INDUSTRIAL SCRAP REDUCTION USING SHAININ TECHNIQUE
Vinay K S*, Praveena Gowda **, H. Ramakrishna***
*P.G.Student of industrial engineering and management, DSCE, Bangalore. Email: vinaysgv2008@gmail.com
** Professor of industrial engineering and management, DSCE, Bangalore.
Email:praveena_gowda@rediffmail.com
***Professor and Head of industrial engineering and management, DSCE, Bangalore.
Email:hrkrishna1963@gmail.com
Abstract— An alternative to the Classical and
Taguchi experimental design is the lesser known
but much simpler Shainin DOE approach
developed and perfected by Dorian Shainin
(Bothe and Bothe, 2000), consultant and advisor
to over 750 companies in America and Europe.
Shainin’s philosophy has been, “Don’t let the
engineers do the guessing; let the parts do the
talking.” Shainin methods refer to a collection of
principles, which make up the framework of a
continually evolving approach to quality. After
the classical design of experiments (DOE) and
Taguchi DOE, the third approach is Shainin
DOE, which is a collection of simple, but
powerful techniques invented or perfected by
Dorian Shainin of the United States.
Index Terms—Shainin Techniques Taguchi
DOE, Multivari analysis, B vs. C, Component
Search, Paired Comparison
I. INTRODUCTION
T he goal of this paper is to provide
overview of Shainin concepts and how it is
used for reducing the Industrial scraps.
Bhote and Bhote (2000) and Bhote (1991, 1988)
give the most complete (although not
comprehensive) treatment of SS. We agree with
reviewers (Nelson 1991, Moore 1993, Hockman
1994 and Ziegel 2001) that these books make many
unsubstantiated, exaggerated claims. What is
worse, we believe that these books are a disservice
to SS, since the hyperbole hides many of the
genuinely useful ideas. A less technical and less
controversial reference that includes many case
studies is Traver (1995).
Overviews of SS have been published in
conference proceedings; see Shainin (1992, 1992b,
1993, 1993b, 1995), Shainin and Shainin (1990)
and Shainin et al. (1997). Other review articles
include Logothetis (1990) and De Mast et al.
(2000). Does, Roes and Trip (1999) cover many of
the specific tools associated with the Shainin
System but not the overall strategy. Ledolter and
Swersey (1997a, 1997b) review two widely
heralded SS tools, pre-control and variables search.
There may be new developments not yet in the
public domain. Steiner and MacKay (2005) build
on what we think are the best elements of SS.
In assessing the Shainin System, it is important to
differentiate between the overall approach that we
think is strong, and the specific analysis methods
that are sometimes weak.
Shainin DOE basically works at eliminating
suspected process variables (Xs) mostly by using
seven different tools:
 Multi-Vari Analysis
 Component Search
 Paired Comparison
 Component Search
 Full Factorials
 B vs. C (Better vs. Current) Analysis
 Root Cause Measurements
II. PHASES OF SHAININ TECHNIQUE
2.1. Multi-Vari analysis
Multi-Vari analysis is used to reduce a large
number of suspected sources of variation to a
smaller family of variables containing the dominant
source of variation. A multi-Vari analysis is based
on the assumption that any process variation can
occur as a result of three factors, viz.
 Positional or Part-to-Part Variation
 Temporal or Time-to-Time Variation
 Cyclical or Stream-to-Stream Variation
Each type of variation is individually measured
using a run of approximately 3 to 5 units produced
consecutively at any given time. After a time lapse,
another run of 3 to 5 units are produced. This
process is repeated until 80 per cent of the out-of-
control variation in the process is captured. A plot
of these results indicates which one out of the three
variations is maximum. Simple tools like the
average and the range are then used to identify
which type of variation out of these three is the
highest.
 Table 2.2: Month wise component rejections
Month Rejection Qty % Rejection
Sep “10 4536 10.272
Oct “10 5656 12.89
Nov “10 5209 11.796
Dec “10 8913 20.184
Jan “11 8981 20.338
Feb “11 7237 16.389
Mar “11 3626 8.211

Table 2.3:Processes where the problem is generated

Process where the problem No of
Fig 2.1: Pareto Chart for Scrap in no‟s vs. Defects Remarks
is generated Machines
By analysing the scrap details from the pareto chart A1 cell & A5
we can further identify the particular root cause Process A 2
cell
from the Suspected source of variation [SSV].
A3 cell & A5
Table 2.1: 5W2H Analysis by Suspected source of Process B 2
cell
Variation
IS Packaging
Process C 2
cell
What ? (part No. Name .actual defect 92425568-015,
as against the spec) Product X
2.3. Measure and Analyse
Measuring and analysing the process can be
Who ?( Internal / External customers ) Internal customer
analysed by determining Best of Best / Worst of
Worst components. By this an activity can be
Where ? (on the geographical location. eliminated and simplified to get a best solution.
@ On Machine 1
Where on the product ( support defect We can do this process by taking one good
concentration chart ) component and one scrap component which we
obtained from cell where the problem is generated
When ? ( from when the defects Start of SOP And severity ranks are assigned to these
noticed. When are all repeating ) components. Where 1 indicates good, 10 indicates
Bad component.
Product X to be scrapped
Why ? ( is he complaining )
& no rework is possible
Table 2.4: Component Search
Through visual inspection Good Bad
How ? ( How detected ? detection
@ Dial indicator & Rota Process
Mechanism ) (BOB) (WOW)
tube

Initial value 1 10

2.2. Component Search First disassembly &

2 9
The first three methods are generally used to reduce reassembly
the SSVs to a smaller number. Once this is
Second disassembly &
achieved and the number of SSVs has been reduced 1 9
to a range of 5 to 15, the Component search
reassembly
technique can be used to further zero-in on one or
two Component that could possibly be the
dominant sources of variation. It is very useful to 2.4 Paired Comparison
separate the important Component from the Paired comparison is similar to component search
unimportant ones, control the former, and open up except that it is used when components cannot be
tolerances on the latter. reassembled or disassembled. This technique uses a
This technique can be used to identify factors of
small sample consisting of good parts and bad parts
variation related to the process parameters and is
especially useful in case the sources of variation are to further narrow down the range of potential
related to input material dimensions which are sources of variation. The good and bad parts are
subject to change during processing. selected on the basis of a suitable parameter related
to the problem.
Selecting the good parts implies selecting the Best technique can be used to further zero-in on one or
of Best (BOB) parts and selecting the bad parts two Component that could possibly be the
implies selecting the Worst of Worst (WOW) parts dominant sources of variation. It is very useful to
separate the important Component from the
with respect to the desired response. Its roots lie in
unimportant ones, control the former, and open up
the assumption that the values of the two groups tolerances on the latter.
must be substantially different and hence This technique can be used to identify factors of
identifiable. Usually 8 good and 8 bad parts are variation related to the process parameters and is
selected. Once the sample is selected, each SSV is especially useful in case the sources of variation are
measured on the 16 parts and the results are related to input material dimensions which are
recorded in the form of a table. subject to change during processing.
The analysis to pinpoint a suspected source of
variation as being significant or not significant is
based on a minimum “end count”. This technique
can be used only when the suspected source of
variation is measurable on both the good and bad
parts. This technique is explained in more detail in
the project illustration.

Table 2.5: Calculation of D/d Ratio

Good (BOB) Bad (WOW)
Initial value 1 10
First disassembly & reassembly 2 9
2.6. Full Factorials
Second disassembly & reassembly 1 9
This is same as the full factorial of Classical DOE
.
except that here Shainin recommends its use once
Median 2 9
the number of SSVs has been bought down to four
Range 1 0 or less than four by using one or more of the above
. techniques. Thus, the maximum size of the full
D- Difference b/w Medians 7 factorial experiment would be 24.Also the
d – Average of ranges 0.5 objectives of this tool remain the same as those of
D/d ratio 14 the variable search technique, i.e., pinpoint the Red
X or the Pink X, separate and quantify the
If D/d ratio is more than 3 then we can conclude important and unimportant factors, and open up
components is the reason for scarp generation , if tolerances on the unimportant ones.
D/d ratio is more than 3 then we can conclude
Machine or Assembly process is the reason for Table 2.6: Pinpointing components
Spec Reject Component < 4cc/min
scarp generation.
Component no Remarks Response Y

Steps involved in paired comparison: 1 B 4

 Measure the parameters of n Good (BOB) and 2 B 4

3 B 4
n Worst (WOW)
4 B 4
 Arrange the parameters in ascending order
5 B 4
 Measure the count. 6 B 4
a) If first and last component falls in same 7 B 4
category, (i.e. B only or W only) then count=0 8 B 4
b) If first and last component are of different B W 38
category, (i.e. B at first and W at last) then take H W 38
count till where there shift from one category to G W 60

other (i.e. B>>W or W>>B) both at top and bottom C W 62

to validate the results. A W 80

D W 80
E W 80
2.5. Component search
F W 80
The first three methods are generally used to reduce
Count= 16 , since count<3
the SSVs to a smaller number. Once this is this is the reason for the problem
achieved and the number of SSVs has been reduced
to a range of 5 to 15, the Component search
III. IMPLEMENTATION From this analysis we can say that the improvement
takes at 95% Confidence Level
3.1. Better vs. Current (B vs. C) analysis
This is a final validation tool useful in validating 3.2. Identifying root causes
the root cause of dent or damage identified using
the other methods. Some applications of this
method are:

 To validate the root cause

 To validate product/process changes made
 To validate improvement actions.
C stands for a current product, B for a better or
improved product. The purpose of a B vs. C test is
to validate or verify that the improvement is
effective and permanent. Often, designers think that
they have achieved a product improvement, only to
discover a week or two later that the improvement
has vanished. They find themselves thrown back to
square one. To avoid these premature judgments, it
is necessary to go back to the old design to see if
the problem has reappeared, then return to the new
design to see if the improvement is reconfirmed.
This back-and-forth process must be repeated
twice. It is the equivalent of turning a light switch
“on” and turning it “off.”
Then the three “B” readings and the three “C”
readings must be arranged in descending rank
order. Only if the three Bs outrank the three Cs can
there be a 95 percent confidence that the B product
is significantly better than the C product. A mixed
rank order would indicate that a significant
improvement could not be validated.
(Note: The sequence of testing three Bs and three
Cs must be randomized.)
The B vs. C test is also very useful in determining a
reliability improvement very quickly. Usually, a
design change for reliability takes months of
exposure in the field to verify. No one can afford
that length of time. And so half-baked
improvements are prematurely made and only
much later does anyone discovers that the design
change was ineffective.
Table 3.1: Better vs. Current (B vs. C) analysis
Component Current
Better (Xb)
No (Xc)
1 10 1 products or processes.
2 10 2
3 9 4
4 10 3 there are some unfortunate aspects about its
5 8 2 promotion. Most notably, many of the specific
6 10 2 tools and the whole approach have not been subject
Average 9.5 2.33
Xb-Xc 7.17
∑(b) 1.21
K*∑b
95% CI 2.96
90% CI 2.61
50% CI 1.41
Fig 3.1: Identify root causes
IV. CONCLUSION
The guiding principles of the Shainin System are
powerful, and, at least in combination, unique.
They include the application of Juran‟s Pareto
principle to the contribution of the causes, the
emphasis on using observational prior to
experimental investigations, and the search for a
dominant cause using the process of elimination
and leveraging. SS deals carefully with the problem
of possible confounding of suspect causes by
conducting a small verification experiment.
We think that the principles and tools related to the
diagnostic journey are generally very strong. Those
related to the remedial journey are much weaker.
The Shainin System, as reflected by the genesis of
the methodology in manufacturing, is best suited
for medium to high volume production. Most of the
tools implicitly assume that many parts are
available for study. Even when using leverage,
where the investigations involve only a small
number of extreme parts, there must be a
substantial amount of measurement to find the
extremes. Like many other systems with strong
statistical components, SS does not handle well
situations where there are few parts to “talk to”
such as in the design and development of new
Although our assessment of SS is strongly positive,
to a peer reviewed public discussion. This may be
due to the fact that much of the specific
terminology is trademarked and is thus legally
protected. We feel this is unfortunate since it has
reduced the dissemination of what we think is an
excellent approach.
Table 4.1: Differences between Shainin tools

discussion of feasible alternatives. In our
experience, there is also a rigidity with which the
methodology is presented.
In many situations, other statistical tools, such as
regression, time series, and analysis of variance,
could be very useful, but are not employed because
they are not formally part of the SS tool bag.
REFERENCES
Greg Blue “Six Sigma for Managers” © McGraw
Hill 2002 0-07-138755
Montgomery D.C. (1996), Introduction to
Statistical Quality Control, 3rd edition, John Wiley
and Sons, New York.
Peter.S.Pandey Robert P, Neumann Roland R
Cavanagh. “The Six Sigma Way” 2nd edition
McGraw-Hill 2000 0-07-135806-4

Arnheiter, Edward D.and Maleyeff, John. “The

integration of lean management and Six
Sigma”. The TQM Magazine, Vol.17 no. 1, 2005
pp.5-18. © Emerald Group Publishing Limited
2004 0954-478X

Bhote, Keki R. (1988), World Class Quality:

Design of experiments made easier, more cost
effective than SPC, American Management
Association (AMACOM), and Saranac Lake, New
York

Bhote, Keki R. (2003), The Power of Ultimate Six

Sigma, Keki Bothe‟s Proven System for Moving
Beyond Quality Excellence to Total Business
Excellence, American Management Association
(AMACOM), and Saranac Lake, New York

Dennis K. J. Lin, „„Making full use of Taguchi‟s

orthogonal arrays” Quality and Reliability
Engineering International, Vol.10, 1994

Juran, J.M. (1988),Juran‟s Quality Control

Handbook, 4th edition, McGraw -Hill, New York.