You are on page 1of 6

STATGRAPHICS Rev.

7/24/2009

Pareto Analysis
Pareto Analysis is a statistical procedure that seeks to discover from an analysis of defect reports or customer complaints which vital few causes are responsible for most of the reported problems. The old adage states that 80% of reported problems can usually be traced to 20% of the various underlying causes. By concentrating ones efforts on rectifying the vital 20%, you can have the greatest immediate impact on product quality. STATGRAPHICS supports two basic types of Pareto analysis: weighted and unweighted. In an unweighted analysis, only the frequency of each problem is considered in determining the vital few causes. In a weighted analysis, the cost or impact of each problem is also considered.

Sample StatFolio: paretochart.sgp Sample data:


The file checksheet.sf6 contains information on n = 166 defects discovered in a manufacturing process. This data, from Montgomery (2005), is shown below: Defect Parts damaged Machining problems Supplied parts rusted Masking insufficient Misaligned weld Processing out of order Wrong part issued Unfinished fairing Adhesive failure Powdery adoline Paint out of limits Paint damaged by etching Film on parts Primer cans damaged Voids in casting Delaminated composite Incorrect dimensions Improper test procedure Salt-spray failure Frequency 34 29 13 17 2 4 3 3 6 1 2 1 5 1 2 2 36 1 4

Notice that the data above has already been tabulated, i.e., the number of defects attributable to each type of problem has been identified. Alternatively, the file could have been structured with 166 rows, each identifying the cause of a single problem, and STATGRAPHICS could have been asked to count the occurrence of each unique cause.

2009 by StatPoint Technologies, Inc.

Pareto Analysis - 1

STATGRAPHICS Rev. 7/24/2009

Data Input
The data input dialog box allows the user to specify how the data is structured.

Data: If the data is Untabulated, with each row corresponding to a single defect, then the name of the column identifying the type of defect should be entered. If the frequency of each type of defect has already been calculated, then a column containing the count for each type should be entered. Labels: a column with labels, one for each unique type of defect. This field is primarily used for tabulated data, in which case the order of the labels should correspond to the frequencies in the data field. For untabulated data, labels are usually automatically generated from the Data entries, so that this field may be left blank. However, if labels are entered here for untabulated data, they will be assigned to the types of defects in the Data column in numeric or alphabetic order, depending upon the type of the Data column. Weights: when creating a weighted or exposure Pareto chart, the weight to be assigned to each type of defect. Weights should be entered in the same order as described for Labels. Select: subset selection.

2009 by StatPoint Technologies, Inc.

Pareto Analysis - 2

STATGRAPHICS Rev. 7/24/2009

Analysis Summary
The Analysis Summary shows the total number of defects and the number of different types.
Pareto Analysis - Frequency
Data variable: Frequency Total counts: 166.0 Number of classes: 19

In the sample data, there are n = 166 defects of k = 19 different types.

Frequency Table
This table shows each of the different types of defects, ordered from most frequent to least frequent:
Pareto Chart with Cumulative Frequencies Class Label Rank Count Weight Incorrect dimensions 1 36 1 Parts damaged 2 34 1 Machining problems 3 29 1 Masking insufficient 4 17 1 Supplied parts ruste 5 13 1 Adhesive failure 6 6 1 Film on parts 7 5 1 Salt-spray failure 8 4 1 Processing out of or 9 4 1 Unfinished fairing 10 3 1 Wrong part issued 11 3 1 Delaminated composit 12 2 1 Voids in casting 13 2 1 Paint out of limits 14 2 1 Misaligned weld 15 2 1 Improper test proced 16 1 1 Primer cans damaged 17 1 1 Paint damaged by etc 18 1 1 Powdery adoline 19 1 1 Total 166 Weighted Score 36 34 29 17 13 6 5 4 4 3 3 2 2 2 2 1 1 1 1 166 Cum. Score 36 70 99 116 129 135 140 144 148 151 154 156 158 160 162 163 164 165 166 Cum. Percent 21.69 42.17 59.64 69.88 77.71 81.33 84.34 86.75 89.16 90.96 92.77 93.98 95.18 96.39 97.59 98.19 98.80 99.40 100.00

Percent 21.69 20.48 17.47 10.24 7.83 3.61 3.01 2.41 2.41 1.81 1.81 1.20 1.20 1.20 1.20 0.60 0.60 0.60 0.60

The table displays the following information for each type of class or defect: Rank - the ranking of each class j from 1 through k. Count - the frequency or number of times each defect type occurred f j . Weight - the weight w j associated with each class j. If weights were not specified on the data input dialog box, then w j 1 for all j.

Weighted score - the score associated with each class:

s j wj f j
2009 by StatPoint Technologies, Inc.

(1)
Pareto Analysis - 3

STATGRAPHICS Rev. 7/24/2009 Cumulative score - the sum of all weighted scores from class 1 through the current class.
Percent - the percent of the total weighted score represented by each class:

pj

100s j
W

(2)

where
W wi f i
i 1 k

(3)

Cumulative percent - the sum of the percentages from class 1 through the current class.

In the example, the 3 most common types of defects were Incorrect dimensions, Parts damaged, and Machining problems. Together, these 3 types of problems accounted for nearly 60% of all defects.

Pareto Chart
The Pareto chart shows the frequency of defects in graphical form:
Pareto Chart for Frequency
180 150
129 116 99 70 36 135 165 166 160 162 163 164 154 156 158 148 151 140 144

frequency

120 90 60 30

Processing out of or

0
Machining problems Film on parts Incorrect dimensions Masking insufficient Salt-spray failure Parts damaged Supplied parts ruste Adhesive failure

Paint damaged by etc

Delaminated composit

Paint out of limits

Improper test proced

Voids in casting

Primer cans damaged

Wrong part issued

Misaligned weld

Each type of defect is represented by a vertical bar. Defect types have been sorted from most frequent to least frequent. Above the bars, a line has been drawn representing the cumulative number of defects from left to right. For example, the leftmost 3 types of defects accounted for 99 of the total 166 defects observed. Note: The X-Axis tab on the Graphical Options dialog box allows you to rotate the labels for each bar and to reduce the font size if necessary. You can also use your mouse to drag the x-axis away from the bottom of the screen if the labels are cut off.

2009 by StatPoint Technologies, Inc.

Unfinished fairing

Powdery adoline

Pareto Analysis - 4

STATGRAPHICS Rev. 7/24/2009


Pane Options

Display: the value to be placed above each bar. Either the percentage of the total weighted score, the score, or no label.

Cumulative Pareto Chart


This form of the Pareto chart uses the bars to create the cumulative curve.
Pareto Chart for Frequency
180 150
129 116 99 164 165 166 158 160 162 163 151 154 156 144 148 135 140

frequency

120 90 60
36 70

30 0
Processing out of or Adhesive failure Unfinished fairing Parts damaged Primer cans damaged Supplied parts ruste Wrong part issued Salt-spray failure Voids in casting Misaligned weld Improper test proced Masking insufficient Delaminated composit Incorrect dimensions Machining problems Paint out of limits Paint damaged by etc Powdery adoline Film on parts

Analysis Options
If there are many types of defects that occur infrequently, the least frequent defects may be combined into a single Other class using Analysis Options.

Combine Classes: the method to be used to combine the least frequent classes:

2009 by StatPoint Technologies, Inc.

Pareto Analysis - 5

STATGRAPHICS Rev. 7/24/2009


None: do not combine any classes. Counts Less Than: combine all classes with scores less than that specified. Percentages Less Than: combine all classes accounting for less than the specified percentage of the total weighted score. Smallest Classes: combine the specified number of classes that have the smallest scores.

Example: Combining Classes with Less Than 2% of All Defects


Pareto Chart for Frequency
180 150
129 116 99 70 36 135 140 144 148 166

frequency

120 90 60 30
Incorrect dimensions

2009 by StatPoint Technologies, Inc.

Processing out of or

Machining problems

Masking insufficient

Film on parts

Supplied parts ruste

Salt-spray failure

Parts damaged

Adhesive failure

Other

Pareto Analysis - 6

You might also like