A Better

Way
Use the root cause analysis
by Matthew Barsalou

helix to choose the right tools

and guide investigations
ROOT CAUSE ANALYSIS (RCA) is frequently performed
in industry to determine the root cause of an occurrence or failure.
In 50 Words An occurrence can be anything from an out
Or Less of specification measurement on a computer nu-
• There are many ap-
proaches to root cause
analysis, but most em- meric control engine lathe that produces parts
phasize the implementa-
tion of corrective actions to a component or part that fails to function
and provide little infor-
mation on how to identify when installed in a customer’s system.
the root cause.
• The plan-do-check-act
cycle can be used to
An RCA can be as simple as an automo-
identify the cause of
failures either as a stand-
tive mechanic troubleshooting a car’s engine or
alone method or as part
of another process. as complicated as a multifunctional team that
meets twice a week over several months.
ROOT CAUSE ANALYSIS

November 2016 • QP 39
 Multiple ways method requires additional tools to identify a root
There are several approaches to RCA. For example, cause.2 One advantage of PDCA is the flexibility it of-
author Duke Okes listed four existing models for RCA fers. It can be used at a high-level to guide the entire
and proposed a fifth.1 The four existing models are: problem-solving process and at an operative level to
The ISO 9001 corrective action process: This provide structure to the actual failure investigation.
consists of corrective and preventive actions, and PDCA alone, however, can’t identify the root cause of
these actions require the identification of the causes of a problem.
nonconformities and the implementation of corrective The eight disciplines (8D) process: This ap-
actions. ISO 9001:2015 requires an RCA, but it does not proach to RCA requires eight steps:
explain how to identify the root cause. 1. Using a team.
Six Sigma’s design, measure, analyze, improve 2. Describing the problem.
and control (DMAIC) method: Here, the analyze 3. Implementing a containment action.
phase is an appropriate time to use RCA tools. The 4. Identifying the root cause.
proper tools to use should be based on what is being 5. Verifying corrective actions.
analyzed, and a large selection of tools is needed. 6. Implementing the corrective actions.
Six Sigma is an appropriate method for an orga- 7. Preventing an occurrence.
nization trying to identify opportunities for massive 8. Congratulating the team after finishing 8D activi-
savings or addressing a problem related to variability, ties.3
but it is not the appropriate method to use when inves- The 8D process is an effective method for managing
tigating defective parts on a production machine at 2 a root cause investigation, but the actual analysis of
a.m. In such situations, an on-the-spot solution may be the failure still requires the use of other tools.
needed to quickly eliminate the problem and get the Massaki Imai proposed PDCA as a problem-solving
machine running again. cycle. Imai’s plan phase of PDCA consists of defining
The plan-do-check-act (PDCA) cycle: This the problem, analyzing the problem, identifying the
root cause and planning countermeasures. The do
phase is where countermeasures are implemented,
Root cause analysis helix / FIGURE 1
and the implemented countermeasures are confirmed
Do Act
during the check phase. The final phase consists of
Evaluate Verify root cause. standardizing the solution.4
hypothesis Plan and implement
empirically. improvement actions. Although Imai’s approaches to RCA can be a use-
ful process, like many other approaches it does not go
into detail regarding how to identify the root cause.
ed

The Red-X method: Dorian Shainin’s problem-

Implement
cept

improvements solving method packages a powerful set of tools for

ction

Induc

is ac

identifying and controlling the source of variation.

Dedu

The tools include component search, muti-vari charts,

tion

thes

paired comparisons, variables search, full factorials, B

Hypo

vs. C, scatter plots and statistical process control using

pre-control charts.5
This approach may be effective when dealing with
Plan Check Repeat
Collect and explore Interpret the results process if variation in an assembly, but it is not ideally suited for
data to form a of the evaluation. necessary identifying the failure of one component or investigat-
tentative hypothesis. Repeat the process
by forming a new ing a single returned part that failed as a result of an
hypothesis to evaluate
or confirm a root cause error during assembly.
and begin planning
corrective actions.
New approach needed
Hypothesis Most approaches to RCA detail how to implement cor-
rejected rective actions, and this is a necessary step. Without

40 QP • www.qualityprogress.com
 ROOT CAUSE ANALYSIS

the successful implementation of corrective actions, the (EDA) can be used to generate the first tentative hypoth-
problem could persist. A more concrete approach to esis, which can later be refined based on empirical data.6
identifying the actual root cause, however, is needed. A quality engineer may graphically display failure
An RCA should be viewed as a subcomponent of data, for example, and realize one machine seems to
methods such as Six Sigma, an 8D or PDCA. It also may be producing more defects than other comparable ma-
be a standalone concept. For example, a quality techni- chines. This can be evaluated empirically by performing
cian would not start a Six Sigma project when called to a statistical hypothesis test to determine whether there
the production floor by a machine operator
whose digital micrometer stopped func-
tioning. He or she may solve the problem Used together, the scientific
within minutes and move to the next crisis.
The use of PDCA may be appropriate in method, quality tools and
such a situation.
The quality technician, for example, may statistics can paint a picture
form the tentative hypothesis, “The battery
is dead,” then test this by replacing the of the actual situation under
battery, check the results by determining
whether the digital micrometer has begun investigation.
functioning and act by taking actions to
ensure the problem will not happen again.
There are many tools available for RCA, and it would is a difference in the means of parts produced by the
be impractical to declare any one tool the best for a spe- various machines. The quality tools can be supported by
cific phase of an investigation. A specific quality tool may using EDA to graphically depict data and form tentative
serve one purpose at the start of an investigation and be hypotheses and statistical methods for analyzing data.
useful for something else as the investigation progresses. Used together, the scientific method, quality tools and
A cause and effect diagram, for example, may be statistics can paint a picture of the actual situation under
used at the start of an issue to identify potential causes investigation and lead a root cause investigator to the
to investigate. Another cause and effect diagram may be true root cause of an event.
needed after the problem has been localized. The person George E.P. Box, Stuart Hunter and William G. Hunter
performing an RCA must be ready and able to quickly described the iterative inductive-deductive process that
switch tools. moves from concepts and ideas through deduction to ac-
The classic seven quality tools are known by many tual observation and through induction back to concepts
names and are important as problem-solving tools in the and ideas.7
field of quality. These tools are useful for the collection Using such an approach, a hypothesis is formed and
and analysis of data and should be used to support the evaluated empirically through observation or testing,
scientific method. such as by performing an experiment. The empirical re-
There are many tools of varying degrees of usefulness sults are used to form a new hypothesis that is evalu-
in addition to the classic seven quality tools. Additional ated empirically. Each cycle of hypothesis forming and
quality tools such as five whys, is/is not analysis and observation should move the investigator closer to the
matrix diagrams may be necessary during an RCA. The root cause.
use of these tools alone may be insufficient to identify The scientific method—together with the iterative
the root cause of a problem, so the tools should be sup- inductive-deductive process—can be combined with
ported by a method. PDCA and supported by the use of quality tools when
The scientific method is a powerful method for per- performing an RCA. Quick iterations of PDCA provide
forming an analysis and also should be applied during a structured yet flexible procedural model for RCA. A
RCA by forming a hypothesis, designing an experiment PDCA cycle could involve a long and detailed experi-
to evaluate the hypothesis and performing the experi- ment or multiple cycles per minute when a root cause
ment. Use of John Tukey’s exploratory data analysis investigator is analyzing a failed component.

November 2016 • QP 41
 ROOT CAUSE ANALYSIS

A technician troubleshooting a machine that will An RCA may be performed for a wide variety of
not start, for example, may go through multiple PDCA reasons. An example could be a quality engineer inves-
cycles when checking the on switch, the power supply tigating the return of a failed component or a quality
and the position of the emergency shutoff switch. An manager trying to understand why a process is not per-
engineer performing an RCA may look at a part and forming at the desired level.
form a hypothesis that is immediately rejected and re- Many approaches to RCA mention PDCA, but they
placed by a new one. The engineer may cycle though fail to connect it to the scientific method which, ac-
multiple hypotheses in a short period of time. cording to Box, should be used for discovery in qual-
ity.8 The scientific method is fundamental to industry,
The RCA helix but it’s often overlooked, which simply changes a dial9
On the other hand, an experiment investigating a hy- instead of performing careful observation and con-
potheses pertaining to a part’s fatigue failure may re- trolled experiments.
quire days before the first endurance test can be com- Approaches to RCA often neglect to use of the sci-
pleted. entific method and are, therefore, incomplete. The sci-
This RCA approach is depicted in Figure 1 (p. 40) entific method—together with the classic seven quality
in the form of a helix. During the plan phase, data are tools and other tools and methods—form a complete
collected and a tentative hypothesis is formed. This is quality toolbox.
evaluated empirically through testing. The evaluation’s While a carpenter may find a hammer to be the op-
results are interpreted during the check phase. timal tool for driving a nail, it is not the proper tool for
Ideally, the cycle is repeated if the results fail to sanding a piece of wood. Like a carpenter, somebody
support the hypothesis and new information is gained performing an RCA must be able to quickly switch to
though the observation. This can be used as a basis for the appropriate tool. QP
a new tentative hypothesis. The results are verified and
corrective actions are planned if the hypothesis is ac- REFERENCES
1. Duke Okes, Root Cause Analysis: The Core of Problem Solving and Correc-
cepted. Verification is important because an accepted tive Action, ASQ Quality Press, 2009.
2. Sylvia Bushell, “Implementing Plan, Do, Check, and Act,” Journal for Quality
hypothesis is not absolute proof the root cause has and Participation, 1992, Vol. 15, No. 5.
been correctly identified. A hypothesis can be disprov- 3. Abdul M. Chaudhry, “To Be a Problem Solver, Be a Classicist,” Quality
Progress, Vol. 32, No. 6, 1999, pp. 47-51.
en if new evidence surfaces. 4. Imai Masaaki, Kaizen: The Key to Japan’s Competitive Success, McGraw Hill,
1986.
An effective approach to RCA should have several 5. Keiki R. Bhote, World Class Quality: Using Design of Experiments to Make It
concepts present. These concepts are the inclusion of Happen, AMACOM, 1991.
6. Matthew Barsalou, “Mix It Up,” Quality Progress, 2014, Vol. 47, No. 5, p. 64.
empirical methods, practicality and a wide selection 7. George E.P. Box, Stuart Hunter and William G. Hunter, Statistics for Experi-
of tools. Although qualitative methods such as brain- menters: An Introduction to Design, Data Analysis and Model Building,
second edition, John Wiley & Sons, 2005.
storming may useful during an RCA, an approach to 8. George E.P. Box, “Statistic for Discovery,” Center for Quality and Productivity
Improvement Report Number 179, March 2000.
RCA must include empirical methods. A problem can’t 9. Lynne B. Hare, “Chapter One,” Quality Progress, 2002, Vol. 35, No. 8, pp. 77-
be identified if the person performing the investigation 79.

fails to actually look at the failed part being investi-

gated.
The approach also must be practical. It can’t be so
theoretical that it’s not usable by a person seeking the
root cause of an event. The RCA helix provides an em-
MATTHEW BARSALOU is a statistical problem resolu-
pirical, easy to use and versatile approach to RCA. tion Master Black Belt (MBB) at BorgWarner Turbo
Systems Engineering GmbH in Kirchheimbolanden,
Germany. He has a master’s degree in business ad-
Be ready to switch ministration and engineering from Wilhem Büchner
Hoschschule in Darmstadt, Germany, and a master’s
An RCA is the search for the underlying cause of a degree in international business and global political
problem. The problem could be the occurrence of an economy from Fort Hays State University in Hays,
KS. An ASQ senior member, Barsalou is an ASQ-certified quality technician
unwanted event, such as a dent in an appliance’s hous- and engineer, Six Sigma Black Belt, and manager of quality/organizational
excellence. He is a Smarter Solutions-certified lean Six Sigma MBB, a
ing, or the lack of an expected occurrence, such as an
technical reviewer for QP, editor of the Statistics Division’s Statistics Digest
appliance that doesn’t start after being turned on. and the ASQ country counselor for Germany.

42 QP • www.qualityprogress.com