ACCEPTANCE SAMPLING

IENG 6104: Quality Engineering

M.Sc. in Industrial Engineering & Management
Industrial Engineering Department
School of Mechanical Engineering & Industrial Engineering
IoT, HH Campus, Ambo University, Ambo

Dr. Er. Ajit Pal Singh

Associate Professor
0921512436 (M)
OBJECTIVES 2

Upon completion of this chapter, the reader is expected to:

 know the advantages and disadvantages of sampling; the types of sampling plans and
selection factors; criteria for formation of lots; criteria for sample selection; and decisions
concerning rejected lots;
 determine the OC curve for a single sampling plan; . determine the equations needed to graph
the OC curve for a double sampling plan;
 know the properties of OC curves;
 know the consumer–producer relationships of risk, AQL, and LQ;
 determine the AOQ curve and the AOQL for a single, sampling plan;
 design single sampling plans for stipulated producer’s risk and for stipulated consumer’s risk.
INTRODUCTION 3

In recent years, acceptance sampling has declined in importance as statistical process control
has assumed a more prominent role in the quality function.
However, acceptance sampling still has a place in the entire body of knowledge that constitutes
quality science.
In addition to statistical acceptance sampling, discussed in this chapter, there are several other
acceptance sampling practices, such as fixed percentage, occasional random checks, and
100% inspection.
FUNDAMENTAL CONCEPTS 4

Lot-by-lot acceptance sampling by attributes is the most common type of sampling. With this
type of sampling, a predetermined number of units (sample) from each lot is inspected by
attributes.
If the number of nonconforming units is less than the prescribed minimum, the lot is accepted; if
not, the lot is not accepted.
Acceptance sampling can be used either for the number of nonconforming units or for
nonconformities per unit.
To simplify the presentation in this chapter, the number of nonconforming units is used;
however, it is understood that the information is also applicable to nonconformities per unit.
Sampling plans are established by severity (critical, major, minor) or on a demerit-per-unit
basis.
FUNDAMENTAL CONCEPTS… 5

A single sampling plan is defined by the lot size, N, the sample size, n, and the acceptance
number, c. Thus, the plan: N = 9000, n = 300, c = 2 means that a lot of 9000 units has 300 units
inspected. If two or fewer nonconforming units are found in the 300-unit sample, the lot is
accepted. If three or more nonconforming units are found in the 300-unit sample, the lot is not
accepted.
Acceptance sampling can be performed in a number of different situations where there is a
consumer–producer relationship. The consumer and producer can be from two different
organizations, two plants within the same organization, or two departments within the same
organization’s facility. In any case, there is always the problem of deciding whether to accept or
not accept the product.
FUNDAMENTAL CONCEPTS… 6

Acceptance sampling is most likely to be used in one of five situations:

(1) When the test is destructive (such as a test on an electrical fuse or a tensile test), sampling is necessary;
otherwise, all of the units will be destroyed by testing.
(2) When the cost of 100% inspection is high in relation to the cost of passing a nonconforming unit.
(3) When there are many similar units to be inspected, sampling will frequently produce as good, if not better,
results than 100% inspection. This is true because, with manual inspection, fatigue and boredom cause a higher
percentage of nonconforming material to be passed than would occur on the average using a sampling plan.
(4) When information concerning producer’s quality, such as X bar and R, p or c charts, and Cpk , is not available.
(5) When automated inspection is not available.
FUNDAMENTAL CONCEPTS… 7

ADVANTAGES AND DISADVANTAGES OF SAMPLING

When sampling is compared with 100% inspection, it has the following advantages:
(1) It places responsibility for quality in the appropriate place rather than on inspection, thereby
encouraging rapid improvement.
(2) It is more economical, owing to fewer inspections (fewer inspectors) and less handling damage during
inspection.
(3) It upgrades the inspection job from monotonous pieceby-piece decisions to lot-by-lot decisions.
(4) It applies to destructive testing.
(5) Entire lots are not accepted, rather than the return of a few nonconforming units, thereby giving
stronger motivation for improvement.
FUNDAMENTAL CONCEPTS… 8

Inherent disadvantages of acceptance sampling are that:

(1) There are certain risks of not accepting conforming lots and accepting nonconforming lots.
(2) More time and effort is devoted to planning and documentation.
(3) Less information is provided about the product, although there is usually enough.
(4) There is no assurance that the entire lot conforms to specifications.
FUNDAMENTAL CONCEPTS… 9

TYPES OF SAMPLING PLANS

There are four types of sampling plans: single, double, multiple, and sequential. In the single sampling
plan, one sample is taken from the lot and a decision to accept or not accept the lot is made based on the
inspection results of that sample.
Double sampling plans are somewhat more complicated. On the initial sample, a decision, based on the
inspection results, is made whether (1) to accept the lot, (2) not to accept the lot, or (3) to take another
sample. If the quality is very good, the lot is accepted on the first sample and a second sample is not
taken; if the quality is very poor, the lot is not accepted on the first sample and a second sample is not
taken. Only when the quality level is neither very good nor very bad is a second sample taken.
FUNDAMENTAL CONCEPTS… 10

If a second sample is required, the results of that inspection and the first inspection are used to make a decision. A double
sampling plan is defined by
N = lot size
n1 = sample size of first sample
c1 = acceptance number for the first sample (sometimes the symbol Ac is used)
r1 = nonacceptance number for the first sample (sometimes the symbol Re is used)
n2 = sample size of second sample
c2 = acceptance number for both samples
r2 = nonacceptance number for both samples
If values are not given for r1 and r2, they are equal to c2 + 1.
FUNDAMENTAL CONCEPTS… 11

An illustrative example will help to clarify the double sampling plan: N = 9000, n1 = 60, c1 = 1, r1
= 5, n2 = 150, c2 = 6, and r2 = 7. An initial sample (n1) of 60 is selected from the lot (N) of 9000
and inspected.
One of the following judgments is made:
(1) If there is 1 or fewer nonconforming unit (c1), the lot is accepted.
(2) If there are 5 or more nonconforming units (r1), the lot is not accepted.
(3) If there are 2, 3, or 4 nonconforming units, no decision is made and a second sample is
taken.
FUNDAMENTAL CONCEPTS… 12

A second sample of 150 (n2) from the lot (N) is inspected, and one of the following judgments is
made:
(1) If there are 6 or fewer nonconforming units (c2) in both samples, the lot is accepted. This
number (6 or fewer) is obtained by 2 in the first sample and 4 or fewer in the second sample, by 3
in the first sample and 3 or fewer in the second sample, or by 4 in the first sample and 2 or fewer
in the second sample.
(2) If there are 7 or more nonconforming units (r2) in both samples, the lot is not accepted. This
number (7 or more) is obtained by 2 in the first sample and 5 or more in the second sample, by 3
in the first sample and 4 or more in the second sample, or by 4 in the first sample and 3 or more in
the second sample.
FUNDAMENTAL CONCEPTS… 13

A multiple sampling plan is a continuation of double sampling in that three, four, five, or as many
samples as desired can be established. Sample sizes are much smaller. The technique is the
same as that described for double sampling; therefore, a detailed description is not given.
Multiple sampling plans of ANSI/ASQ Z1.4 use seven samples.
In sequential sampling, items are sampled and inspected one after another. A cumulative record
is maintained, and a decision is made to accept or not accept the lot as soon as there is
sufficient cumulative evidence.
FUNDAMENTAL CONCEPTS… 14

All four types of sampling plans can give the same results; therefore, the chance of a lot being
accepted under a single sampling plan is the same under the appropriate double, multiple, or
sequential sampling plan.
Thus, the type of plan for a particular unit is based on factors other than effectiveness.
These factors are simplicity, administrative costs, quality information, number of units inspected,
and psychological impact.
Perhaps the most important factor is simplicity. In this regard, single sampling is the best and
sequential sampling the poorest.
TABLE 1 Random Number 15
FUNDAMENTAL CONCEPTS… 16

 Administrative costs for training, inspection, record keeping, and so on, are least for single
sampling and greatest for sequential sampling.
 Single sampling provides more information concerning the quality level in each lot than
double sampling and much more than multiple or sequential sampling.
 In general, the number of units inspected is greatest under single sampling and least under
sequential. An ASN curve, shown later in the chapter, illustrates this concept.
FUNDAMENTAL CONCEPTS… 17

A fifth factor concerns the psychological impact of the four types of sampling plans. Under
single sampling there is no second chance; however, in double sampling, if the first sample is
borderline, a second chance is possible by taking another sample.
Many producers like the second-chance psychology provided by the double sample.
In multiple and sequential sampling, there are a number of “second chances”; therefore, the
psychological impact is less than with double sampling.
Careful consideration of the five factors is necessary to select a type of sampling plan that will
be best for the particular situation.
FUNDAMENTAL CONCEPTS… 18

FORMATION OF LOTS
(A) Lots should be homogeneous, which means that all product in the lot is produced by the same machine, same
operator, same input material, and so on. When units from different sources are mixed, the sampling plan
does not function properly. Also, it is difficult to take corrective action to eliminate the source of nonconforming
units.
(B) Lots should be as large as possible. Because sample sizes do not increase as rapidly as lot sizes, a lower
inspection cost results with larger lot sizes.
For example, a lot of 2000 would have a sample size of 125 (6.25%), but an equally effective sampling plan for a
lot of 4000 would have a sample size of 200 (5.00%). When an organization starts a just-in-time procurement
philosophy, the lot sizes are usually reduced to a 2- or 3-day supply. Thus, the relative amount inspected and the
inspection costs will increase. The benefits to just-in-time are far greater than the increase in inspection costs;
therefore, smaller lot sizes are to be expected.
FUNDAMENTAL CONCEPTS… 19

The reader is cautioned not to confuse the packaging requirements for shipment and materials
handling with the concept of a homogeneous lot.
In other words, a lot may consist of a number of packages and may also consist of a number of
shipments.
If two different machines and/or two different operators are included in a shipment, they are
separate lots and should be so identified.
The reader should also be aware that partial shipments of a homogeneous lot can be treated as
if they are homogeneous lots.
FUNDAMENTAL CONCEPTS… 20

SAMPLE SELECTION
The sample units selected for inspection should be representative of the entire lot. All sampling
plans are based on the premise that each unit in the lot has an equal likelihood of being selected.
This is referred to as random sampling.
The basic technique of random sampling is to assign a number to each unit in the lot. Then a
series of random numbers is generated that tells which of the numbered units are to be sampled
and inspected.
Random numbers can be generated from a computer, an electronic hand calculator, a 20-sided
random-number die, numbered chips in a bowl, and so on. They may be used to select the
sample or to develop a table of random numbers.
FUNDAMENTAL CONCEPTS… 21

A random-number table is shown in here as Table 1. To use the table, it is entered at any
location and numbers are selected sequentially from one direction, such as up, down, left, or
right. Any number that is not appropriate is discarded.
For locating convenience, this table is established with 5 digits per column. It could have been
established with 2, 3, 6, or any number per column.
In fact, the digits could have run across the page with no spaces, but that format would make
the table difficult to read.
Any number of digits can be used for a random number.
FUNDAMENTAL CONCEPTS… 22

An example will help to illustrate the technique.

Assume that a lot of 90 units has been assigned numbers from 1 to 90 and it is desired to select a sample of 9.
A two-digit number is selected at random, as indicated by the number 53.
Numbers are selected downward on the right of the third column, and the first three numbers are 53, 15, and
73.
Starting at the top of the next column, the numbers 45, 30, 06, 27, and 96 are obtained. The number 96 is too
high and is discarded.
The next numbers are 52 and 82. Units with the numbers 53, 15, 73, 45, 30, 06, 27, 52, and 82 comprise the
sample.
FUNDAMENTAL CONCEPTS… 23

Many units have serial numbers that can be used as the assigned number. This practice avoids
the difficult process of assigning numbers to each unit.
In many situations, units are systematically packed in a container and the assigned number can
be designated by the location.
A three-digit number would represent the width, height, and depth in a container as shown in
Figure 1.
Thus, the random number 328 could specify the unit located at the third row, second level, and
eighth unit from the front.
For fluid or other well-mixed products, the sample can be taken from any location, because the
product is presumed to be homogeneous.
 24

FIGURE 1 Location and

Random Numbers

FIGURE 2 Dividing a Lot for Stratified Sampling

FUNDAMENTAL CONCEPTS… 25

It is not always practical to assign a number to each unit, utilize a serial number, or utilize a
locational number.
Stratification of the lot or package with samples drawn from each stratum can be an effective
substitute for random sampling. The technique is to divide the lot or package into strata or layers
as shown in Figure 2.
Each stratum is further subdivided into cubes, as illustrated by stratum 1. Within each cube,
samples are drawn from the entire volume.
The dividing of the lot or package into strata and cubes within each stratum is an imaginary
process done by the inspector. By this technique, pieces are selected from all locations in the lot
or package.
FUNDAMENTAL CONCEPTS… 26

Unless an adequate sampling method is used, a variety of biases can occur. An example of a
biased sample occurs when the operator makes sure that units on the top of a lot are the best
quality, and the inspector selects the sample from the same location.
Adequate supervision of operators and inspectors is necessary to ensure that no bias occurs.
FUNDAMENTAL CONCEPTS… 27

NONACCEPTED LOTS
Once a lot has not been accepted, there are a number of courses of action that can be taken.
(1) The nonaccepted lot can be passed to the production facilities and the nonconforming units
sorted by production personnel.
This action is not a satisfactory alternative, because it defeats the purpose of sampling
inspection and slows production. However, if the units are badly needed, there may be no other
choice.
FUNDAMENTAL CONCEPTS… 28

(2) The nonaccepted lot can be rectified at the consumer’s plant by personnel from either the
producer’s or the consumer’s plant.
Although shipping costs are saved, there is a psychological disadvantage, because all the
consumer’s personnel are aware that producer X had product that was not accepted.
This fact may be used as a crutch to explain poor performance when using producer X ’s
material at a future time.
In addition, space at the consumer’s plant must be provided for personnel to perform the sorting
operation.
FUNDAMENTAL CONCEPTS… 29

(3) The nonaccepted lot can be returned to the producer for rectification.
This is the only appropriate course of action, because it results in long-run improvement in the
quality. Because shipping costs are paid in both directions, cost becomes a motivating factor to
improve the quality.
Also, when the lot is sorted in the producer’s plant, all the employees are aware that consumer Y
expects to receive quality units.
This, too, is a motivating factor for quality improvement the next time an order is produced for
consumer Y .
This course of action may require the production line to be shut down, which would be a loud and
clear signal to the supplier and operating personnel that quality is important.
FUNDAMENTAL CONCEPTS… 30

It is assumed that nonaccepted lots will receive 100% inspection and the nonconforming units
discarded.
A resubmitted lot is not normally reinspected, but if it is, the inspection should be confined to the
original nonconformity.
Because the nonconforming units are discarded, a resubmitted lot will have fewer units than the
original.
STATISTICAL ASPECTS 31

OC CURVE FOR SINGLE SAMPLING PLANS

An excellent evaluation technique is an operating characteristic (OC) curve .
In judging a particular sampling plan, it is desirable to know the probability that a lot submitted
with a certain percent nonconforming, 100p0, will be accepted.
The OC curve will provide this information, and a typical OC curve is shown in Figure 3.
When the percent nonconforming is low, the probability of the lot being accepted is large and
decreases as the percent nonconforming increases.
FIGURE 3: OC Curve for the Single Sampling Plan N = 3000, n = 89,
and c = 2 32
STATISTICAL ASPECTS… 33

The construction of an OC curve can be illustrated by a concrete example.

A single sampling plan has a lot size N = 3000, a sample size n = 89, and an acceptance
number c = 2.
It is assumed that the lots are from a steady stream of product that can be considered infinite,
and therefore the binomial probability distribution can be used for the calculations.
Fortunately, the Poisson is an excellent approximation to the binomial for almost all sampling
plans; therefore, the Poisson is used for determining the probability of the acceptance of a lot.
STATISTICAL ASPECTS… 34

In graphing the curve with the variables 100Pa (percent of lots accepted) and 100p0 (percent
nonconforming), one value, 100p0, will be assumed and the other calculated. For illustrative
purposes, we will assume a 100p0 value of, say, 2%, which gives an np0 value of
P0 = 0.02
np0 = (89)(0.02) = 1.8
STATISTICAL ASPECTS… 35

Acceptance of the lot is based on the acceptance number c = 2 and is possible when there are
0 nonconforming units in the sample, 1 nonconforming unit in the sample, or 2 nonconforming
units in the sample. Thus,
Pa = P0 + P1 + P2
= P2 or less
= 0.731 or 100Pa = 73.1%
The Pa value is obtained from Table C in the Appendix for c = 2 and np0 = 1.8.
Table C in the Appendix for c = 2 and np0 = 1.8 36
STATISTICAL ASPECTS… 37

A table can be used to assist with the calculations, as shown in Table 2 .

The curve is terminated when the Pa value is close to 0.05.
Because Pa = 0.055 for 100 p0 = 7%, it is not necessary to make any calculations for values
greater than 7%.
Approximately 7 points are needed to describe the curve with a greater concentration of points
where the curve changes direction.
TABLE 2: Probabilities of Acceptance for the Single Sampling Plan: n=89, c = 2 38
STATISTICAL ASPECTS… 39

Information from the table is plotted to obtain the OC

curve shown in Figure 3.
The steps are (1) assume p0 value, (2) calculate np0
value, (3) attain Pa values from the Poisson table using
Figure 3
the applicable c and np0 values, (4) plot point (100p0,
100Pa), and (5) repeat steps 1, 2, 3, and 4 until a smooth
curve is obtained.
To make the curve more readable, the label Percent of
Lots (expected to be) Accepted is used rather than
Probability of Acceptance.

FIGURE 3: OC Curve for the Single Sampling Plan N = 3000, n = 89, and c = 2
STATISTICAL ASPECTS… 40

Once the curve is constructed, it shows the chance of a lot being accepted for a particular
incoming quality.
Thus, if the incoming process quality is 2.3% nonconforming, the percent of the lots that are
expected to be accepted is 66%.
Similarly, if 55 lots from a process that is 2.3% nonconforming are inspected using this sampling
plan, 36 [(55)(0.66) = 36] will be accepted and 19 [55 - 36 = 19] will be unacceptable.
This OC curve is unique to the sampling plan defined by N = 3000, n = 89, and c = 2.
If this sampling plan does not give the desired effectiveness, then the sampling plan should be
changed and a new OC curve constructed and evaluated.
STATISTICAL ASPECTS… 41

OC CURVE FOR DOUBLE SAMPLING PLANS

The construction of an OC curve for double sampling plans is somewhat more involved,
because two curves must be determined.
One curve is for the probability of acceptance on the first sample; the second curve is the
probability of acceptance on the combined samples.
A typical OC curve is shown in Figure 4 for the double sampling plan N = 2400, n1 = 150, c1 = 1,
r1 = 4, n2 = 200, c2 = 5, and r2 = 6.
The first step in the construction of the OC curve is to determine the equations.
If there is one or fewer nonconforming unit on the first sample, the lot is accepted. Symbolically,
the equation is (Pa)I = (P 1 or less)I
FIGURE 4: OC Curve for Double Sampling Plan 42
STATISTICAL ASPECTS… 43

To obtain the equation for the second sample, the number of different ways in which the lot can
be accepted is determined.
A second sample is taken only if there are 2 or 3 nonconforming units on the first sample.
If there is 1 or less, the lot is accepted; if there are 4 or more, the lot is not accepted. Therefore,
the lot can be accepted by obtaining
(1) Two nonconforming units on the first sample and 3 or less nonconforming units on the
second sample, or
(2) Three nonconforming units on the first sample and 2 or less nonconforming units on the
second sample.
STATISTICAL ASPECTS… 44

The and’s and or’s are emphasized to illustrate the use of the additive and multiplicative
theorems.
Where an and occurs, multiply, and where an or occurs, add, and the equation becomes
(Pa)II = (P2)I(P3 or less)II + (P3)I(P2 or less)II
Roman numerals are used as subscripts for the sample number.
The equations derived above are applicable only to this double sampling plan; another plan will
require a different set of equations.
STATISTICAL ASPECTS… 45

Figure 5 graphically illustrates the technique.

Note that the number of nonconforming units in each term in the second equation is equal to or
less than the acceptance number, c2.
By combining the equations, the probability of acceptance for the combined samples is
obtained:
(Pa)combined = (Pa)I + (Pa)II
FIGURE 5: Graphical Description of the Double Sampling Plan:
N = 2400, n1 = 150, c1 = 1, r1 = 4, n2 = 200, c2 = 5, and r2 = 6 46
STATISTICAL ASPECTS… 47

Once the equations are obtained, the OC curves are found by assuming various P0 values and
calculating the respective first and second sample Pa values.
For example, using Table C and assuming a p0 value of 0.01 (100p0 = 1.0),
(np0)I = (150)(0.01) = 1.5, (np0)II = (200)(0.01) = 2.0
(Pa)I = (P1 or less)I = 0.558
(Pa)II = (P2)I(P3 or less)II + (P3)I(P2 or less)II
(Pa)II = (0.251)(0.857) + (0.126)(0.677) = (Pa)II = 0.300
(Pa)combined = (Pa)I + (Pa)II ; (Pa)combined = 0.558 + 0.300 ; (Pa)combined = 0.858
STATISTICAL ASPECTS… 48

These results are illustrated in Figure 4. When

the two sample sizes are different, the np0
values are different, which can cause a
calculating error.
Another source of error is neglecting to use the
“or less” probabilities. Calculations are usually
to three decimal places.
The remaining calculations for other points on
the curve are
STATISTICAL ASPECTS… 49
STATISTICAL ASPECTS… 50

Similar to the construction of the OC curve for single sampling, points are plotted as they are
calculated, with the last few calculations used for locations where the curve changes direction.
Whenever possible, both sample sizes should be the same value to simplify the calculations and
the inspector’s job.
Also, if r1 and r2 are not given, they are equal to c2 + 1.
The steps are (1) assume p0 value, (2) calculate (np0)I and (np0)II values, (3) determine Pa value
using the three equations and Table C, (4) plot points, and (5) repeat steps 1, 2, 3, and 4 until a
smooth curve is obtained.
OC Curve for Multiple Sampling Plans 51

The construction of an OC curve for multiple

sampling plans is more involved than for
double or single sampling plans; however, the
technique is the same. A multiple sampling
plan with four levels is illustrated in Figure 6
and is specified as

FIGURE 6: OC Curve for a Multiple Sampling Plan

OC Curve for Multiple Sampling Plans 52

Using the equations above and

varying the fraction
nonconforming, p0, the OC
curve of Figure 6 is constructed.
This is a tedious task and one
that is ideally suited for the
computer.
Comment 53

 An operating characteristic curve evaluates the effectiveness of a particular sampling plan. If

that sampling plan is not satisfactory, as shown by the OC curve, another one should be
selected and its OC curve constructed.
 Because the process quality or lot quality is usually not known, the OC curve, as well as other
curves in this chapter, is a “what if” curve. In other words, if the quality is a particular percent
nonconforming, the percent of lots accepted can be obtained from the curve.
Difference Between Type A and Type B OC Curves 54

The OC curves that were constructed in the previous sections are type B curves.
It was assumed that the lots came from a continuous stream of product, and therefore the
calculations are based on an infinite lot size.
The binomial is the exact distribution for calculating the acceptance probabilities; however, the
Poisson was used because it is a good approximation.
Type B curves are continuous.
Type A curves give the probability of accepting an isolated finite lot. With a finite situation, the hyper
geometric is used to calculate the acceptance probabilities. As the lot size of a type A curve
increases, it approaches the type B curve and will become almost identical when the lot size is at
least 10 times the sample size (n/N≤0.10).
Difference Between Type A and Type B OC Curves… 55

A type A curve is shown in Figure 7, with the FIGURE 7: Types A and B OC Curves
small open circles representing the discrete
data and a discontinuous curve; however, the
curve is drawn as a continuous one. Thus, a 4%
value is impossible, because it represents 2.6
nonconforming units in the lot of 65 [(0.04)(65)
= 2.6], but 4.6% nonconforming units are
possible, as it represents 3 nonconforming units
in the lot of 65 [(0.046)(65) = 3.0]. Therefore,
the “curve” exists only where the small open
circles are located.
 56

In comparing the type A and type B curves of Figure 7, the type A curve is always lower than the
type B curve.
When the lot size is small in relation to the sample size, the difference between the curves is
significant enough to construct the type A curve.
Unless otherwise stated, all discussion of OC curves will be in terms of type B curves.
OC Curve Properties 57

Acceptance sampling plans with similar properties can give different OC curves. Four of these
properties and the OC curve information are given in the information that follows.
(1) Sample size as a fixed percentage of lot size. Prior to the use of statistical concepts for
acceptance sampling, inspectors were frequently instructed to sample a fixed percentage of the
lot. If this value is, say, 10% of the lot size, plans for lot sizes of 900, 300, and 90 are N = 900, n
= 90, c = 0; N = 300, n = 30, c = 0; N = 90, n = 9, c = 0 .
Figure 8 shows the OC curves for the three plans, and it is evident that they offer different levels
of protection. For example, for a process that is 5% nonconforming, 100Pa = 2% for lot sizes of
900, 100Pa = 22% for lot sizes of 300, and 100Pa = 63% for lot sizes of 90.
OC Curve Properties… 58

FIGURE 8: OC Curves for Sample Sizes FIGURE 9: OC Curves for Fixed

that are 10% of the Lot Size Sample Size (Type A)
OC Curve Properties… 59

(2) Fixed sample size. When a fixed or constant sample size is used, the OC curves are very
similar. Figure 9 illustrates this property for the type A situation where n = 10% of N. Naturally, for
type B curves or when n < 10% of N , the curves are identical. The sample size has more to do
with the shape of the OC curve and the resulting quality protection than does the lot size.
(3) As sample size increases, the curve becomes steeper. Figure 10 illustrates the change in the
shape of the OC curve. As the sample size increases, the slope of the curve becomes steeper
and approaches a straight vertical line. Sampling plans with large sample sizes are better able to
discriminate between acceptable and unacceptable quality. Therefore, the consumer has fewer
lots of unacceptable quality accepted and the producer fewer lots of acceptable quality that are
not accepted.
OC Curve Properties… 60

(4) As the acceptance number decreases, the curve becomes steeper. The change in the shape
of the OC curve as the acceptance number changes is shown in Figure 11. As the acceptance
number decreases, the curve becomes steeper. This fact has frequently been used to justify the
use of sampling plans with acceptance numbers of 0. However, the OC curve for N = 2000, n =
300, and c = 2, which is shown by the dashed line, is steeper than the plan with c = 0.
A disadvantage of sampling plans with c = 0 is that their curves drop sharply down rather than
have a horizontal plateau before descending. Because this is the area of the producer’s risk
(discussed in the next section), sampling plans with c = 0 are more demanding of the producer.
OC Curve Properties… 61

FIGURE 10: FIGURE 11:

OC Curves OC Curves Illustrating
Illustrating Change Change in
in Sample Size Acceptance Number
OC Curve Properties… 62

Sampling plans with acceptance numbers greater than 0 can actually be superior to those with 0;
however, these require a larger sample size, which is more costly.
In addition, many producers have a psychological aversion to plans that do not accept lots when
only one nonconforming unit is found in the sample.
The primary advantage of sampling plans with c = 0 is the perception that nonconforming units
will not be tolerated and should be used for critical nonconformities.
For major and minor nonconformities, acceptance numbers greater than 0 should be considered.
Consumer–Producer Relationship 63

When acceptance sampling is used, there is a conflicting interest between the consumer and the
producer. The producer wants all conforming lots accepted, and the consumer wants all
nonconforming lots not accepted.
Only an ideal sampling plan that has an OC curve that is a vertical line can satisfy both the
producer and consumer.
An “ideal” OC curve, as shown in Figure 12, can be achieved only with 100% inspection, and the
pitfalls of this type of inspection were mentioned earlier in the chapter.
Therefore, sampling carries risks of not accepting lots that are acceptable and of accepting lots that
are unacceptable. Because of the seriousness of these risks, various terms and concepts have
been standardized.
Consumer–Producer Relationship… 64

The producer’s risk, which is represented by the symbol α, is the probability of nonacceptance
of a conforming lot. This risk is frequently given as 0.05, but it can range from 0.001 to 0.10 or
more.
Because a is expressed in terms of the probability of nonacceptance, it cannot be located on an
OC curve unless specified in terms of the probability of acceptance.
This conversion is accomplished by subtracting from 1.
Thus, Pa = 1 - α, and for α = 0.05, Pa = 1 - 0.05 = 0.95.
Figure 13 shows the producer’s risk, α, or 0.05 on an imaginary axis labeled “Percent of Lots
Not Accepted.” (Note: α=alpha)
 65

FIGURE 12: Ideal OC Curve FIGURE 13: Consumer–Producer Relationship

Consumer–Producer Relationship… 66

Associated with the producer’s risk is a numerical definition of an acceptable lot, which is called
acceptance quality limit (AQL). The AQL is the quality level that is the worst tolerable process average
when a continuing series of lots is submitted for acceptance sampling.
It is a reference point on the OC curve and is not meant to convey to the producer that any percent
nonconforming is acceptable.
It is a statistical term and is not meant to be used by the general public. The only way the producer can be
guaranteed that a lot will be accepted is to have 0% nonconforming or to have the number nonconforming
in the lot less than or equal to the acceptance number.
In other words, the producer’s quality goal is to meet or exceed the specifications so that no
nonconforming units are present in the lot.
Consumer–Producer Relationship… 67

For the sampling plan N = 4000, n = 300, and c = 4, the AQL = 0.7% for 100 = 5%, as shown in
Figure 13.
In other words, units that are 0.7% nonconforming will have a nonacceptance probability of 0.05, or
5%.
Or, stated another way, 1 out of 20 lots that are 0.7% nonconforming will not be accepted by the
sampling plan.
The consumer’s risk, represented by the symbol β, is the probability of acceptance of a
nonconforming lot. This risk is frequently given as 0.10.
Because β is expressed in terms of probability of acceptance, no conversion is necessary. (Note:
β=Beta)
Consumer–Producer Relationship… 68

Associated with the consumer’s risk is a numerical definition of a nonconforming lot, called
limiting quality (LQ).
The LQ is the percent nonconforming in a lot or batch for which, for acceptance sampling
purposes, the consumer wishes the probability of acceptance to be low.
For the sampling plan in Figure 13 , the LQ = 2.6% for 100β = 10%. In other words, lots that are
2.6% nonconforming will have a 10% chance of being accepted.
Or, stated another way, 1 out of 10 lots that are 2.6% nonconforming will be accepted by this
sampling plan.
 69

AVERAGE OUTGOING QUALITY

The average outgoing quality (AOQ)
is another technique for the
evaluation of a sampling plan.
Figure 14 shows an AOQ curve for
the sampling plan N = 3000, n = 89,
and c = 2.
This is the same plan as the one for
the OC curve shown in Figure 3. .

FIGURE 14: Average Outgoing Quality Curve for the Sampling Plan N = 3000, n = 89, and c = 2
 70

The information for the construction of an average outgoing quality curve is obtained by adding one
column (an AOQ column) to the table used to construct an OC curve.
Table 3 shows the information for the OC curve and the additional column for the AOQ curve.
The average outgoing quality in percent nonconforming is determined by the formula AOQ = (100p 0)
(Pa).
This formula does not account for the discarded nonconforming units; however, it is close enough for
practical purposes and is simpler to use.
Note that to present a more readable graph, the AOQ scale is much larger than the incoming process
quality scale. The curve is constructed by plotting the percent nonconforming (100 p0) with its
corresponding AOQ value.
TABLE 3: Average Outgoing Quality (AOQ) for the Sampling Plan
N=3000, n=89,and c=2 71
 72

The AOQ is the quality that leaves the inspection operation.

It is assumed that any nonaccepted lots have been rectified or sorted and returned with 100%
good product.
When rectification does not occur, the AOQ is the same as the incoming quality, and this
condition is represented by the straight line in Figure 14.
 73

• Analysis of the curve shows that when the incoming quality is 2.0% nonconforming, the
average outgoing quality is 1.46% nonconforming, and when the incoming quality is 6.0%
nonconforming, the average outgoing quality is 0.64% nonconforming.
• Therefore, because nonaccepted lots are rectified, the average outgoing quality is always
better than the incoming quality.
• In fact, there is a limit that is given the name average outgoing quality limit (AOQL). Thus, for
this sampling plan, as the percent nonconforming of the incoming quality changes, the average
outgoing quality never exceeds the limit of approximately 1.6% nonconforming.
 74

A better understanding of the concept of acceptance sampling can be obtained from an example.
Suppose that, over a period of time, 15 lots of 3000 each are shipped by the producer to the
consumer. The lots are 2% nonconforming and a sampling plan of n = 89 and c = 2 is used to
determine acceptance. Figure 15 shows this information by a solid line.
The OC curve for this sampling plan (Figure 3) shows that the percent of lots accepted for a 2%
nonconforming lot is 73.1%. Thus, 11 lots (15 * 0.731 = 10.97) are accepted by the consumer, as
shown by the wavy line.
Four lots are not accepted by the sampling plan and are returned to the producer for rectification,
as shown by the dashed line. These four lots receive 100% inspection and are returned to the
consumer with 0% nonconforming, as shown by a dashed line.
 75

FIGURE 15: How Acceptance Sampling Works

 76

A summary of what the consumer actually receives is shown at the bottom of the figure. Two
percent, or 240, of the four rectified lots are discarded by the producer, which gives 11,760 rather
than 12,000. The calculations show that the consumer actually receives 1.47% nonconforming,
whereas the producer’s quality is 2% nonconforming.
It should be emphasized that the acceptance sampling system works only when nonaccepted lots
are returned to the producer and rectified.
The AQL for this particular sampling plan at a = 0.05 is 0.9%; therefore, the producer at 2%
nonconforming is not achieving the desired quality level.
The AOQ curve, in conjunction with the OC curve, provides two powerful tools for describing and
analyzing acceptance sampling plans.
 77

AVERAGE SAMPLE NUMBER

The average sample number (ASN) is a comparison of the average amount inspected per lot by
the consumer for single, double, multiple, and sequential sampling. Figure 16 shows the
comparison for the four different but equally effective sampling plan types.
In single sampling the ASN is constant and equal to the sample size, n.
For double sampling the process is somewhat more complicated because a second sample may
or may not be taken.
The formula for double sampling is ASN = n1 + n2(1 - P1), where P1 is the probability of a decision
on the first sample. An example problem will illustrate the concept.
Example Problem 1 78

Given the single sampling plan n = 80 and c = 2 and the equally effective double sampling plan n1 = 50, c1 = 0, r1 = 3, n2 =
50, c2 = 3,and r2 = 4, compare the ASN of the two by constructing their curves.
For single sampling, the ASN is the straight line at n = 80.
For double sampling, the solution is
P1 = P0 + P 3 or more
Assume that p0 = 0.01; then np0 = 50(0.01) = 0.5.
From Appendix Table C. P0 = 0.607
P 3 or more = 1 - P2 or less = 1 - 0.986 = 0.014
ASN = n1 + n2(1 - [P0 + P3 or more]) = 50 + 50(1 - [0.607 + 0.014]) = 69
Repeating for different values of p0,the double sampling plan is plotted as shown in Figure 16 .
FIGURE 16: ASN Curves for Single, Double, Multiple, and Sequential
Sampling 79
 80

The formula assumes that inspection continues even after the non acceptance number is
reached.
It is frequently the practice to discontinue inspection after the non acceptance number is reached
on either the first or second sample.
This practice is called curtailed inspection, and the formula is much more complicated.
Thus, the ASN curve for double sampling is somewhat lower than what actually occurs.
 81

An analysis of the ASN curve for double sampling in Figure 16 shows that at a fraction
nonconforming of 0.03, the single and double sampling plans have about the same amount of
inspection.
For fraction nonconforming less than 0.03, double sampling has less inspection because a decision
to accept on the first sample is more likely.
Similarly, for fraction nonconforming greater than 0.03, double sampling has less inspection
because a decision not to accept on the first sample is more likely and a second sample is not
required.
It should be noted that in most ASN curves, the double sample curve does not get close to the
single sample one.
 82

Calculation of the ASN curve for multiple sampling is much more difficult than for double
sampling. The formula is
ASN = n1P1 + (n1 + n2)PII + …. + (n1 + n2 + …. + nk) Pk
where nk is the sample size of the last level and Pk is the probability of a decision at the last level.
Determining the probabilities of a decision at each level is quite involved-more so than for the OC
curve, because the conditional probabilities must also be determined.
Figure 16 shows the ASN curve for an equivalent multiple sampling plan with seven levels. As
expected, the average amount inspected is much less than single or double sampling.
 83

The reader may be curious about the two extra scales in Figure 16.
Because we are comparing equivalent sampling plans, the double and multiple plans can be
related to the single sampling plans where c = 2 and n is the equivalent single sample size by the
additional scales.
To use the horizontal scale, multiply the single sample size n by the fraction nonconforming. The
ASN value is found from the vertical scale by multiplying the scale fraction with the single sample
size.
 84

ANSI/ASQ Z1.4 shows a number of ASN curve comparisons indexed by the acceptance number,
c.
These curves can be used to find the amount inspected per lot for different percent
nonconforming without having to make the calculations.
When inspection costs are great because of inspection time, equipment costs, or equipment
availability, the ASN curves are a valuable tool for justifying double or multiple sampling.
 85

AVERAGE TOTAL INSPECTION

The average total inspection (ATI) is another technique for evaluating a sampling plan. ATI is the
amount inspected by both the consumer and the producer.
Like the ASN curve, it is a curve that provides information on the amount inspected and not on
the effectiveness of the plan.
For single sampling, the formula is
ATI = n + (1 - Pa)(N - n)
 86

It assumes that rectified lots will receive 100% inspection.

If lots are submitted with 0% nonconforming, the amount inspected is equal to n, and if lots are
submitted that are 100% nonconforming, the amount inspected is equal to N.
Because neither of these possibilities is likely to occur, the amount inspected is a function of the
probability of nonacceptance (1 - Pa). An example problem will illustrate the calculation.
Example Problem 2 87

Determine the ATI curve for the single sampling plan N = 3000,
n = 89, and c = 2.
Assume that p0 = 0.02. From the OC curve ( Figure 3 ), Pa =
0.731.
ATI = n + (1 - Pa)(N - n) = 89 + (1 - 0.731)(3000 - 89) = 872
Repeat for other p0 values until a smooth curve is obtained, as
shown in Figure 17.

FIGURE 17: ATI Curve for N = 3000, n = 89, and c = 2

 88

Examination of the curve shows that when the process quality is close to 0% nonconforming, the
average total amount inspected is close to the sample size n.
When process quality is very poor, at, say, 9% nonconforming, most of the lots are not accepted,
and the ATI curve becomes asymptotic to 3000.
As the percent nonconforming increases, the amount inspected by the producer dominates the
curve.
 89

Double sampling and multiple sampling formulas for the ATI curves are more complicated. These
ATI curves will be slightly below the one for single sampling.
The amount below is a function of the ASN curve, which is the amount inspected by the
consumer, and this amount is usually very small in relation to the ATI, which is dominated by the
amount inspected by the producer.
From a practical viewpoint, the ATI curves for double and multiple sampling are not necessary,
because the equivalent single sampling curve will convey a good estimate.
SAMPLING PLAN DESIGN 90

SAMPLING PLANS FOR STIPULATED PRODUCER’S RISK

When the producer’s risk a and its corresponding AQL are specified, a sampling plan or, more
precisely, a family of sampling plans can be determined.
For a producer’s risk, α, of, say, 0.05 and an AQL of 1.2%, the OC curves for a family of sampling
plans as shown in Figure 18 are obtained.
Each of the plans passes through the point defined by 100Pa = 95% (100α = 5%) and p0.95 =
0.012.
Therefore, each of the plans will ensure that product 1.2% nonconforming will not be accepted
5% of the time or, conversely, accepted 95% of the time.
 91

The sampling plans are obtained by assuming a value for c and finding its corresponding np0
value from Table C.
When np0 and p0 are known, the sample size n is obtained. In order to find the np0 values using
Table C, interpolation is required.
To eliminate the interpolation operation, np0 values for various α and β values are reproduced in
Table 4.
In this table, c is cumulative, which means that a c value of 2 represents 2 or less.
 92

Calculations to obtain the three sampling plans of Figure 18 are as follows

FIGURE 18: Single Sampling Plans for Stipulated Producer’s Risk

and AQL
TABLE 4: np Values for Corresponding c Values and Typical Producer’s
and Consumer’s Risks 93
 94

While all the plans provide the same protection for the producer, the consumer’s risk, at, say, β = 0.10, is
quite different.
From Figure 18 for the plan c = 1, n = 30, product that is 13% nonconforming will be accepted 10% (β =
0.10) of the time; for the plan c = 2, n = 68, the product that is 7.8% nonconforming will be accepted 10%
(β= 0.10) of the time; and, for the plan c = 6, n = 274, product that is 3.8% nonconforming will be accepted
10% (β = 0.10) of the time.
From the consumer’s viewpoint, the latter plan provides better protection; however, the sample size is
greater, which increases the inspection cost.
The selection of the appropriate plan to use is a matter of judgment, which usually involves the lot size.
This selection would also include plans for c = 0, 3, 4, 5, 7, and so forth.
 95

Sampling Plans for Stipulated Consumer’s Risk

When the consumer’s risk b and its corresponding LQ are specified, a family of sampling plans
can be determined. For a consumer’s risk, β, of, say, 0.10 and a LQ of 6.0%, the OC curves for a
family of sampling plans as shown in Figure 19 are obtained. Each of the plans passes through
the point defined by Pa = 0.10 (β = 0.10) and p0.10 = 0.060. Therefore, each of the plans will
ensure that product 6.0% nonconforming will be accepted 10% of the time.
 96

The sampling plans are determined in the

same manner as used for a stipulated
producer’s risk.
Calculations are as follows:
 97

The sampling plans for c = 1, c = 3, and c = 7 were arbitrarily

selected to illustrate the technique.
Although all the plans provide the same protection for the
consumer, the producer’s risk, at, say, a = 0.05, is quite
different. From Figure 19 for the plan c = 1, n = 65, product
that is 0.5% nonconforming will not be accepted 5% (100α =
5%) of the time; for the plan c = 3, n = 111, product that is
1.2% nonconforming will not be accepted 5% (100α
= 5%) of the time; and for the plan c = 7, n = 196,
product that is 2.0% nonconforming will not be
accepted 5% (α = 0.05) of the time.
FIGURE 19: Single Sampling Plans for
Stipulated Consumer’s Risk and LQ
 98

From the producer’s viewpoint, the latter plan provides better protection; however, the sample
size is greater, which increases the inspection costs. The selection of the appropriate plan is a
matter of judgment, which usually involves the lot size. This selection would also include plans for
c = 0, 2, 4, 5, 6, 8, and so forth.
 99

SAMPLING PLANS FOR STIPULATED PRODUCER’S AND CONSUMER’S RISK

Sampling plans are also stipulated for both the consumer’s risk and the producer’s risk. It is difficult
to obtain an OC curve that will satisfy both conditions.
More than likely there will be four sampling plans that are close to meeting the consumer’s and
producer’s stipulations.
Figure 20 shows four plans that are close to meeting the stipulations of a = 0.05, AQL = 0.9 and b =
0.10, LQ = 7.8.
The OC curves of two plans meet the consumer’s stipulation that product that is 7.8%
nonconforming (LQ) will be accepted 10% (β= 0.10) of the time and comes close to the producer’s
stipulation.
 100

These two plans are shown by the dashed lines in Figure 20 and are c = 1, n = 50 and c = 2, n =
68. The two other plans exactly meet the producer’s stipulation that product that is 0.9%
nonconforming (AQL) will not be accepted 5% (α= 0.05) of the time. These two plans are shown
by the solid lines and are c = 1, n = 39 and c = 2, n = 91.
In order to determine the plans, the first step is to find the ratio of p0.10 ÷ p0.95, which is
p0.10 ÷ p0.95 = 0.078 ÷ 0.009 = 8.667
From the ratio column of Table 4 , the ratio of 8.667 falls between the row for c = 1 and the row
for c = 2.
Thus, plans that exactly meet the consumer’s stipulation of LQ = 7.8% for β= 0.10 are
101
 102

Which of the four plans to select is based on one of four additional criteria? The first additional
criterion is the stipulation that the plan with the lowest sample size be selected.
The plan with the lowest sample size is one of the two with the lowest acceptance number.
Thus, for the example problem, only the two plans for c = 1 are calculated, and c = 1, n = 39 is
the sampling plan selected. A second additional criterion is the stipulation that the plan with the
greatest sample size be selected.
The plan with the greatest sample size is one of two with the largest acceptance number.
Thus, for the example problem, only the two plans for c = 2 are calculated, and c = 2, n = 91 is
the sampling plan selected.
 103

A third additional criterion is the stipulation that the plan exactly meets the consumer’s stipulation
and comes as close as possible to the producer’s stipulation. The two plans that exactly meet the
consumer’s stipulation are c = 1, n = 50 and c = 2, n = 68. Calculations to determine which plan
is closest to the producer’s stipulation of AQL = 0.9, α= 0.05 are

Because p0.95 = 0.007 is closest to the stipulated value of 0.009, the plan of c = 1, n = 50 is
selected.
 104

The fourth additional criterion for the selection of one of the four sampling plans is the stipulation that
the plan exactly meet the producer’s stipulation and comes as close as possible to the consumer’s
stipulation. The two plans that are applicable are c = 1, n = 39 and c = 2, n = 91. Calculations to
determine which is the closest to the consumer’s stipulation of LQ = 7.8, β= 0.10 are

Because p0.10 = 0.058 is closest to the stipulated value of 0.078, the plan of c = 2, n = 91 is selected.
Some Comments 105

The previous discussions have concerned single sampling plans. Double and multiple sampling plan design, although
more difficult, will follow similar techniques.
In the previous discussion, a producer’s risk of 0.05 and a consumer’s risk of 0.10 were used to illustrate the technique.
The producer’s risk is usually set at 0.05 but can be as small as 0.01 or as high as 0.15.
The consumer’s risk is usually set at 0.10 but can be as low as 0.01 or as high as 0.20.
Sampling plans can also be specified by the AOQL. If an AOQL of 1.5% for an incoming quality of, say, 2.0% is
stipulated, the probability of acceptance is
AOQL = (100p0)(Pa)
1.5 = 2.0Pa
Pa = 0.75 or 100Pa = 75%
 106

Figure 21 shows a family of OC curves for various

sampling plans that satisfy the AOQL criteria.
To design a sampling plan, some initial stipulations are
necessary by the producer, consumer, or both. These
stipulations are decisions based on historical data,
experimentation, or engineering judgment.
In some cases the stipulations are negotiated as part of
the purchasing contract.

FIGURE 21: AOQL Sampling Plans

SAMPLING PLAN SYSTEMS 107

The task of designing a sampling plan system is a tedious one.

Fortunately, 10 sampling plan systems are available.
(1) ANSI/ASQ Z1.4 108

One such system that is almost universally used for the acceptance of product is ANSI/ASQ
Z1.4. This system is an AQL, or producer’s risk system.
The standard is applicable, but not limited, to attribute inspection of the following:
(1) end items, (2) components and raw materials, (3) operations, (4) materials in process, (5)
supplies in storage, (6) maintenance operations, (7) data or records, and (8) administrative
procedures. Sampling plans of this standard are intended to be used for a continuing series of
lots, but plans may be designed for isolated lots by consulting the operating characteristic (OC)
curve to determine the plan with the desired protection.
(1) ANSI/ASQ Z1.4 … 109

The standard provides for three types of sampling: single, double, and multiple.
For each type of sampling plan, provision is made for normal, tightened, or reduced inspection.
Tightened inspection is used when the producer’s recent quality history has deteriorated.
Acceptance requirements under tightened inspection are more stringent than under normal
inspection.
Reduced inspection is used when the producer’s recent quality history has been exceptionally
good.
(2) ANSI/ASQ Standard Q3 110

This standard is to be used for inspection of isolated lots by attributes.

It complements ANSI/ASQ Z1.4 that is appropriate for a continuous stream of lots.
This standard indexes tables by LQ values and is applicable to type A or type B lots or batches.
The LQ values are determined by the same techniques used to determine AQL values
(3) Dodge-Romig Tables 111

In the 1920s, H. F. Dodge and H. G. Romig developed a set of inspection tables for the lot-by-lot
acceptance of product by sampling for attributes. These tables are based on LQ and AOQL.
For each of these concepts there are tables for single and double sampling.
No provision is made for multiple sampling.
The principal advantage of the Dodge–Romig tables is a minimum amount of inspection for a
given inspection procedure.
This advantage makes the tables desirable for in-house inspection.
(4) Chain Sampling 112

A special type of lot-by-lot acceptance sampling plan for attributes was developed by H. F. Dodge
called Chain Sampling.
It is applicable to quality characteristics that involve destructive or costly tests where the
acceptance number is zero and sample size is small.
These plans make use of the cumulative results of previous samples.
(5) Sequential Sampling 113

Sequential sampling is similar to multiple sampling except that sequential sampling can
theoretically continue indefinitely.
In practice, the plan is truncated after the number inspected is equal to three times the number
inspected by a corresponding single sampling plan.
Sequential sampling, which is used for costly or destructive tests, usually has a subgroup size of
1, thereby making it an item-by-item plan.
(6) Skip-Lot Sampling 114

Skip-lot sampling was devised by H. F. Dodge in 1955.

It is a single sampling plan based on the AOQL for minimizing inspection costs when there is a
continuing supply of lots from the same source.
It is particularly applicable to chemical and physical characteristics that require laboratory
analyses.
As companies emphasize statistical process control (SPC) and just-in-time (JIT) procurement,
this type of sampling has become more applicable.
(7) ANSI / ASQ S1 115

The purpose of this standard is to provide procedures to reduce the inspection effort when the
supplier’s quality is superior.
It is a skip-lot scheme used in conjunction with the attribute lot-by-lot plans given in
ANSI/ASQZ1.4; it is not to be confused with Dodge’s skip-lot scheme described in the previous
section.
This sampling plan is an alternate to the reduced inspection of ANSI/ASQZ1.4, which permits
smaller sizes than normal inspection.
(8) MIL-STD-1235B 116

The standard is composed of five different continuous sampling plans. Inspection is by attributes
for nonconformities or nonconforming units using three classes of severity: critical, major, and
minor.
The continuous sampling plans are designed based on the AOQL.
In order to be comparable with ANSI/ASQ Z1.4 and other standards, the plans are also indexed
by the AQL.
The AQL is merely an index to the plans and has no other meaning.
(9) Shainin Lot Plot Plan 117

The Shainin lot plot plan is a variable sampling plan used in some industries. It was developed by
Dorian Shainin while he was chief inspector at Hamilton Standard Division of United Aircraft
Corporation.
The plan uses a plotted frequency distribution (histogram) to evaluate a sample for decisions
concerning acceptance or nonacceptance of a lot.
The most significant feature of the plan is the fact that it is applicable to both normal and
nonnormal frequency distributions.
Another feature is its simplicity. It is a practical plan for in-house inspection as well as receiving
inspection.
(10) ANSI/ASQ Z1.9 118

This standard is a lot-by-lot acceptance sampling plan by variables and closely matches
ANSI’ASQ Z1.4 for attributes.
The standard is indexed by numerical values of the AQL that range from 0.10 to 10.0%.
Provision is made for single and double sampling with normal, tightened, and reduced
inspections.
Sample sizes are a function of the lot size and the inspection level.
The standard assumes a normally distributed random variable.