Boston University

Metropolitan College

AD 734: Quality Management

Lecture 11: Supplier Quality Assurance

Spring 2024
Topics

● Managing Suppler Quality

● Acceptance Sampling
– AQL vs. AOQL
– MIL-STD-105 Standards
– limitations of acceptance sampling

● Deming’s k1,k2 Rule

– game

● Supplier Certification
– quality audits
2
Learning Objectives

1. Develop a quality system that includes all of a firm’s supply

chain members, including external suppliers.

2. Describe the use of acceptance sampling and its limitations.

3. Implement MIL-STD-105 standards.

4. Understand the implications of Deming’s k1k2 rule.

5. Implement a supplier certification system.

6. Conduct a quality audit.

3
 Supplier Quality Assurance
● Producers (of both products and services) have only indirect
control over external supplier processes.
● Two options exist for managing supplier quality:
– a process-thinking approach that certifies suppliers (similar to
quality certification, but more focused on specific products).
– a product-thinking approach that attempts to identify
substandard products before that are used by the producer.
● The process-thinking approach is better but may be difficult
to implement with some suppliers
– for example, your firm may be a low volume or low priority
customer of the suppler.
4
Product-Thinking System Options
● No inspection
– generally used when inspection costs are high (e.g.,
destructive tests) or when external failure costs are low.
● Sampling inspection
– traditional approach uses acceptance sampling.
● 100% inspection
– generally used when external failure costs are high or
inspection costs are low.
● Multiple 100% inspection
– generally used when external failures result in death or
potential litigation (and inspection errors exist).
5
 Acceptance Sampling

● Acceptance sampling generally involves taking a random

sample from a batch (also referred to as a “lot”):
– the entire lot is either accepted or rejected based on the sample.
– rejected lots are typically returned to the supplier.
● Acceptance sampling most often uses attribute inspections:
– typical “single sampling” plan calls for inspecting each item in a
sample (of n items), and rejecting the lot if the number of
nonconforming items found in the sample exceeds an acceptance
number (abbreviated Ac).

6
Acceptance Sampling - Description
The lot is either accepted
or rejected depending
on the number of
nonconformances in the
sample.

Supplier Sample
Process

Lot (or Batch) 7

Acceptance Sampling (for Attributes)

Inspect
Receive n items yes Reject Rejected lots
Lot (let d=# d>Ac? are sent back to
Lot the producer
(Size N) nonconfor
-mances)
no

Accept Lot
Accepted lots are
received by the
customer

Ac = “Acceptance Number” based on the acceptance sampling plan. 8

Acceptance Sampling “Errors”

● A Type I error occurs when the lot meets quality

standards, but the lot is rejected:
– the probability of a Type I error is denoted by 
–  is referred to as the producer’s risk

● A Type II error occurs when the lot does not meet quality
standards, but the lot is accepted:
– the probability of a Type II error is denoted by 
–  is referred to as the consumer’s risk

9
Acceptance Sampling Performance

● Standard acceptance sampling plans are often employed

based on inspection costs, target quality levels, and lot
sizes.
● Most acceptance sampling plans limit the producer’s risk,
and therefore  is usually low.
● The performance of the acceptance sampling plan is based
on the “probability of acceptance “ (Pa) for sub-standard
batches of product.
– this value is called the operating characteristic (OC).

10
Theoretical Foundation: Attributes
● For proportion data (fraction nonconforming in the lot is
p), the binomial probability model can be used to forecast
acceptance sampling plan results.
– in Excel: Pa=BINOMDIST(Ac,n,p,true)
● For count data (average unit count in the lot is u), the
Poisson probability model can be used to forecast
acceptance sampling plan results.
– in Excel: Pa=POISSON(Ac,nu,true)
● Formulas valid as long as n is no more than 15% of the lot
size (N).
11
Example (n=20, Ac=1)
p Pa
0.00 1.000
0.01 0.983
0.02 0.940
0.03 0.880
0.04 0.810
0.05 0.736
0.06 0.660
0.07 0.587
0.08 0.517 BINOMDIST(1,20,0.1,TRUE) – this is the
0.09 0.452 probability that the number of nonconforming
0.10 0.392 items in the sample is 0 or 1, when the lot
0.11 0.338 includes 10% nonconforming items.
0.12 0.289
0.13 0.246
0.14 0.208
0.15 0.176 12
Example: OC Curve (n=20, Ac=1)

1.0
0.9
0.8 For example, a batch with
0.7 10% nonconforming items
0.6 will be accepted about 40%
of the time.
Pa

0.5
0.4
0.3
0.2
0.1
0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40

p = Fraction Nonconforming in the Lot

13
Average Outgoing Quality
● For an acceptance sampling plan, the average outgoing
quality (AOQ) is the quality level that would result from
implementing the plan: 𝐴𝑂𝑄 = 𝑝 × 𝑃𝑎
– assumes that all rejected lots are 100% inspected
● The average outgoing quality limit (AOQL) is the maximum
value (worst-case) of the AOQ - obtained using AOQ Curve.

Inspect
Receive p n items yes Reject
Lot (let d=# d>Ac?
1-Pa Lot
(Size N) nonconfor
-mances)
no Pa
14
Accept Lot
Example (n=20, Ac=1)
p Pa AOQ
0.00 1.000 0.000
0.01 0.983 0.010
0.02 0.940 0.019
0.03 0.880 0.026
0.04 0.810 0.032
0.05 0.736 0.037
0.06 0.660 0.040
0.07 0.587 0.041
0.08 0.517 0.041
0.09 0.452 0.041 0.1*0.392 (For lots having 10%
0.10 0.392 0.039 nonconforming items, the acceptance
0.11 0.338 0.037 sample plan will accept items at a 3.9%
0.12 0.289 0.035 nonconformance rate.)
0.13 0.246 0.032
0.14 0.208 0.029
0.15 0.176 0.026
15
Example: AOQ Curve

0.05

0.04
AOQL  4.1%

0.03
AOQ

0.02

0.01

0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40

p
n=20, Ac=1 16
 0.020 0.010
0.018
0.016 n = 50 Ac = 1 0.008 n = 100 Ac = 1
0.014
0.012 AOQL ≅ 0.017 0.006 AOQL ≅ 0.0089
AOQ

AOQ
0.010
0.008 0.004
0.006
0.004 0.002
0.002
0.000 0.000
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25
p = Proportion Nonconformances in the Lot p = Proportion Nonconformances in the Lot

0.007 0.005

0.006
n = 150 Ac = 1 0.004 n = 200 Ac = 1
0.005

0.004 AOQL ≅ 0.0061 AOQ 0.003 AOQL ≅ 0.0047

AOQ

0.003 0.002
0.002
0.001
0.001

0.000 0.000
0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25

p = Proportion Nonconformances in the Lot p = Proportion Nonconformances in the Lot

Note the differences in AOQL as n increases

18
Attribute Sampling Systems

● An attribute sampling system consists of three components:

– several sampling schemes (e.g., single, double, multiple)
– a number of sampling plans (e.g., reduced, normal, tightened)
– rules for switching among plans
● Example: ANSI/ASQC Z1.4 (MIL-STD-105)
– called an AQL-based system since the key input is the
“Acceptable Quality Limit” (AQL), which is the target (worst
case) proportion of nonconformances.
– applicable to proportion and count data

19
MIL-STD-105: Procedure for Use

1. Specify the AQL, inspection level (loosely based on

inspection costs), and lot size

2. Find the sample size code (associated with the cost of

inspection)

3. Determine the type(s) of plan(s)

4. Find the normal single sampling plan from the

appropriate table

20
 Sample-Size Codes
Special Inspection Levels General Inspection Levels
Lot Size S-1 S-2 S-3 S-4 I II III
2-8 A A A A A A B
9-15 A A A A A B C
16-25 A A B B B C D
26-50 A B B C C D E
51-90 B B C C C E F
91-150 B B C D D F G
151-280 B C D E E G H
281-500 B C D E F H J
501-1200 C C E F G J K
1201-3200 C D E G H K L
3201-10,000 C D F G J L M
10,001-35,000 C D F H K M N
35,001-150,000 D E G J L N P
150,001-500,000 D E G J M P Q
500,001 or more D E H K N Q R

Inspection Level II: most common level (move to the left for more expensive inspections)
 Single Sampling (Normal) Plan

Ac = Acceptance Number If sample size exceeds lot size, do 100% inspection

Re = Rejection Number  = Use first plan in direction of arrow 22
Notes

1. Be sure to change both the sample size (n) and the

acceptance number (Ac) when following an up or down
arrow.
2. The rejection number is usually one more than the
acceptance number (except for multiple inspection plans).
3. The AQL levels are in rates per 100, for example:
– AQL=1.5 means 1.5 per 100, or 1.5%
– AQL=0.65 means 0.65 per 100, or 0.65%
– AQL=250 means 250 per 100 (use for count data, in this case
the average rate of occurrence is 2.5 per unit)

23
Find the acceptance sampling “normal” plan:

0.1% AQL, lot size of 100, inspection level I:

1.5% AQL, lot size of 700, inspection level S-4:

0.015% AQL, lot size of 20000, inspection level II:

0.25% AQL, lot size of 750, inspection level III:

24
 Proportion data, 0.40% AQL, lot size of 500, inspection level II
– normal plan: n=32, Ac=0.

0.012 AOQL=1.13%
0.010
In other words, our quality assurance system
0.008 seeks to ensure that incoming materials have a
nonconformance rate of 4000 PPM, yet the
AOQ

0.006
system ensures that incoming materials have a
0.004 nonconformance rate of about 11300 PPM.

0.002

0.000
0.00 0.05 0.10 0.15 0.20 0.25

This is a clear example of the phrase “you 25

cannot inspect quality into a product.”
Limitations of Acceptance Sampling

● Typically, acceptance sampling involves:

– infrequent sampling (usually after all production is complete)
– no accounting for order of production
– no displays
– little or no understanding of the manufacturing processes
● As such, this approach usually provides little or no guidance
in identifying problems.
– for example, it cannot distinguish between common cause
and special cause variation.
● Today, this approach is most often used for “incoming
inspection” of products from suppliers.
26
Deming’s k1,k2 Rule
● Deming determined that, for a process in statistical control
(i.e., stable), acceptance sampling was never optimal.
● Deming’s k1,k2 rule states that either zero or 100% inspection
should be done, based on three variables:
– let k1 be the cost per unit to inspect, k2 be the cost per unit of an
external failure, and p be the process’s nonconformance rate.
– perform a 100% inspection if: 𝑘1 < 𝑝
𝑘2
– otherwise, perform no (0%) inspection.
● Note that this rule assumes a perfect inspection system.
27
Examples

● If k1=$1 and k2=$10, and p=0.2, Deming’s rule would call

for 100% inspection, since:
𝑘1 1
= = 0.1 < 𝑝
𝑘2 10

● If k1=$5 and k2=$10, and p=0.2, Deming’s rule would call

for zero inspection, since:

𝑘1 5
= = 0.5 > 𝑝
𝑘2 10

28
Comparison of 0% and 100% Inspection

Type of Expected k1=1, k2=10, k1=5, k2=10,

Inspection Cost p=0.2 p=0.2

𝑝𝑘2
0% 2.50 2.50
1−𝑝

𝑘1
100% 1.25 6.25
1−𝑝
𝑘1 𝑘1
<𝑝 >𝑝
𝑘2 𝑘2

29
 k1,k2 Game: Concept

● A simulation exercise is used to illustrate Deming’s k1,k2 rule.

● For each inspection system option, ten items are “made” with
each item classified as conforming or nonconforming based
on the value from a random number generator.
● The application of 100% inspection, zero inspection, and
acceptance sampling inspection are explored:
– average cost of each system is calculated.
● Assumes that customer contract requires replacement of
nonconforming items free of charge.

30
k1,k2 Game: Setup

● Lot Size = 10
● p = 0.2
● k2 = $10.00
● k1 varies by supplier (red, blue)

31
 Expected Costs for Red Supplier
Acceptance
sampling plan k1 = 1 inspection cost
uses AQL=4% k2 = 10 failure cost
and inspection p= 0.2 probability of a nonconforming item
level II.
N= 10 Number of items in a lot
n= 3 Number of items in a sample from the lot
Ac = 0 Acceptance number (Ac)

Expected Costs for Blue Supplier

k1 = 5 inspection cost
k2 = 10 failure cost
p= 0.2 probability of a nonconforming item
N= 10 Number of items in a lot
n= 3 Number of items in a sample from the lot
Ac = 0 Acceptance number (Ac)
 k1,k2 Game: 100% Inspection
PERFORM THE FOLLOWING STEPS FOR EACH LOT

● Generate 10 random items and indicate “G” (conforming)

or “B” (nonconforming), where B occurs 20% of the time.
– record the number of nonconforming items.
– calculate the inspection cost for the 10 items.
● Generate as many additional random items as necessary to
obtain conforming replacement items for those
nonconforming items in the lot.
– record the total number of new items inspected.
– calculate the inspection cost for the replacement items.
● Calculate the total cost.
33
 Deming (k1, k2) Worksheet
100% Inspection

k1 = k2 =

Number of Number of
Inspection Inspection cost for
Lot nonconforming replacements Total Cost
Cost replacements
items in the lot inspected

1 $ $ $
2 $ $ $
3 $ $ $
4 $ $ $
5 $ $ $
6 $ $ $
7 $ $ $
8 $ $ $
9 $ $ $
10 $ $ $
Average total cost $
Average total cost per conforming item sold $
 k1,k2 Game: No Inspection
PERFORM THE FOLLOWING STEPS FOR EACH LOT

● Generate 10 random items and indicate “G” (conforming) or

“B” (nonconforming), where B occurs 20% of the time.
– record the number of nonconforming items.
– calculate the failure cost for the nonconforming items.
● Generate as many additional random numbers as necessary
to obtain good replacement items for those nonconforming
items in the lot.
– record the number of nonconforming items generated.
– calculate the failure cost for the nonconforming replacements.
● Calculate the total cost.
35
 Deming (k1, k2) Worksheet
No Inspection

k1 = k2 =

Number of Number of Failure cost of

Failure
Lot nonconforming nonconforming nonconforming Total Cost
Cost
items in the lot replacements replacements

1 $ $ $
2 $ $ $

3 $ $ $
4 $ $ $

5 $ $ $
6 $ $ $

7 $ $ $
8 $ $ $

9 $ $ $
10 $ $ $

Average total cost $

Average total cost per conforming item sold $
 k1,k2 Game: (3,0) Inspection
● Generate 10 random items and indicate “G” (conforming) or “B”
(nonconforming), where B occurs 20% of the time; and use a
random number generator to choose the 3 items to inspect.
– record the inspection cost for the three items in the sample.
– record the number of nonconforming items in the sample.
● If the sample includes no bad items, the lot is accepted.
– record # nonconforming items shipped and the failure cost.
– generate & record # replacements & the inspection cost.
● If the sample includes 1+ bad items, the lot is rejected.
– calculate the cost to inspect the remainder of the lot.
– generate & record the number of replacement items made.
– record the cost of inspecting the replacements.
37
 Deming (k1, k2) Worksheet
(3, 0) Acceptance Sampling Plan

k1 = k2 =

If Lot is Accepted If Lot is Rejected

Number of
Cost to Cost to
nonconforming Number of Number of Number of Number of
Lot Inspect the Cost to inspect inspect Cost to inspect
items in the nonconforming Failure Cost replacements nonconforming replacements Total Cost
sample replacements remainder of replacements
sample items in inspected items in inspected
lot
remainder of lot remainder of lot
1 $ $ $ $ $ $
2 $ $ $ $ $ $
3 $ $ $ $ $ $
4 $ $ $ $ $ $
5 $ $ $ $ $ $
6 $ $ $ $ $ $
7 $ $ $ $ $ $
8 $ $ $ $ $ $
9 $ $ $ $ $ $
10 $ $ $ $ $ $
Average total cost $
Average total cost per conforming item sold $
 Deming's k1k2 Game: Results
Red Supplier p = 0.2
k1 = 1 k2 = 10

0% Inspection 100% inspection Sampling Inspection

Avg Cost/Item Sold

Deming Expectation 2.5 1.25 n/a

Deming's k1k2 Game: Results

Blue Supplier p = 0.2
k1 = 5 k2 = 10

0% Inspection 100% inspection Sampling Inspection

Avg Cost/Item Sold

Deming Expectation 2.5 6.25 n/a

 k1,k2 Game: Simulation (Example Results)

Data:
k1 = 1.5 Inspection cost
k2 = 10 Failure cost
p= 0.2 Probability of a nonconforming item
N= 10 Number of items in a lot
n= 3 Acceptance sampling sample size
Ac = 0 Acceptance sampling acceptance number

Results: Avg Cost Exp Cost

𝑘1 1.5 100% Inspection $1.892 $1.875
= = 0.15 < 𝑝
𝑘2 10 No Inspection $2.720 $2.500
$2.039
 100% Inspection is optimal Acceptance Sampling

Green = Lowest Cost

Red = Highest Cost
 k1,k2 Game: Simulation (Example Results)

Data:
k1 = 2.5 Inspection cost
k2 = 10 Failure cost
p= 0.2 Probability of a nonconforming item
N= 10 Number of items in a lot
n= 3 Acceptance sampling sample size
Ac = 0 Acceptance sampling acceptance number

Results: Avg Cost Exp Cost

𝑘1 2.5 100% Inspection $3.085 $3.125
= = 0.25 > 𝑝
𝑘2 10 No Inspection $2.540 $2.500
Acceptance Sampling $2.705
 0% Inspection is optimal
Green = Lowest Cost
Red = Highest Cost
Notes

42
 Supplier Certification

● Supplier certification transfers the responsibility for quality

from the customer to the supplier.
– can give a firm’s production workers the autonomy to control
quality (operator certification).
– more likely it trusts a supplier to produce high quality without
significant firm oversight.
● In these systems, the quality department performs an audit
function rather than an inspection function.
– these systems tend to be customized & administrated internally
– their structure is often be based on quality certifications (e.g.,
ISO 9001).
43
Supplier’s Focus & Benefits
● A supplier would proactively:
– control their processes, and
– find root cause and take corrective action when necessary.

● If properly implemented, overall quality costs are reduced.

– increased direct labor costs are offset by reductions in
inspection costs and quality department overhead.
– removal of quality assurance activities by the customer
improves on time delivery rates.
● In return, when a supplier becomes certified, it will often
receive a price increase for its parts.

44
Supplier’s Responsibility
A certified supplier must complete a training program with
written and hands-on exams, followed by a “Probationary
Stage.” They are also evaluated in regards to:
– their ability to apply basic quality methodologies (e.g., the Six
Sigma body of knowledge);
– their technical understanding of the parts manufactured (e.g.,
their function, quality specifications, inspection methods, etc.);
and
– their performance during a trial where quality is managed jointly
(along with the customer) for a certain number of production
orders.

45
Supplier Certification: Example Framework

Probationary Stage

46
After the supplier is certified the system enters the “Audit
Stage” with each meeting organized as follows:
1. Preparation
2. Audit team selection
3. Checklist development
4. Opening meeting
5. Implementation
6. Analysis
7. Exit meeting
8. Reporting & corrective action
9. Follow-up
10. Closure
47
Audit Requirements

● The auditors should:

– be unbiased,
– include individuals experienced in the procedures evaluated,
– be firm in their assessments, but flexible to identify unusual
circumstances that affect how procedures are implemented,
and
– understand the “letter” and the “spirit” of each procedure.

● Although a flexibility is good, too much customization of

procedures can lead to confusion, misunderstandings, and
have detrimental effects on future audits.

48