be made.
Now ask yourself can you trust that data? Are you sure it is reliable? Are you sure that
the data is measured consistently? If more than one person or piece of measuring or
test equipment is involved then are you sure that if each were to measure/inspect the
same item that they would all reach the same conclusion?
If your answer to any of the above questions is no then there is a real possibility that
your measurement system could be producing unreliable data. This could lead your
team to draw the wrong conclusions about whether your processes are in control and
capable. Unreliable data can lead us to believe there is a problem with the process
when actually everything is OK or it may prevent us from spotting a problem. This is
likely to cost the business money, through unnecessary scrap or rework, or unnecessary
improvement projects, or through being unaware that there is a problem with a process
or product, and risking customer complaints or more serious problems such as safety
incidents.
Therefore, before collecting and using data to make decisions about any process or
product, it is important to check that the measurement system is good enough by doing
a Measurement System Analysis.
1.2: What is Measurement System Analysis?

Measurement System Analysis (usually referred to as MSA) is a
structured procedure which we use to assess the ability of a
measurement system to provide good quality data.
All data is collected using some form of Measurement System. A
measurement system is any system used to collect data. For example it can
be a piece of equipment (such as a clock), some software, or it could be a
person (such as an inspector recording the count of defects in a batch of
materials) or a combination of these.
A Measurement System is the combination of people, equipment,
materials, methods and environment involved in obtaining
measurements
Examples:
Data: time taken from receipt of an invoice until payment
Measurement System: a piece of computer software which measures the
time from start to finish, or possibly someone measuring the time using a
stopwatch or a clock
Data: Dimensions of a machined part
Measurement System: a vernier calliper, a coordinate measurement
machine (CMM) or a Go/No Go gauge in combination with the person
operating the piece of equipment, the procedures they are following and
environmental conditions such as the temperature or vibration level in the
working environment
Data: Number of errors on an engineering drawing
Measurement System: a person checking each drawing and counting the
number of errors in combination with the assessment standards, procedures
being used and environment conditions such as changes in the natural light
conditions.
Potential Pitfall: Where the measurement system is a device such as a
vernier calliper or CMM and you have multiple devices it is important to
recognise that each device is a separate system and each must
separately be analysed. For example just because one CMM machine
is a good or bad system does not mean that other identical machines
will have the same characteristics.
Turn to Exercise 1 in the Workbook to think about data and the
associated measurement systems within your own working area.
MSA How-to Guide

1.3: Different Types of Measurement System Analysis
Although the principles are similar, there are 2 main types of
Measurement System Analysis (MSA). The appropriate one to use
depends on the type of data being collected using the measurement
system.
Where the measurement system is collecting data which is measured on
a continuous scale such as time, weight, dimensions or pressure then
we will be collecting and comparing numerical results (such as the
weights of parcels in grams). This type of data is called continuous
data. In this case we use an MSA method called Gauge R&R
Continuous data comes from measurements on a
continuous scale such as: temperature, time, distance,
weight, dimensions.
Gauge R&R is short for Gauge Repeatability and Reproducibility. It
is the Measurement System Analysis method used to analyse
measurement systems which collect continuous data
Where the measurement system collects data which
categorises each item (such as pass/fail, grading items into
different categories or counting defects such as scratches in
paintwork or errors on a drawing) then we will be analysing
agreement in the categories chosen. This type of data is Called Attribute
data. In this case we use an MSA method called Attribute Agreement
Analysis
Attribute data is based on upon counting how many units fall into
discrete distinctions such as: pass/fail or percentage defective
Attribute Agreement Analysis is the Measurement System
Analysis method used to analyse measurement systems for
attribute data
Turn to Exercise 2 in the Workbook to check your understanding of

the difference between continuous and attribute data
1.4: How does Measurement System Analysis work?

Regardless of the type of MSA being used, there are two key things we
are usually interested in when we analyse a measurement system
1) If the same person or piece of equipment measures the same item
over and over again do we consistently get the same data? For example
if the same person measures the weight of a parcel 3 times would they
record the same weight each time? If a team member assesses whether
an expenses claim should be paid or rejected and looks at the same
claim 3 times would they be consistent in their decision?
This is important to know. If a person or a piece of equipment is not able
to consistently measure the same item then clearly this would result in
unreliable data.
This check for consistency within an inspector is referred to using
slightly different terminology depending on the type of measurement
being made.
For measurement of continuous data such as dimensions, weight or time
then this check for consistency is referred to as Repeatability
Repeatability assesses whether each person can measure the
same item multiple times with the same measurement device and
get the same value when measuring continuous data.
For measurement of attribute data such as the invoice checking example
above then this check for consistency is referred to as Within Appraiser
Agreement
Within Appraiser Agreement assesses whether each person can
assess the same item multiple times using the same measurement
method and get the same result when measuring attribute data
MSA How-to Guide

1.4: How does Measurement System Analysis work? (cont.)
2) If different people or pieces of equipment measure the same item would
they each get the same result? For example if three different people weighed
the same parcel multiple times would they each record the same weight? If
every member of the accounts team were asked to assess whether they
would pay or reject an expenses claim would they all draw the same
conclusion?
Again this is important to understand. Problems such as no fault found part
returns are often a result of a lack of agreement between two different
measurement systems. Obviously any disagreement indicates a problem
with the reliability of data collected from such a system
This check for consistency between inspectors is also referred to using
slightly different terminology depending on the type of measurement being
made.
For measurement of continuous data such as dimensions, weight or time
then this check for consistency between inspectors is referred to as
Reproducibility
Reproducibility assesses whether different people can measure the
same item multiple times with the same measurement device and get
the same average value
For measurement of attribute data such as the invoice checking example
above then this check for consistency is referred to as Between Appraiser
Agreement
Between Appraiser Agreement assesses whether different people
can assess the same item using the same measurement method and
get the same result when measuring attribute data

these definitions.
1.4: How does Measurement System Analysis work? (cont.)

Measurement System Analysis works by setting up and running a
controlled experiment to check the repeatability and reproducibility or
the within appraiser and between appraiser agreement of the system
Essentially a number of items (such as machined parts or drawings for
inspection) are selected and deliberately measured multiple times by
each person or piece of equipment. We then analyse the results using
Minitab software (full details on how to do this are provided in the
appendices), to establish how consistent the measurement system is,
and to decide whether improvements need to be made before using it to
collect data.
For example, to run such an experiment in the parcel weighing example
you would select a range of parcels (around 10 would be sufficient in this
case) and get a number of different people who work in the area to each
weigh the parcels a number of times. As we will explain in section 2
usually three different people are selected to make the measurements
and each item is measured by each person three times. This provides
sufficient data to be able to check for consistency both within and
between those people making the measurements.
If a problem is found with the measurement system then the next step is
to find the root cause of the problem and try to eliminate as much
variation as possible within the measurement system. In the parcel
weighing example differences between operators may exist due to them
having different methods for using the scales or due to them interpreting
or recording the weight display in different ways.
Teams are often surprised to discover that measurement systems they
have used for years are actually inconsistent. Carrying out formal
Measurement System Analysis is a good structured way for teams to
assess and improve their measurement system.
You may well already make informal checks of this kind. However using
the formal structured approach provided by Measurement System
Analysis allows us to be consistent and thorough in the way that we
check the reliability of our data.
MSA How-to Guide

1.5: Choosing the Appropriate Type of Study
What sort of
data do I have?
Continuous
Attribute
Use Gauge R&R
Use Attribute
Agreement Analysis
The flow chart above can be used to select the types of measurement
study to conduct
SEEK
GUIDANCE
If you are not sure what type of data your variable is then ask a local
Black Belt to help you choose the most appropriate type of MSA.
Summary
A Measurement System is the combination of people, equipment,

materials, methods and environment involved in obtaining
measurements.
If left unchecked there is a real risk that the measurement systems
we use could be delivering unreliable data without us being aware
of the existence of a problem. Therefore it is important to analyse
how effective our critical measurement systems are.
Measurement System Analysis is a structured procedure which
we use to assess the ability of a measurement system to provide
good quality data, before using it to collect data on which decisions
are based.
Measurement System Analysis provides a structured approach
for teams to assess whether the measurement systems they use
are fit for use.
There are 2 main types of Measurement System Analysis:
Gauge R&R, which is the method used to evaluate
measure systems for continuous data
Attribute Agreement Analysis, which is the method
used to evaluate measurement systems for attribute data
the key points so far.
Section 2
Conducting an MSA
.................................................
In this section:
A step-by-step guide for conducting an MSA
.................................................
The key steps we will cover are as follows:
Step 1
Be Prepared
Step 2
Planning the Study
Step 3
Conducting the Study
Step 4
Interpreting the Results for Continuous / Attribute Data
Step 5
Taking Action if the Results are Unacceptable
Step 6
Maintaining the Improvement
MSA How-to Guide

Case Study Examples
We will introduce you to the application of these steps using two example
scenarios which you can work through one for a measurement system
which collects continuous data and one for a system which collects
attribute data. You will then have the opportunity to turn to the workbook
and practise applying the tools yourself using further case study examples
Case Study - Scenario 1
Julie is the production manager of a factory where mobile phones are manufactured and
assembled. There has recently been a noticeable increase in mobile phones being
returned by customers because the touch screens arent functioning correctly. She
speaks to the failure analysis team, who examine every phone which is returned, and
they have done some analysis which indicates that the screens are being manufactured
with an incorrect thickness.
Julie sets up a team to investigate this this is a major issue for the company, as every
phone which is returned needs to be replaced at significant cost. She is also concerned
that the reputation of the companys phones will be damaged, which may quickly result
in loss of business to their competitors.
On initial investigation, the team cant find a reason for the screen thickness being
th
incorrect. The thickness (in millimetres) of every 10 screen made is measured, using a
vernier calliper, by one of the experienced measurement technicians. The data collected
over the last few months indicates that there has been no issue with screen thickness.
Julie and her team cant understand why there seems to be a problem with screen
thickness, yet all of their measurements show that the thickness is fine. One of the
engineers suggests that perhaps there is an issue with the measurement of screen
thickness perhaps the data being collected is incorrect or unreliable in some way.
The measurement technicians dont believe there can be an issue with the
measurements, as the measurement technique is so straightforward and they all have
several years experience in making the measurements. However, the team decides to
carry out an MSA to eliminate this as a possible root cause.
Since the data being collected is continuous data (screen thickness measured in
millimetres), the team recognise that they need to conduct a Gauge R&R study to
determine repeatability and reproducibility.
Case Study Scenario 2

Tom works in the Quality department of the same major mobile phone
manufacturer as Julie. There has been an increase in the number of
mobile phones being rejected at the final test stage due to scratched
screens. Tom cannot understand why this is happening, as each screen is
inspected for scratches before assembly, and immediately has a protective
film placed over it. This inspection is referred to as the in-line inspection,
and there has been no increase in the number of scratches detected at the
in-line inspection.
This issue results in around 5% of manufactured screens being scrapped.
The screens are unable to be reworked adding a significant cost to the
company. There is also an additional cost of fitting a replacement screen to
each of the affected phones.
Tom feels that there might be an issue with the in-line inspection. The
inspection is done by a team of quality technicians. They do a visual check
of every screen manufactured, by placing it under an inspection lamp and
looking for scratches. Each screen is then categorised as a Pass or a Fail
and the results are recorded by the technicians in a database. Tom
discusses the issue with the technicians, however they are adamant that
there isnt a problem with the inspection as it has always worked well in the
past. Nevertheless, Tom feels it would be worthwhile to conduct an MSA,
so that inspection can be eliminated as a potential root cause.
The measurement system in this case consists of the person doing the
inspection, plus the inspection lamp, methods, acceptance criteria being
used and any environmental factors.
As the data being collected is either a Pass or a Fail, this is attribute data
and therefore Attribute Agreement Analysis is the appropriate technique to
determine the within and between appraiser agreement.
In both these measurement systems the parts are measured or inspected
by people. Throughout this guide we will call these people the appraisers.
Appraisers: these are the people who measure the parts
Remember repeatability and within appraiser agreement are asking the
question do the appraisers agree with themselves when asked to do the
same assessment a number of times? Reproducibility and between
appraiser agreement are asking the question do the appraisers agree
with each other when asked to do the same assessments?
MSA How-to Guide

Step
1
Be Prepared
Ensure the pre-requisites are in place
We must remember that this activity builds on the process basics and it is
important to ensure that these foundations are in place.
1a) There are then a few more prerequisites that we should check:
Check If the measurement system involves a piece of equipment, we must ensure
equipment that, where relevant, the calibration is up to date.
calibrated
If the calibration is not up to date, there is a risk that there will be
some Bias in the measurements, which means that the results will be
different than their reference value. Bias is often also known as
accuracy. An example of bias is if your bathroom scales are
incorrectly set up (incorrectly calibrated) so that they consistently overestimate your weight by a set amount such as 1kg.
For visual inspections it is also important to check that the appraisers have
current eye examinations that meet the required standards.
The equipment having a current calibration sticker does not guarantee
that it is measuring correctly. Bear in mind that the equipment may
have been dropped, otherwise damaged or tampered with since it was
last calibrated. Inspect the gauges carefully for any signs of
damage and if in doubt contact your local Metrology department for
advice
1b) Resolution is used to assess the increments of measure that a particular

Check the gauge can work to when measuring parts.
equipment
resolution Where we are measuring continuous data it is important to check that
measurement equipment has adequate resolution for the characteristic
being measured.
For example, a set of bathroom scales are often marked in 1kg increments
so would not have sufficient resolution for weighing baking ingredients if
measurement was required to the nearest 10g.
As a guide, a suitable resolution for the measurement system is usually
1/10 of the tolerance or better. So for example if you needed to be able to
measure baking ingredients out to be 100g +/- 5g then the tolerance would
be 10g meaning that the required resolution of the measurement system
should be 1g. In this case we would need to have scales which could
measure to increments of 1g.
For attribute agreement ensure that reference standards are available and
are able to distinguish between pass and fail conditions.
Reference standards provide clear guidance for appraisers to allow
them to distinguish between pass and fail conditions. These could be
written procedures, pictures or physical samples of parts which display
both the pass and fail conditions.
1c)
Check
measurement
system is
suitable
1d)
Check
equipment
is stable
For continuous data it is also important to ensure that the method being
used is suitable for the full range of measurements you will be making for
example that it is equally suitable for measuring the smallest and largest
parts. For example, when weighing airline baggage, bathroom scales
would probably be suitable for weighing small holdalls, but would they be
suitable (and give reliable results) for extra large suitcases?
We must also be confident that the measurement system is stable over
time that is, there is no reason to expect that the performance of the
measurement system will vary over time. One way to check this is to
measure the same part at regular intervals, and plot on an SPC chart.
MSA How-to Guide

For more information on SPC Charts see the How To Guide for
SPC Charts.
Checking the equipment is calibrated, has suitable resolution

and range and is stable over time
Julie and her team worked with their local Black Belt (Alison) to
ensure that the pre-requisites are in place:
The vernier calliper used to measure the screen thickness has

been calibrated within the last year, which is fine.
The specification limits for screen thickness are 1.5 2mm, so

the tolerance is 0.5 mm. For the vernier calliper to have sufficient
resolution, it must be able to measure in 0.05mm increments - it
can actually measure down to 0.01mm, so meets this
requirement easily.
The team are satisfied that the measurement system is suitable

for the full range of measurements.
The team are also satisfied that the vernier calliper is stable
they have what they call a reference standard screen, which is
stored in a cupboard and measured once per week (it is always
measured by the same technician). The results from this
standard screen have been very consistent (stable) over the
past year (the data is plotted on an SPC chart so they can see
this at a glance)
Continuous Data
Checking the equipment is calibrated, has suitable resolution

and range and is stable over time
Tom carries out the following checks to ensure the pre-requisites are
in place:
Each appraiser has current eye examinations to the company

standard.
Reference standard screens are available, represent the full

range of expected defects and have current calibration stickers.
The reference standards are in good condition
The inspection lamps have all got stickers confirming that their
lux levels have been checked in the last 12 months
Attribute Data
1e)
Ensure
clear
operational
definitions
Without an operational definition, there will almost certainly be

misunderstandings and inconsistencies in the way that data is collected
using the measurement system.
An Operational Definition is a clear description of:
What to measure; ensuring that all aspects of the measure are

carefully defined.
How to measure it; including what equipment to use, how to use

the equipment or standards and how to record the data.
Dont assume that its obvious to everyone how to make the
measurement!
Even measures that seem obvious are open to interpretation:
For example, when making measurements using a ruler, does

everyone start measuring at the point where zero is marked, or might
some people start from the end of the ruler?
If asking people to decide whether something is defective or not,

does everyone have the same understanding of what defective
actually means? For example, if you were assessing the quality of
the paintwork on a car door, some people might feel that a few tiny
scratches is acceptable, whereas others may class the door as
defective if it has any scratches at all.
Ensure everyone has been trained in how to follow the operational

definition
Turn to Exercise 5 in the Workbook to practise writing operational
definitions.
MSA How-to Guide
Attribute Data
Case Study Attribute Scenario - Operational Definitions

Tom sets up a team, including Alan, the Quality Technician team
leader, to discuss running the Attribute Agreement Analysis. The team
check that there is an operational definition in place for the in-line
inspection, and in fact there is as follows:
Review the reference standard screens
Wearing gloves hold the bottom left hand corner of the screen
using the silicon coated tweezers supplied.
Do not touch any other part of the screen
Hold the screen 5 centimetres below the inspection lamp.
Tilt the screen in all directions if any scratches are seen,

contact the production engineer immediately.
The team feel that this operational definition is clear and they feel that
they can proceed with the study.
KATE:
Sales
Continuous Data
Case Study Continuous Scenario - Operational Definitions

Julies team realise there is no operational definition for the
measurement of the screen thickness, so they produce one:
The screen should be held between the thumb and

forefinger of one hand.
Set the vernier calliper to zero
The vernier calliper should be used to measure the thickness at a

point in the middle of the screen.
The thickness should be recorded, in millimetres, to two decimal

places.
1f)
Observe trial
measurements
Finally, observe a few people making measurements, and note down any
observations. Does it look like the operational definition is being followed?
Is it confusing? Are there any misunderstandings or discrepancies?
If any issues are found during this observation, it is worth addressing these
before moving on to conduct the MSA.
KEY POINT
When doing this observation, and also when conducting the study, it
is important to clearly communicate to those involved that it is the
Measurement System as a whole which is being assessed, not the
individual people involved. To ensure that this is clear, peoples
names should not be used within the study, rather they should be
referred to as Person A, Person B etc, or something similar
(ensuring that it is still possible to trace back to the real names where
required). It is strongly advisable to think carefully about how to
communicate the purpose of the MSA to the team. Where possible
always try to communicate with the team face to face in a team
meeting
Attribute Data
Initial Observations
Over the next few days, Tom and Alan observe the technicians
inspecting screens on a few occasions as far as they can see they
are following the operational definition.
Continuous Data
Initial Observations
There are 3 measurement technicians in Julies team Kevin, Mary
and Karen they will all participate in the study.
Julie observes each of them doing a few measurements, and although
it appears that the 3 technicians are all following the same procedure,
she feels that it is quite a difficult measurement to make. However, the
technicians all say that they find it quite easy, so she decides to
proceed with the Gauge R&R.
MSA How-to Guide

Step
2
2a)
Selecting
appropriate
parts for the
study
Planning the Study

2a) Selecting appropriate parts for the study
2b) Identifying appropriate people to involve in the study
2c) Deciding when to run the study
2d) Planning the data collection
Once we are confident that the prerequisites are in place, we can begin to
plan the MSA.
As we discussed in section 1, the principal of MSA is that we set up an
experiment to measure several parts several times, and use statistical
analysis (with the help of the Minitab software package) to look at how
much variation there is in the results. We can then decide whether this
amount of variation is acceptable, and if it is not, we can do further
analysis to understand where the variation is coming from, and help us find
the root cause(s).
The first step in planning an MSA is to consider which parts (or items) we
are going to measure during the study. Remember that MSA can be
applied to anything from manufactured parts to engineering drawings to
invoices. For simplicity in this guide in all cases we will call the thing that is
being measured the part.
Part: this refers to whatever is being measured or assessed. For
example this could be a physical component which is being measured,
a telephone call which is being timed or a document which is being
checked for errors.
A common standard is to use a minimum of 10 parts for a Gauge R&R
Study and minimum of 20 parts or samples for an Attribute Agreement
Analysis. In both cases a standard approach is for each part to be
measured by 3 different people, 3 times each a total of 90
measurements.
Ideally, we want to select a range of parts some from the lower end of
the process range, some from the upper, and some in-between. Also
include some borderline parts if possible (ones which are marginally inside
or outside the required tolerance limits or standard). The reason for this is
that we want to ensure that the amount of variation in the measurement
system is the same across the whole range of parts, and if it isnt, we want
to be able investigate why.
Potential Pitfall: Be careful not to confuse the process range with
the customer specification. The process range is the full range of
parts produced by the process. This may be wider or narrower than
the tolerance specified by the customer. A failure to use the full
process range will result in the measurement system analysis not
being fully representative of the actual measurement system. If you
have concerns about selecting a fully representative sample then
contact your local Black Belt for advice.
SEEK
GUIDANCE
If you are unsure as to which parts to choose for the study, or if its
not possible to obtain 10 parts, contact a local Black Belt for
guidance.
Also seek guidance if it isnt possible to measure the same part
more than once.
MSA How-to Guide

2b)
Identify
people to
involve
The next step is to identify the appropriate people to involve in the study.
First you need to choose the appraisers the people who will be
doing the repeat measurements of the parts. These should be
people who normally make the measurements. The standard
approach for MSA is to use 3 appraisers
The 3 people should represent everyone who normally do the
measurements therefore, if there are several teams or shifts
who do the measurements, try to ensure they are represented.
It is also important to have an observer (or facilitator) someone
who wont take part in the actual study, but who will observe the
study, coordinate the measurement of the parts and note the
results. Their role will be discussed later in this section.
If possible, arrange to have some help from a Black Belt when
planning the study, and also when analysing and interpreting the
results (especially the first few times you conduct a Gauge R&R).
It is recommended that when running an Attribute Agreement

Analysis wherever possible to have an expert assess each of the
parts, in addition to the three appraisers we have already mentioned.
Where it is possible, use an expert (or a group of experts) to assess
each part and determine the correct result for each. The expert
decision is referred to as the Standard. This can then be used to
assess the level of agreement between the appraisers and the
agreed standard.
SEEK
GUIDANCE
If unsure which 3 appraisers to choose, or if you feel that it might be

better to have more (or less) than 3 people (e.g. if you have 4 teams
doing the measurements), discuss with a local Black Belt.
If you are unsure how to determine the standard result for attribute
data, also discuss with a local Black Belt.
Selecting Parts for the Study

Continuous Data
With Alison the Black Belts help, Julie and the team select 10 screens
from those processed during the previous 24 hours they select every
th
20 screen processed, as they know from their knowledge of the
screen manufacturing process that this is likely to cover the full range
of thicknesses typically produced, from 1.5mm to 2.0mm. They agree
that Kevin, Mary and Karen will each measure each of the 10 screens
3 times.
The team are now looking forward to conducting the study and ready
to begin.
Selecting Parts for the Study

Three of the technicians who normally do the in-line inspection are
chosen to take part in the study Steven, Jim and Emma. These 3
technicians are chosen from a possible 8, as they represent all 3 shifts
and are of varying levels of experience.
Again with the assistance of Alison the Black Belt Tom selects a
sample of 20 screens from those recently processed. They ensure that
a selection of good and bad screens are chosen, including a few
which are clearly borderline (i.e. quite difficult to decide whether to
pass or fail).
Attribute Data
Although it isnt possible to label the screens, Tom feels that he will be
able to keep track of them as an observer. He puts the screens into
numbered boxes between measurements, and decides that he will
randomly introduce them into the inspection queue when the study is
being carried out. Alison agrees to assist with the observation in that
she will monitor the database and extract the results as they are
entered.
Tom and Alison decide that each technician will inspect each screen a
total of 3 times.
In order to generate a standard set of results, they also decide that
Mark, the final test engineer will assess each of the screens once he
is seen as the expert in determining whether a screen is actually a
pass or a fail. Marks assessment will be the standard against which
the others will be compared to.
MSA How-to Guide

2c)
Deciding
when to
run the
study
We also need to consider when to run the study. In some cases, running
the study will be quite straightforward, and it can simply be done at a time
which suits the appraisers and the observer. However, it is important to
consider the following when arranging a time:
Will you be using any software, databases etc which might

interfere with the normal running of the process if so, could you
run the study at the end of the day, or before the process starts in
the morning?
How long will it take to assess each part? Will you need to plan
the study over a few days, or can it all be done at once?
We will later discuss the need for the appraisers to measure the
parts for a second and third time without being able to remember
their previous results this may mean that the study needs to be
spread over a few days, rather than the 3 sets of measurements
being done in a single day.
Continuous Data
Deciding When to Run the Study

Julie and the team estimate that the study will take a maximum
of 2 hours to run. They decide to do the first set of measurements in
the morning, the second set immediately after lunch and the third set at
the end of the day.
2d) We can now start to plan the data collection:
If possible, set up a Minitab worksheet, or excel spreadsheet, into

Planning
which the observer can type the data directly to save time and
the data
possible errors.
collection
If this isnt possible, print out a blank Minitab worksheet for

recording the data.
The worksheets for Gauge R&R and Attribute Agreement Analysis

each have their own layout.
For detailed guidance on how to set up and understand the layout of
Minitab worksheets for Gauge R&R and Attribute Agreement Analysis
go to Appendix 1
2e)
Randomisation of
the study
When Minitab is used to set up the worksheet then it will automatically

randomise the data. This means putting the measurements into a random
order.
Randomising the parts before each of the sets of measurements is
important, as this ensures that if there are any time-related or
environmental causes of variation in the measurement system, these will
be spread randomly throughout the study.
For example, some possible time-related changes could include people
getting tired towards the end of the study or equipment heating up as the
day progresses.
Where Minitab is not available to randomise the order of measurement
then the random order can be determined simply by writing the numbers 110 or 1 20 on pieces of paper, and selecting them one at a time from a
hat or a box.
For detailed guidance on the options of randomising the data in
Minitab worksheets for Gauge R&R and Attribute Agreement Analysis
go to Appendix 1
SEEK
GUIDANCE
If you are unsure about using Minitab or randomising the study then
contact a local Black Belt for guidance.
MSA How-to Guide

Step
3
Conducting the Study

3a) Collecting the data
3b) Role of the observer
3c) Enter the data
3a) Now we have the parts, the appraiser and the data collection sheet in place
Collecting we are ready to conduct the MSA study.
the data
Following the randomised order created on the worksheet each part should
be measured /appraised three times by each of the appraisers.
It is critical that the result of each measurement is not influenced by
previous measurements the appraisers must not know the results for
their previous measurements or for the measurements of the other
appraisers otherwise they may accidentally bias the results (i.e. change
their results so that they match the previous ones).
Ideally the appraisers should be unaware that they are even measuring the
same parts or items more than once this is often described as the
measurements being done blind. The observer has an important role to
play in this, which well discuss on the next page.
Attribute Data
Randomise the Study

Tom and Alison also use Minitab to randomise the order of
measurement of the screens, and are confident that measurements will
be done blind as the technicians are unlikely to realise that they are
inspecting the same screen more than once.
Continuous Data
Randomise the Study

Julie discusses how to randomise the study with Alison (the Black
Belt). They decide to put each of the screens in a numbered box
(numbered from 1-10). Alison will be the observer in the study and will
hand the screens to the technicians for the measurement one at a
time, to ensure that the technicians dont recognise them from previous
measurements. They decide to use the Minitab worksheet to
randomise the order of the trials.
3b) The observers role in conducting the MSA is key. They will be required to:
Role of the
Plan, or be involved in planning the study.
Identify which parts/items are which, in such a way that the appraisers
Observer
arent aware of any labelling or marking. This can be done in a number
of ways:
If physical items are being assessed (e.g. manufactured parts,

drawings or invoices), label each part discretely so that only the
observer can see the label. Alternatively, for smaller items, place
each item in a labelled box, and pass the items one-by-one to the
appraisers without the label being seen.
For items such as recorded calls, save each call in a separate file
or on a separate tape or CD, and again ensure that the appraisers
cant see the label/filename.
Ensure that the items are in random order before each set of
measurements begin. Pass the items to the appraiser one-by-one for
measurement
Note the results, ensuring that none of the appraisers can see the
results, or hear each others results.
Enter the results into the Minitab worksheet immediately if possible.
Ensure that the appraisers are referred to as A, B and C, or 1, 2 and 3

names should be noted (to allow for investigation of any unusual results)
but not published.
Note any comments during the study anything of note which happens
while the measurements are being made. For example:
If the measurement equipment or software has a fault or error

during the measurement and needs to be corrected or replaced
If one of the parts or equipment is dropped
If the appraiser has any problems in making the measurement or if

there are any noticeable differences in the way the appraisers
make the measurements
If there are any interruptions or distractions
If there are any obvious opportunities for improvement.

3c) During or immediately after the study the data should be entered into
Entering Minitab.
the data Whilst it is most efficient for the data to be directly entered into Minitab it is
important to be aware that it is quite easy to accidentally delete, mix up or
mistype data. It is therefore advisable to write the measurements down
prior to entering them into the Minitab worksheet
For detailed guidance on how to avoid the common pitfalls associated
with entering the data into Minitab please go to Appendix 2
MSA How-to Guide

Step
4
Interpreting the Results for Continuous Data

4a) Analysing the data in Minitab
4b) Interpretation of graphical output
4c) Interpretation of numerical output
4d) Communicating the results
Now the data has been collected we are ready to start to analyse, interpret
and communicate the results
As the methods for analysis are very different for Gauge R&R and for
Attribute Agreement Analysis we will take each in turn using Julie and Toms
case studies as examples.
We will begin with Gauge R&R (for continuous data) and then move on to
Attribute Agreement Analysis (for attribute data). If you only wish to follow
the procedure for Attribute Agreement Analysis then please turn to page 48
now.
Case Study Data Analysis

Open the file Julies GAUGE R&R DATA 1. This is the data
collected from Julies measurement study.
To start with it is worthwhile just to look at the results. Can
you see any difference in the results:
With the same technician measuring the same screen

repeatably?
With different appraisers measuring the same screen?
Continuous Data
4a)
Analysing
the data in
Minitab
For example part 5

measured by
appraiser 1 (Kevin)
was measured at
1.55mm whereas
this part measured
by appraiser 2 (Mary)
was measured at
1.41mm
To learn how to use the Minitab Gauge R&R function to analyse this
data turn to Appendix 3
MSA How-to Guide

Minitab uses a technique called Analysis of Variance (ANOVA) to analyse
the data collected during the study.
If we have a measurement system which is performing well, and
consistently, there will be very little difference between the repeated
measurements made on each of the parts during the study we describe
this as there being very little variation in the measurement system.
If the measurement system is not performing well, there will be noticeable
(or significant) variation i.e. there will be a difference between the
repeated measurements made on each of the parts.
The ANOVA technique quantifies the amount of variation in the
measurement system so that we can easily decide whether the
measurement system is acceptable for use or not (using some guidelines
which we will discuss shortly.)
Minitab also analyses the data in a such a way that, if there is an
unacceptable amount of variation in the measurement system, we can
easily see where that variation is coming from, by using the graphs which
are produced, which will help us find the root cause(s) of the variation
between measurements.
Minitab can either analyse the data and produce a numerical output or a
set of graphs. Both are used in tandem to analyse the data. Each of these
methods will be explained on the following pages.
4b) First of all we will look at the graphs which are produced by Minitab. We
Interpret will look at each of the graphs shown below in turn.
graphical
output
For details on how to set up and produce the graphs in Minitab see
Appendix 3
Further information on interpreting each of the graphs is given in
Appendix 4
SEEK
GUIDANCE
If you are unsure as to how to construct or interpret any of the

graphs, ask for help from a local Black Belt.
MSA How-to Guide

The first graph to look at is the Components of Variation graph.
This is the most important of the graphs to look at as it tells us how well the
measurement system is performing overall. For a good measurement
system, we would expect the Gauge R&R and the Repeat and Reprod bars
all to be very small compared to the Part-to-Part bar. This indicates that
most of the variation seen in the study comes from genuine variation
between the parts rather than from variation due to the repeatability or
reproducibility of the measurement system. If there is a problem with the
measurement system then the Gage R&R bar will be relatively tall
representing 30% or more on the left hand scale.
In cases where the Gage R&R bar is too tall (as in the example above), we
can also use this graph to help us determine whether the problems with the
measurement system are due to Repeatability, Reproducibility or both. We
look to see whether one of the bars labelled Repeat and Reprod is
noticeably taller than the other, or whether they are similar in height.
If the Repeat bar is taller, this indicates that there is an issue with
Repeatability.
If the Reprod bar is taller, this indicates that there is an issue with
Reproducibility
If they are a similar height, this indicates that there is an issue with
both Repeatability and Reproducibility.
They also see that the Reprod bar is quite a bit taller than the Repeat
one this suggests that most of the causes of the measurement
system being poor are likely to be related to Reproducibility.
Continuous Data
Interpreting Julies Data The Graphical Output

Julie and her team see that the Gauge R&R bar is quite tall compared
to the Part-to-Part bar, which indicates that a significant proportion of
the variation they are seeing in their measurements is coming from
variation in the measurement system.
They continue to look at the rest of the graphs to see if they can find
some clues as to what the issues are.
Turn to Exercise 6 in the Workbook to learn more about interpreting

the components of variation graph
For further information on interpreting this graph, refer to Appendix 4
MSA How-to Guide

Next we will look at the R Chart by Appraiser.
This chart shows the Range of the results for each appraiser for each of the
10 parts.
There are 3 segments in the graph one for each appraiser.
Appraiser 1
Appraiser 2
Appraiser 3
Parts 1-10
In each appraisers segment, the
range is plotted for each of the
parts measured. (You will see
the part numbers (1-10) along
the bottom axis, repeated 3
times once for each appraiser)
So in the above example, the
first point plotted is the range of
Kevin (appraiser 1)s results for
part number 1.
The range is the maximum value minus the minimum value, so if we look at
Kevins results for part number 1 in the Minitab worksheet, we can see that
his measurements were 1.63, 1.60 and 1.60.
So the range was 1.63-1.60=0.03. We can see this plotted as the first point
on the graph.
Interpreting the R-Chart by Appraiser

We interpret this chart by saying that if any of ranges is particularly large
(above the upper red line) then that indicates there is inconsistency in the
measurements made by that particular appraiser on that particular part .
High ranges are often due to typing errors check for this first.
Continuous Data
Interpreting Julies Data The Graphical Output

On studying the R Chart by Appraiser from their study, Julie and her
team can see that there are two points with high ranges (above the
upper red line ) - these correspond to screen numbers 2 and 8 when
measured by Mary (appraiser 2). They check for typing errors, but are
confident that the data has been entered correctly. They note this
observation and move on to look at the next graph.
MSA How-to Guide

Next we will look at the Xbar Chart by Appraiser.
As with the previous chart, there is a segment for each appraiser. In each
appraisers segment, the average measurement for each of the parts is
plotted.
So in the above example,
the first point plotted is the
average of Kevin (appraiser
1)s results for part number
1.
Looking at Kevins results
for part number 1, we can
see that his measurements
were 1.63, 1.60 and 1.60,
therefore the average is
1.61. We can see this
plotted as the first point on
the graph.
Interpreting the Xbar chart by Appraiser

If the measurement system is good, and the appraisers are consistent with
each other, we would expect to see at least 50% of the points either above
or below the red lines (and very few points between the red lines).
Continuous data
Julie notes that most of the parts are outside the control limits and so
moves on to look at the next graph
Although this chart is an X-Bar Chart it is NOT being used in this

context to assess control. The red lines in this case are NOT process
control limits but rather represent the variation due to the measurement
system. Therefore in this case it is desirable for the points to be outside
the red lines.
MSA How-to Guide

Next we will look at the Measurement by Parts graph.
9 measurements
and average value
for part number 10
The scale of the Y-axis for this graph is not labelled in Minitab but represents
the measurement (in this case the measurement in mm for each part)
This graph shows a circle for each measured value of each part (i.e. all
measurements made by all appraisers). So in the case study example there
are 9 circles for each part, as each part was measured 9 times in total. The
average value for each of the parts is also shown by a crossed circle.
Interpreting the Measurement by Parts Graph

If the measurement system is consistent, there should be very little
difference between the measurements for each individual part (in other
words, the circles for each part should almost be on top of each other or
overlapping).
If any of the parts has a noticeably larger spread in results than the others,
this particular part might be worth investigating.
Similarly any parts with noticeably less spread in results than the others
(such as part 7 below) may also be worth investigating as this can also
indicate something unusual within the measurement system.
Julie and her team notice from this graph that screens 4, 5, 8 and 10
appear to have a greater spread than the rest of the screens.
The team need to consider why these screens were more difficult to
measure? They note this finding and move on to the next graph.
Continuous Data
MSA How-to Guide

Next we will look at the Measurement by Appraiser graph.
Variation
In this graph, the crossed circles show the average of all of the results for
all of the parts as measured by that appraiser. So for appraiser 1 (Kevin),
the average of all of his measurements for all 9 screens was 1.84, as
shown.
There is also a box plot for each appraiser the represents the amount of
variation in the results for all of the parts as measured by that appraiser.
The middle 50% of the variation is represented by the height of the box and
the full range of variation by the full length of the plot.
Interpreting the Measurement by Appraiser Graph
If the measurement system is consistent, we would expect the average
values to be similar (as we would expect all appraisers to get very similar
results for all of the parts). If this is the case, the line connecting the
averages would be perfectly horizontal.
For the same reason, we would also expect the spread for each appraiser
(represented by the size of the boxes and the length of the whiskers) to be
similar.
If the line connecting the averages is not a straight line (i.e. if any of the
appraisers has either a noticeably higher or lower average than the others),
this suggest that they are perhaps doing something consistently different to
the others (poor reproducibility)
If any of the appraisers has a noticeably larger variation than the others,
this may indicate that that they doing something inconsistent from one
measurement to the next (poor repeatability)
Again Julie and the team note their findings and move onto the final
graph.
Continuous Data
Julie and her team notice that appraiser 2 (Mary) has a slightly larger
spread of results than Kevin and Karen, and also a lower average (the
connecting line is not horizontal), which suggests that she might be
doing something differently.
MSA How-to Guide

Finally, we will look at the Appraiser*Part Interaction graph.
This graph overlays the average

measurements for each item as
measured by each appraiser. So for
screen number 1, Kevins average was
1.61 (dot), Marys was 1.57 (square)
and Karens was 1.54 (diamond). The
average measurements for each
appraiser are connected by straight
lines.
Interpreting the graph
If the measurement system is perfectly consistent, we would expect all of the
points for each part to be on top of each other (similar to part 2 in the
example above), and the lines connecting would be parallel with each other.
We interpret this graph by saying that if the lines connecting the averages
are not parallel, and there is a noticeable spread in the points for any of the
parts, it is worth investigating why the measurements for these particular
parts are less consistent than the others.
Continuous Data
Julie and the team notice from this graph that the points are
noticeably separate for screen numbers 4, 5, 8 and 10, and the
connecting lines arent parallel this confirms what they saw in some
of the previous graphs.

4c)
Interpretation of
numerical
output
Interpreting the Numerical Data

In addition to the graphical output Minitab also provides us with a lot of
numerical data. However the only output we are interested in for now is the
Total Gage R&R % Tolerance (as highlighted on the session window
below).
Interpreting the Numerical Data

As we have seen Minitab provides us with a lot of numerical data.
However the only output we are interested in for now is % Tolerance
(as highlighted on the session window below).
In the above example from the case study, you will see that the Total Gage
R&R %Tolerance is 76.99% this tells us that the variation in the
measurement system is 76.99% of the process tolerance (upper tolerance
limit-lower tolerance limit). [The process tolerance for the screen is 2.0mm
1.5mm = 0.5mm]
MSA How-to Guide

How do we decide whether this is acceptable or not?
The guideline for whether a measurement system is acceptable or not, is
* This is a guideline
criteria remember to check
the latest version of the
SABRe standard for the
limits for your application.
This standard also sets
out the requirements for
bias which are outside the
scope of a gauge R&R
study.
SEEK
GUIDANCE
If the %Tolerance is outside the acceptance criteria, the results should

be discussed with a local Black Belt, who will help you assess the
risks of continuing to use the system remember, making decisions
with an unreliable measurement system can cost your organisation a
lot of money.
The other numerical output which we are interested in is the Number of
Distinct Categories.
An acceptable measurement system will have 5 or more distinct categories.

This means that it is able to discriminate between different parts 5 distinct
categories means that it can discriminate between 5 different sizes of parts
across the total study variation
If the discrimination is poor, the gauge is poor at distinguishing parts from
each other and will only classify them as, for example, high and low (2
distinct categories) , or high/medium/low (3 distinct categories). This is not
adequate for process control and analysis.
Julie immediately contacts her local Black Belt Alison for advice on
how to proceed.
Continuous Data
Interpreting Julies Numerical Data

So, as the % tolerance of 76.99% is greater than 30% and the number
of distinct categories is <5, Julie and her team conclude that the
measurement system is UNACCEPTABLE and in need of
improvement.
4d) The results of the Gauge R&R can now be communicated as necessary.
Communicating
results Try and keep the communication simple include the %Tolerance results,
along with an explanation that >30% is deemed unacceptable, and that the
results will need to be discussed with a Black Belt.
Also include one or two of the graphs in the communication dont include
too many though, choose the ones which clearly show what the problem is,
and the ones which you find easiest to explain (as you will have noticed, in
many cases the same problem will be highlighted by several of the graphs)
Summary Interpretation of Minitab Output

Julie and the team get together to consider the results of the
study and to discuss possible opportunities for improvement
Continuous Data
Their summary of the study is:
The %Tolerance of 76.99% is unacceptable and the

measurement system should not be used until it has been
improved.
Poor Reproducibility indicates that the technicians are not

consistent with each other in their measurement of screen
thickness.
Mary was the least Repeatable in measuring the screen

thickness. The Range Chart indicates that Mary was particularly
inconsistent in her measures for parts 2 and 8.
There appears to be a particular problem with screen numbers

4,5, 8 and 10
Once the results of the Gauge R&R Study have been interpreted action
must be taken to address any repeatability and/or reproducibility issues
identified by the study as unacceptable.
To continue to follow the continuous data Gauge R&R case study
please turn to page 61
MSA How-to Guide

Step
4
Interpreting the Results for Attribute Agreement Analysis

4a) Analysing the data in Minitab
4b) Interpretation of graphical output
4c) Interpretation of numerical output
4d) Communicating the results
We will now examine how to analyse, interpret and communicate the results
for Attribute Agreement Analysis
Tom and Alison carry out the study, and Tom notes some
interesting observations. Tom enters the data into Minitab and is
now ready to analyse the results. Open Minitab file TOMS
ATTRIBUTE DATA 1 to view the results worksheet which includes
Toms observations.
Attribute Data
4a)
Analysing
the data in
Minitab
We can now begin to analyse the data in Minitab.

For detailed guidance on how to carry out an Attribute Agreement
Analysis in Minitab go to Appendix 8
How does Minitab analyse the data?
If we have an attribute measurement system which is working perfectly,
each appraiser will get the same result (or come to the same decision) for
every part every time they assess it:
Each appraisers results will be in 100% agreement (this is

conceptually the equivalent of Repeatability).
All of the appraisers will get the same results or reach the same
decisions as each other they will be in 100% agreement with each
other (this is conceptually the equivalent of Reproducibility).
Their results will all be in 100% agreement with the expert or

standard (this is the equivalent of Bias)
Minitab analyses the results of the Attribute Agreement Analysis and tells
us what percentage of the time the appraisers agree with themselves and
also what percentage of the time they agree with each other these 2
percentages allow us to assess both the % With Appraiser Agreement
and % Between Appraiser Agreement for the measurement system.
It also tells us what percentage of the time the appraisers are in agreement
with the Standard.
Having determined the percentages described above, we then use some

guidelines to determine whether the measurement system needs to be
improved or not.
Attribute Data
Below is an extract from Toms worksheet

Column 2 contains the number which was designated to each screen.
Column 3 contains the number of the technician who inspected the
screen:
Appraiser 1 is Steven
Appraiser 2 is Jim
Appraiser 3 is Emma
Column 4 contains the decision (pass/fail) made by the technician
assessing each screen.
Column 5 contains the decision made by Mark (final test engineer).
Refer to Appendix 8 for a guide to analysing the data in Minitab

MSA How-to Guide
In row 3, screen 16 was assessed by Mark as a Fail, but Steven

classified it as a Pass.
In row 10, screen 7 was also assessed as Fail by Mark and a Pass by
Steven.
Attribute Data
Before analysing the data in Minitab, Tom and Alison have a look at
the results. They immediately notice some inconsistencies, a few of
these can be seen in the worksheet below are:
4b) As with Gauge R&R, Minitab produces both numerical and graphical
Interpreta- output (turn to Appendix 3 for a guide to how to produce these) we will
tion of look at the graphical output first.
graphical
output
The left hand graph (Within Appraisers) shows the agreement of the
appraisers as a percentage score. The dot for each appraiser shows the
percentage of the measurements where the appraiser agrees with him or
herself.
So, in the case study example above, appraiser 1 (Steven) agrees with
himself 85% of the time this means that of the 20 parts inspected, he
made the same decision every time for 17 of the parts (85% of 20 = 17).
However his decisions werent consistent for the other 3 parts (if you look
at the worksheet you will see that his decisions were inconsistent for parts
3, 4 and 16).
Refer to Appendix 8 for information on interpreting the crosses and

red lines shown on the graph above these represent confidence
intervals.
MSA How-to Guide
The right hand graph (Appraiser vs. Standard) shows how well each of the
appraisers agrees with the standard (i.e. in the case study, how often each
of the appraisers reached the same decision as Mark, who is seen as the
expert). In the example above, we can see that appraiser 2 (Jim) agreed
with Mark 65% of the time this means that of the 20 parts measured, Jim
reached the same decision as Mark on 13 occasions (65% of 20=13). If
you look at the worksheet you will see that on screens 1, 3, 4, 7, 9, 11 and
16, Jims decision disagreed with Marks on at least one of the 3
inspections.
Interpreting the Results

Ideally we would like 100% agreement in all cases (both Within
Appraisers and Appraisers vs. Standard); however, this is unlikely to be
the case, so we use the following rules of thumb:
An unacceptable Within Appraiser score means that the appraiser

is not making consistent decisions with him/herself when repeatedly
measuring some of the parts.
An unacceptable Appraiser vs. Standard scores means that the
appraisers are not making the same decision as the expert on some
of the parts.
When using the above general guidelines you must take into
account the risk associated with the decision being made by the
measurement system. For example whilst 80% agreement may be
acceptable for the assessment of expenses claims it may be
unacceptable for the pass/fail decision on a critical part
MSA How-to Guide
Attribute Data
Tom and Alan can see from these graphs that the % Within Appraiser
Agreement (equivalent of Repeatability) for appraisers 1 and 3 is 85%
and appraiser 2 is 75%, all of which are above the acceptable
guideline.
The Appraiser vs. Standard results for Steven (70%) is acceptable;
however Jim and Emmas results (both 65%) are not acceptable their
decisions are not consistent with Marks, therefore the root cause of
this needs to be investigated.
They can see some of the inconsistencies in the worksheet:
These are some of the

results obtained by Emma
(technician 3). We can see
several instances where she
consistently
reaches
a
different decision from the
standard (Mark) it looks
like she particularly has a
problem with screens 1, 3,
11 and 12.
4c)
Interpretation of
numerical
output
Minitab also produces some numerical output in the session window this
provides more detail on the results seen in the graphs.
The first part of the output we will look at is the Within Appraisers
assessment agreement.
This gives us the % of times that each appraiser agreed with themselves
consistently on their judgement about the part.
These are the percentages which we saw on the Within Appraisers graph.
Next we look at how well the appraisers agreed consistently with the
standard. This is called Each Appraiser vs. Standard
These are the percentages which we saw on the Appraiser vs. Standard
graph.
MSA How-to Guide

The final piece of output we are interested in here is the Between
Appraisers assessment agreement. This gives us the agreement
expressed as a percentage of how consistently the appraisers decisions
agreed with each other.
Again, we look at the percentage. This percentage tells us for how many
parts the appraisers agree with each others decisions. In this example,
they agree for 65% of the parts so of the 20 parts inspected, they all
consistently made the same decision on 13 of those parts. For the
remaining parts, their decisions differed in some way (more detailed
information on this can be obtained by studying the worksheet).
The rule of thumb for interpreting the percentages is as before:
An unacceptable Between Appraiser scores means that the

appraisers are not making consistent measurements with each other
when repeatedly measuring some of the parts (the equivalent of poor
reproducibility)
Attribute Data
Tom and Alan can see that the percentage agreement between
appraisers is 65%, so they conclude that the Measurement System is
Unacceptable the 3 technicians are Inconsistent with each other in
making decisions on whether to pass or fail the screens.
Fleiss Kappa Statistic

The Minitab session window also shows us some statistics known as Fleiss
Kappa statistics for each of Within Appraisers, Appraiser vs Standard
and Between Appraisers. The statistic we are interested in is Kappa, as
highlighted below. These statistics gives us an alternative measure for
assessing how well the appraisers agree with themselves, the standard and
each other. Rather than simply looking at the observed agreement, the
Kappa statistic calculates the likelihood that the agreements seen could
have occurred by chance.
The Kappa Statistic estimates the probability that agreements

occur by chance.
The Kappa value ranges from -1 to 1
The higher the Kappa value, the stronger the agreement
Kappa = 1 means perfect agreement
Kappa = 0 means the same as would be expected by chance

(50:50)
The guidelines for interpreting Fleiss Kappa Statistics are as follows:
So if we look at the Kappa values for the Within Appraisers (from the
case study), we see that appraisers 1 and 3 have Kappa values of
>0.75 which is excellent, whereas appraiser 2 (Jim) has Kappa values
of 0.641148, which is fair to good.
MSA How-to Guide

Fleiss Kappa Statistic
The Kappa values for the Between Appraisers, has Kappa values of
0.67 which again is fair to good.
The Kappa values for Each Appraiser vs. Standard, has Kappa
values as low as 0.49 for appraiser 3 which sits just within the fair to
good guidance but indicates nevertheless the opportunity for
improvement to the measurement system.
There are other variations, for example, for categorical data (see a
Black Belt)
Summary Interpretation of Minitab Output
4d)
Communicate
results
If all of the %Agreements are greater than or equal to 70%, the

measurement system is acceptable, and can be used to collect data
If any of the %Agreements are <70%, the graphs and further

numerical data can be studied to help further understand and
investigate the reasons for the variation in the measurements
If any of the %Agreements are <70%, the measurement system is

unacceptable, and should be discussed with a Black Belt
The results of the Attribute Agreement Analysis can now be communicated

as necessary.
Try and keep the communication simple include the %Agreement results,
along with an explanation that <70% is deemed unacceptable, and that the
measurement system is not fit for purpose until improvements have been
made.
Also include the graphs in the communication to help explain how the
problems were found and also to help demonstrate the improvements.
MSA How-to Guide

Summary of Minitab Output
Tom talks through the results with Alison, Alan and the 3 technicians
who were involved in the study. In his communication he thinks
carefully about how to explain the results of the analysis to team. He
stresses that the problems highlighted by the study are a reflection of
weaknesses in the systems, standards, procedures and training being
used rather than in the people themselves. He also continues to keep
the results of the study anonymous (referring to the appraisers only as
1, 2 and 3) so that the team focussed on the overall results rather than
identifying themselves within the trial.
The minimum result for Within Appraisers was 75%. This is
above the 70% required, which means that the % agreement
within appraisers is acceptable. This means that each of the
technicians is fairly consistent with themselves.
The % agreement between appraisers however is

unacceptable, as the Between Appraisers result was 65%.
This means that the technicians are inconsistent with each
other, which suggests that the method (or the equipment)
being used by the technicians to determine whether each
screen is a pass or a fail differs between technicians.
Appraiser 2 & 3s Appraiser vs. Standard scores were

unacceptable at 65%. This means that their measurements
are not consistent with Marks, which might suggest that the
criteria for determining whether each screen is a pass or a
fail is unclear.
The above inconsistencies (or variation) in the measurement

system are resulting in some scratched screens being
passed further through the process, resulting in wasted
processing costs and some good screens are being rejected
resulting in unnecessary scrap.
Attribute Data
Step
5
Taking Action if the Results are Unacceptable

5a) What to do if results of Measurement System Analysis are
unacceptable
5b) Common reasons for unacceptable Measurement systems
5c) Identifying improvement opportunities
5d) Confirming the improvement
5a) In the case of both Gauge R&R and Attribute Agreement Analysis if the
Results are results are unacceptable, as previously discussed, the first thing to do is to
unacceptable discuss the results with a local Black Belt.
We can then use the findings from the Minitab graphs to help find the root
causes of the variation in the measurement system.
If the causes arent immediately obvious, techniques such as brainstorming
and cause and effect can be used to list all of the potential causes, and
then each of these can be investigated until the root causes are found.
5b)
Common
reasons for
unacceptable
measurement
systems
Here are some common reasons for unacceptable measurement systems,

and some questions which can be asked to help find the root causes:
If there is an issue with a particular part, compare this part with the
other parts to see if this leads to any possible causes of the variation
If there is an issue with the measurement system as used by one or
more of the appraisers, observe the appraisers and ask them for
possible reasons.
Are the operational definitions clear and being followed by everyone?
Has everyone been trained how to use them?
Are the results being recorded consistently look for rounding errors
such as some people rounding up, others rounding down. Is everyone
recording to the same number of decimal places?
Is everyone using the same version of the current standard? The
same applies to checklists, drop down menus, other reference
documents.
Is one or more of the appraisers becoming tired, rushed, bored?
Do the appraisers understand why they are doing the measurement?
If not, they may not realise the importance of following the operational
definition consistently.
MSA How-to Guide
Are the appraisers comfortable when making the measurement?

Are they having to stretch or twist to read the measurements or note
the results?
Is their eyesight OK?
Are the lighting conditions adequate?
If using software, is everyone using the same version and selecting
the same programme?
Do the appraisers load the parts consistently?
Is the procedure for loading and setting up the parts very complex?
Could it be made easier e.g. by using a jig?
Are temperature and humidity levels stable?
Does the equipment need to be left to stabilise after switching on?
Could the gauge or equipment be worn? Corroded? Contaminated?
Could the method of conducting the Measurement System have had
an impact on its results for example could inspectors be looking
extra hard for defects or did they spend much more time than
normal measuring or inspecting the parts?
Use the observations noted during the test to look for any clues.
Turn to Exercise 7 in the Workbook to identify further possible

reasons for your own measurement systems
After further discussion and observation of the inspections done

both at the in-line inspection and at final test, Tom and his team feel
that the following were the root causes of the problems seen
Attribute Data
One of the inspection lamps wasnt working properly, resulting in

the light being too dim, and it was difficult to see some of the
scratches. This was causing scratches to be missed, and
screens with scratches being passed instead of scrapped.
One of the sets of tweezers which was being used to hold the
screens appeared to be scratching the screens this was
resulting in screens which passed at in-line inspection to fail at
Final test.
There seemed to be some confusion as to what a scratch
actually meant some of the technicians thought that small
scuffs were OK to pass, whereas others classed every mark on
the screen as a fail.
Note: The team realised that they should have noticed these issues
during their initial observations when preparing for the study (step 1).
This highlighted to them how important the initial observation step is.
The technicians and Alison discuss their observations from the

results and the study, and feel that the following are possible
root causes of the poor reproducibility and repeatability
The screens are quite small and difficult to hold. Each technician
had a slightly different technique for holding them.
Mary is left handed and she seems to find the vernier calliper
more difficult to use because of this.
One of the engineers took a closer look at screen numbers 4, 5,
8 and 10 and found that these screens are actually slightly
warped, therefore the thickness result obtained depends on
exactly where on the screen the measurement is taken.
Continuous Data
MSA How-to Guide

5c)
Identify
improvement
opportunities
Identifying Improvement Opportunities

Once the causes of the variation have been found, we need to identify
solutions which will remove these causes. Sometimes the solutions will be
obvious, but sometimes solutions may need to be more creative. The ideal
solution is one which makes the variation impossible for example, the use
of a jig to prevent parts from moving, or the use of software to check for
errors.
Julie and the team use brainstorming to identify some possible
solutions to the problems identified.
Solution: They decide to manufacture a jig in which to place the
screens. The jig ensures that the screens are securely held, and
also has markings to ensure that the thickness is measured in
the same place on every screen (even if the screen is slightly
warped). The jig is easy to use for both left and right-handed
people.
They also pass the warped screens to the Screen Manufacturing

engineer for further investigation into the reasons for warping.
Continuous Data

Tom and his team use brainstorming to identify some possible
solutions to the problems identified.
Attribute Data
Solutions:
New bright lights are purchased for the in-line inspection area,
and a weekly check introduced to ensure that they are
functioning correctly.
The damaged tweezers are replaced as a short term solution

however a team is put together to look at developing a more
suitable way of holding the screens without scratching them.
Laminated photographs showing examples of passing and failing

screens are placed next to the in-line inspection station, and the
operational definition updated to reference these.
5d)
Confirm
the
improvement
Confirming the Improvement

Once the improvements to the measurement system have been
implemented, the MSA must be repeated (following the same procedure as
before) to ensure that there is in fact an improvement to the results, and
also to look for further improvement opportunities.
In order to check whether the measurement system had in fact

improved, they repeated the Gauge R&R study, using the same
procedure as before.
Open the file Julies GAUGE R&R DATA 2. This file contains the
data collected from Julies new measurement study.
Continuous Data

Once the jig was manufactured, the team wrote a new operational
definition for the measurement system and retrained all the
technicians in its use.
Re-run the analysis in Minitab and turn to Exercise 8 in the Workbook

to check if you have interpreted the Minitab output correctly
They then repeat the attribute agreement analysis, using the same
sample of parts (where practical) and method as before.
Open the file TOMS ATTRIBUTE DATA 2. This file contains the data
collected from Toms new measurement study.
Attribute Data

Tom and his team train everyone involved in the use of the new
operational definition, and ensure that the maintenance team have
been trained in doing the weekly checks on the bright lights.
Re-run the analysis in Minitab and turn to Exercise 9 in the Workbook

to check if you have interpreted the Minitab output correctly
MSA How-to Guide

Step
6
Maintaining the Improvement

6a) Ensuring the gains are sustained
It is also important to ensure that the improvements are sustained, and that the
performance of the measurement system is checked regularly.
6a)
Ensuring
the gains
are
sustained
Checking the Stability of the Measurement System

To monitor the ongoing stability of the measurement system, a regular (for
example daily or weekly) measurement can be made of the same part and the
results plotted on an SPC chart. This will detect any changes in the
performance of the measurement system over time.
For more information on SPC Charts see the How To Guide for
SPC Charts.
The part which will be measured regularly is often referred to as the
Reference Part, and should be stored carefully between measurements. The
best reference part for this purpose is one that is borderline as this will most
effectively test the ability over time of the Measurement System to be able to
discriminate between acceptable and unacceptable parts. It is recommended
that a part is selected which falls near the middle of the tolerance limits, or it
may be desirable to measure several parts which have low, high and midtolerance values. Alternatively, some measurement equipment is supplied with
standard parts, which can also be used for this purpose.
Note: Measuring the same part on a daily basis may not be possible, or may
not make sense, especially where attribute data is being collected. Asking
someone to inspect the same item every day and determine whether it is a
Pass or a Fail probably wont give meaningful results, as the appraisers will be
biased by their previous results. Instead, it may be more useful to regularly
choose a part at random from the process, ask both an appraiser and an
expert to assess it, and then compare their results, If their results are found to
be different on any occasion, this indicates a problem with the measurement
system, which should be investigated.
In either of the above cases, if a problem is found with the measurement
system, it is recommended that the appropriate type of MSA is carried out as
soon as possible, to further understand where the problem lies.
It is recommended that you discuss the most suitable frequency for
ongoing stability checks with a local Black Belt

Having confirmed the improvements, Julie and her team now need to
think about how they will ensure that the improvements are sustained.
They select a screen which has a thickness very close to the target of
1.75mm, and decide that they will measure this screen on a weekly
basis, to monitor the stability of the improved measurement system.
They update the SOP for the process to state that this screen (which
is referred to as the Reference Screen) will be measured every
Monday morning, by one of the dayshift technicians, and the results
plotted on an SPC Chart.
The data which they collect for the first 10 days is shown below, along
with an SPC chart showing the results for the first month.
Continuous Data
As you can see from the table, the measured thickness of the Golden
Screen for the first 10 days ranges from 1.74958mm to 1.75098mm,
so all of the results are very close to the expected value of 1.75mm,
which indicates that the measurement system is stable. This is also
confirmed by the SPC chart, as all of the points are between the
control limits (red lines).
MSA How-to Guide

Tom and his team feel that inspecting a Reference Screen for
scratches on a daily basis probably wont give meaningful results, as it
will be very difficult to inspect the same screen repeatedly without
being influenced by previous results. They did consider randomly
introducing the Reference Part into the inspection queue without the
technicians being aware (as they did for the Gauge R&R); however,
they are concerned that this will be difficult and time-consuming to
implement and monitor. They are also concerned that it will be difficult
to store the screen for a long period of time without it being scratched
further.
Here are the results for the first

11 weeks. As you can see, on the
6th week, Mark and the technician
dont agree in their result. On
investigation, it is quickly
discovered that the technician
who had done this inspection is
a new employee, and they realise
that, although they trained all of
the existing quality technicians in
the use of the new operational
definition, they omitted to train one of the trainers, who trains new
employees, This is quickly rectified they re-train both the new
employee and the trainer - and the training documentation is updated
to include the new operational definition.
They then run an Attribute Agreement Analysis study, which confirms
that the performance of the measurement system is acceptable, and
continue with their weekly checks.
Attribute Data
Instead, they agree with Mark that, every Wednesday, he will take a
screen at random from the production line, and assess whether he
thinks it is a pass or a fail. He will then enter this result into a
database, which will automatically compare his result to that obtained
at the in-line inspection. The IT department are able to set up the
database so that if the 2 results differ, a warning message will appear
as soon as Mark enters his result, which will enable him to put the
screen aside for further investigation. Tom will also automatically
receive an email alerting him to the result.
Repeating the MSA

For both continuous and attribute measurement systems, the MSA should
be repeated at regular intervals (such as every 6 months or annually), to
ensure that any changes in the repeatability or reproducibility of the
measurement system are detected. The MSA should also be repeated in
any of the following circumstances (where relevant):
Before implementing SPC for the first time in a process SPC charts
will only be useful if the data being used to construct them is good
quality.
If a large process variation is seen and it is suspected to be caused by

the measurement system
After any significant changes or improvements to the process if the

performance of the process changes or improves significantly (for
example if the tolerance limits change or the acceptance criteria
change) a measurement system which was previously acceptable,
may no longer be good enough therefore the MSA must be
repeated to check whether or not this is the case.
After any significant changes in the people making the measurement

for example the involvement of new or less experienced people
After a significant change in the measurement location or conditions

factors such as background noise (e.g. in an office), distractions,
volume of work can all impact the repeatability and reproducibility of a
measurement system.
After any major maintenance of measurement equipment - especially

if components have been adjusted or replaced.
After measurement equipment has been moved factors like

temperature, humidity, excess vibration can all affect the performance
of measurement equipment.
As part of the acceptance for a new machine dont assume that a

new machine will have good repeatability and reproducibility
As part of the training for new people to qualify that their inspection
ability is comparable with the standard and with other members of the
team
If you are unsure when or how often to repeat the MSA, discuss with
a local Black Belt.
MSA How-to Guide

Repeating the MSA
After discussions with Alison, Julie and her team decide they will
repeat their Gauge R&R every 3 months, until they gain confidence
that the measurement system is performing consistently well. They
update the relevant SOPs to state that the Gauge R&R must be
repeated during the first week of every quarter.
Continuous Data
After the first year, they find that the measurement system is
performing consistently well, and after further discussions with Alison,
they decide to reduce the frequency of the stability check to weekly,
and that the Gauge R&R only needs to be done once per year, unless
they have reason to suspect that there might be an issue with the
measurement system.
Since the vernier calliper is calibrated every March, they decide that
they will perform the Gauge R&R immediately after each annual
calibration. This will enable them to check that the calibration hasnt
affected the repeatability or reproducibility.
Although they find that the results of the Attribute Agreement Analysis
are usually acceptable, the study occasionally highlights problems
which they would otherwise have been unaware of, so they feel happy
that they have chosen the correct frequency for the study.
Attribute Data
Repeating the MSA

Since there are often changes in the people doing the in-line
inspection (due to temporary staff being used when there are peaks in
production volumes), Tom and his team decide that they will repeat
their Attribute Agreement Analysis once per month. They are happy to
do this, as it doesnt interfere significantly with the production process,
and the analysis isnt too time consuming as the data can be imported
from the production database into Minitab.
Summary
A Measurement System is the combination of devices, hardware,

software, procedures, standards, people and training involved in
obtaining measurements
If left unchecked there is a real risk that the measurement systems we

use could be delivering unreliable data without us being aware of the
existence of a problem. Therefore it is important to analyse how
effective our critical measurement systems are.
Measurement System Analysis is a structured procedure which we

use to assess the ability of a measurement system to provide good
quality data, before using it to collect data on which decisions are
based.
Measurement System Analysis provides a structured approach for

teams to assess whether the measurement systems they use are
repeatable and reproducible
Gauge R&R, which is used when the measurement system is being

used to collect Continuous data
Attribute Agreement Analysis, which is used when the measurement

system is being used to collect Attribute data.
MSA How-to Guide
Measurement System Analysis

How-to Guide
Change History
Revision
Date
Description of Change
Author
Owner
Approval
V6.1
20/08/2013
Guide reformatted for

SABRe
D Prodger
D Prodger
D Prodger
Document update policy

This document may be updated periodically. Major amendments will be shown as an update from one
revision number to a higher revision number (e.g. revision 1 to revision 2) and therefore the content of
the higher revision will be regarded as the latest requirements. A minor amendment will be shown as a
number change after a decimal point (e.g. revision 1.1 to revision 1.2) and therefore any of these
revisions may be regarded as the latest requirements until a major amendment is introduced
Rolls-Royce plc 2013

The information in this document is the property of Rolls-Royce
plc and may not be copied, communicated to a third party or
used for any purpose other than that for which it is supplied,
without the express written consent of Rolls-Royce plc.
While the information is given in good faith based upon the
latest information available to Rolls-Royce plc, no warranty or
representation is given concerning such information, which must
not be taken as establishing any contractual or other
commitment binding upon Rolls-Royce plc or any of its
subsidiary or associated companies.

Measurement Systems Analysis - How To

Uploaded by

Measurement Systems Analysis - How To

Uploaded by

Measurement System Analysis

MSA How-to Guide

2013 Rolls-Royce plc

2 | 2013 Rolls-Royce plc

Introduction to Measurement System Analysis

What is Measurement System Analysis?

Different Types of Measurement System Analysis

How does Measurement System Analysis work?

Choosing the appropriate type of study: Continuous or Attribute

Ensuring the prerequisites are in place

Selecting appropriate parts for the study

Identifying appropriate people to involve in the study

Deciding when to run the study

Planning the data collection

Collecting the data

Role of the observer

Gauge R&R for Continuous Data

Analysing the data in Minitab

Interpretation of graphical output

Interpretation of numerical output

Communicating the results

Attribute Agreement Analysis for Attribute Data

Analysing the data in Minitab

Interpretation of numerical output

Interpretation of graphical output

Communicating the results

What to do if results of MSA are unacceptable

Common reasons for unacceptable measurement systems

Identifying improvement opportunities

Confirming the improvement

Ensuring the gains are sustained

MSA How-to Guide

Indicating where you

4 | 2013 Rolls-Royce plc

Tips on the commonly

2013 Rolls-Royce plc | 5

MSA How-to Guide

6 | 2013 Rolls-Royce plc

1.2: What is Measurement System Analysis?

MSA How-to Guide

Turn to Exercise 2 in the Workbook to check your understanding of

8 | 2013 Rolls-Royce plc

1.4: How does Measurement System Analysis work?

2013 Rolls-Royce plc | 9

MSA How-to Guide

Turn to Exercise 3 in the Workbook to check your understanding of

10 | 2013 Rolls-Royce plc

1.4: How does Measurement System Analysis work? (cont.)

2013 Rolls-Royce plc | 11

MSA How-to Guide

Use Gauge R&R

A Measurement System is the combination of people, equipment,

12 | 2013 Rolls-Royce plc

Planning the Study

Conducting the Study

Interpreting the Results for Continuous / Attribute Data

Taking Action if the Results are Unacceptable

Maintaining the Improvement

2013 Rolls-Royce plc | 13

MSA How-to Guide

14 | 2013 Rolls-Royce plc

Case Study Scenario 2

MSA How-to Guide

16 | 2013 Rolls-Royce plc

1b) Resolution is used to assess the increments of measure that a particular

2013 Rolls-Royce plc | 17

MSA How-to Guide

Checking the equipment is calibrated, has suitable resolution

The vernier calliper used to measure the screen thickness has

The specification limits for screen thickness are 1.5 2mm, so

The team are satisfied that the measurement system is suitable

Checking the equipment is calibrated, has suitable resolution

Each appraiser has current eye examinations to the company

Reference standard screens are available, represent the full

The reference standards are in good condition

18 | 2013 Rolls-Royce plc