80/20 Pareto Analysis

It can ensure that your cause analysis and corrective
Pareto Analysis
actions are focused on the vital few:
What is it?
A type of bar chart used to distinguish the vital few areas
♦ Customer complaints -- Identify each complaint by
of importance from the trivial many. The tool is named
category (e.g., late delivery, damaged shipment,
after Vilfredo Pareto, a 19th century economist who studied
bil ing error, etc.).
the distribution of wealth in Italy and found 20% of Italians
possessed 80% of the wealth.

♦ Employee complaints -- Identify each employee
concern by category (e.g., supervisor,
After World War II, Joseph Juran taught Pareto's principle
compensation, lack of direction, training,
as the 80/20 rule to Japanese and American managers.
tools/equipment, etc.).


♦ Causes of project delays -- Identify the number
Quote: The Pareto principle is employed to identify
and type of each delay in a series of projects. This
the "vital few," whether customers, customer
wil help you address the most common causes.
needs, product features, process features, or
inputs. Identification of the vital few helps to

assure that resources and attention are
♦ Reasons for manufacturing interruptions -- Identify
concentrated where they wil do the most
anything that causes a slow down or stoppage of
good.
production in your key bottleneck or constraint

--from Juran's Quality Control Handbook, 4th
(e.g., lack of materials, equipment breakdown,
Edition, by J.M. Juran and Frank M. Gryna,
yield below optimum, rate below optimum,
©1988 by McGraw Hil , ISBN#0-07-033176-6.
production of scrap, repairs/rework, etc.)




A Pareto chart plots the number of times a particular event
or cause occurs versus the total number of instances of

that problem or category of events.

Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

The Pareto chart enables you to test for the 80/20 rule,
How do I use It?
i.e., that few causes are often responsible for a great
majority of effects. It helps you identify those vital few
Identify the data to analyze. Clarify your purpose, i.e.,
areas where your efforts wil have the greatest impact.
What do you hope to learn?



As the table indicates, Pareto's 80/20 rule has wide
applicability.
Decide on the time period for the study. You usual y
need 50 data points for a Pareto chart.
20% of…
Often account for 80% of…
Take into account seasonality, or typical patterns within
Your products
Sales volume
the time period selected, to ensure the data wil be
representative for your purpose.
Your customers
Profits

Your equipment
Breakdowns

Decide on the categories and develop clear
Your products or services
Help desk requests
definitions of each.
Your suppliers
Delays or supply problems


Remember, once col ected, data
The carpets in your house
The wear
can only be stratified into
categories if properly coded.
The clothes in your closet
What you wear

Operating system software
Computer processing time
Process steps
Problems or errors
♦ In defining categories, be sure each category is
discrete, with no overlap between them, so each
Students
Absences
event or item fal s into only one category.
Motorists
Accidents
♦ If recording times or quantities, decide how much
precision is needed, i.e., how many decimal
Diseases
Deaths
places.
Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

POINTER: See Stratification for possible categories.
You consider the fol owing types of questions in
determining what questions to ask and what data to

See Checksheet for more on data
col ect.

col ection.

♦ WHO—Cal er's Position, function;



♦ WHAT—Type of question or inquiry;

For example, imagine that you manage a benefits cal
♦ WHICH—Forms and form sub-sections in
center, where employees of your company cal to ask
question;
questions about their health or retirement benefits, etc.


♦ WHEN—Cal length, Enrol ment date,

Cal er's stage in the benefits cycle;
You are interested in reducing the amount of time your
agents spend answering cal s, so you col ect some
♦ WHERE—Location of the cal er;
data.

♦ WHY—Suspected cause of problem.



As cal s come in, your agents ask a few key questions.

An on-screen form with pre-defined response
categories might be used to record data.
You decide upon some key WHO questions and
develop possible response fields like those shown on

the next page.







Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

Position
Department
Fol owing is an example of data col ected by the
fictitious Benefits Cal Center.
♦ Executive
♦ Manufacturing

♦ Manager
♦ Engineering
Benefits Center Data by Position of the Caller
♦ Technical
♦ Marketing
Positions
Total # of # of Avg. Cal Cal s as
EEs
cal s Length % of EE
♦ Supervisor
♦ R&D
(min.s)
#s
♦ Foreman
♦ HR
Executive
315
293
22.5
93
♦ Line worker
♦ IT
Manager
1,183
472
5.2
40
♦ Clerical
♦ Finance
Technical
1,820 1,257
6.4
69
♦ Other
♦ Other
Supervisor
5,422 1,264
2.6
23


Foreman
6,923
872
3.8
13

Always test your categories before data
Line worker
36,422 4,286
2.2
18
col ection. You may find categories
overlap, or too many responses fal in
Clerical
1,267
67
19.3
12
the Other category.
Other
624
29
3.4
5


Totals
53,976 8,540
65.4
273
Record the data, i.e., the number of times each event

fal s into a particular category. You can do this by:



♦ Using existing data, or



♦ Gathering new data.
Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

Chart the data as shown in the example below.

HINT: Note that the actual count for each

category is written above the bar in the


/
previous example of a Pareto chart for the

number of cal s by position.








Add the cumulative line, from zero to the right hand

axis, to show how many bars it takes to account for

80% of the total.

# $ % ' (
(
)
*
+
, - (
&
(
.
! "
&

For each bar, the point on the cumulative line
♦ The horizontal (X) axis is used for the categories, with
represents the per cent of that bar plus al preceding
equal widths for each.
bars.


♦ Put the "Other" category last.

Interpret the chart.
♦ The left hand vertical axis (Y) should be equal to the
total number of occurrences (n).
♦ Check to see if the Pareto principle applies.
♦ The right hand vertical axis should be 120%.
♦ Do 20% of the categories account for 80% of the
♦ Graph each occurrence as a bar, with al bars
total occurrences?
touching and arranged from largest to smal est.

♦ The height of each bar equals the count for that

category.
Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

HINT: Even when the 80/20 rule does not
Use Different Categories
apply, if a few (<20%) of the categories
account for 60% or more of the
If you identified multiple stratification factors for the
occurrences, you may stil have identified
data being col ected, you can stratify it in different ways
the vital few categories having the greatest
and produce various Pareto charts.
impact on the situation or problem.



If not, you should look for other ways to
One of the stratification factors identified for the Cal
stratify the data (Step - ), until the
Center data was the "Type of Question" asked by the
80/20 principle applies and you can
cal er.
identify a vital few.



Below is data for the number of cal s by "Type of
In the previous example, the Pareto principle does not
Question".
apply. About 80% of cal s are made by three of the
eight groups:
Type of Question
Number of cal s
Enrol ment
121
♦ Line workers,
Health
1,127
Retirement
32
♦ Supervisors, and
Stock Options
94
Savings Plan
177
♦ Technical personnel.

Change of status
6,942

Disability
29

Check to make sure the "Other" category
Other
18
is not too large. If it is one of the larger
bars, then re-stratify the data with
Totals
8540
specific categories for responses

currently being captured as "Other".


The Pareto chart for this data is on the next page.

Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

' $
2
Stratify according to Impact.

/
Before taking action, it is always advisable to stratify

the data in terms of impact, e.g., time or money, versus

just the count of occurrences.




For the Benefits Center, the count is number of cal s.

One way to think of impact is in terms of the length of

those cal s, i.e., the longer the cal s, the more

3
+
$
.
.
resources required by the Center to cover phones.

+
&
("
#
,
4
5
#



The Pareto Chart for Average Cal time by Position of

the cal er is shown below.


In the example above, the Pareto principle does apply.

0 &
+
'
Over 80% of cal s are for a change of status, such as:

/


♦ Getting married,


♦ Divorced,

1

1
1

1
1
1
1
1
♦ The birth of a child, or

, - (
&
(
' (
(
*
+
)
.
# $ &


!
"


♦ A change of address, etc.


Note that two relatively smal groups of employees,

Executive and Clerical, are the largest bars for average
This might suggest you could reduce significant time
cal time. For some reason, these two groups are
spent answering this type of cal if you can develop an
making the longest cal s to the center.
automated means for employees to make simple
changes.

Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

11 Normalize the data
This Pareto chart is shown below.
Counts can have very different areas of opportunity,

, $
, (
6
i.e., the range of possible values for the count. For this


reason, it is often best to normalize the data by

converting counts to a rate or percentage.


For example…
!

Areas of Opportunity Can
…So Convert

Vary…
Counts to Rates

, - (
&
' ( %
*
+
# $ &

)
(
.
♦ Sales volume can vary
Sales $

(
!
"
based upon the number of
# Store visitors


visitors to a store.


♦ The number of complaints
# Complaints

can vary based upon the
# Customers

number of customers.
The normalized data shows that on a percentage
basis, far more Executives, Technical personnel and
♦ Payrol errors can vary
# Payrol errors
Managers cal ed the center.
based upon the number of
#checks

employees paid.
processed


Clearly the higher "ranking" employees made greater
use of the Benefits Center during the period in
In our cal center example, since there are only 315
question.
Executives and over 36,000 line workers, it is not

surprising that the count of cal s by Executives is

dramatical y smal er. Instead, you might look at the per

cent of cal ers by position:




# of cal s by position ÷ total # of employees by position

Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---

12
' $
2
" 7
, - (
&
6
7
Work on the Largest Bar

/
Once the Pareto chart has identified the vital few
#

areas, a good next step is often to take the largest bar

and do further Pareto analysis of that subset to further

localize the problem or causes.






For the Benefits Center example, the charts shown

$
.
+
3
.
previously indicate that as a group, Executives make a
("
4
&
+
5
,
#
#
disproportionately higher number, and some of the

longest cal s to the Center.

" !
#





It would make sense to do further Pareto analysis of

just the Executive data to better understand why.

POINTER: It is common to begin cause analysis on the

largest bars. See Improvement Methods in
The Pareto Chart at the top of the next column looks
Chapter 6 such as Process Mapping, Is/Is
only at Executive cal s.
Not Analysis, Cause-Effect diagram, etc. for

help with this.

It indicates that the most frequent type of cal made by

Executives is related to stock options.













Making Better Joe Kilbride, Kilbride Consulting, Inc.
Excerpt from Chapter 7: Making Data

---