Statistical Quality Control in Testing Laboratory

Quality is a discipline that focuses on product and service excellence. Quality is achieved
through customers’ perception; therefore, organizations must understand customer
needs and expectations in order to meet and exceed them. Customer needs and
expectations can be achieved through quality improvement. Quality is important to the
consumers.

Quality has been defined from several perspectives. For example, quality may have a
different meaning to the engineer who designs the product, or to the manufacturer
involved in the production of a product, or the laboratory who provide the testing
service to the customer. Although we define quality from many different perspectives,
the final judge of the product or service quality is the customer, and therefore, quality is
the customer’s perception of the degree to which the product or service meets his or her
expectations.

To improve quality, it is important to have an understanding of systems and processes.

A system converts inputs into useful products or services through a conversion process.
The products or services are the output of the system. Quality is concerned with the
variation in the output of a system or the products and the processes of the system. The
quality of the products and services are improved by reducing the defects and variation.
Quality is inversely proportional to the variation in the products and processes. This
means that as the variation is reduced and controlled, significant improvement in
quality can be achieved. The improved quality and reliability in products and services
lead to higher perceived value and increased market share, thereby, increasing revenue
and profitability.

Definition of Quality

Customer-driven quality approach

Quality means different things to different people. Therefore, quality can be defined
from several different perspectives. From the perspective of the customer or the end
user of the product or service, the quality of a product or service is the customer’s
perception of the degree to which the product or service meets his or her expectations.
This also means that the quality of a product or service can be determined by the extent
to which the product or service satisfies the needs and requirements of the customers.
This definition is a customer-driven quality approach that aims at meeting or exceeding
customer expectations.

Manufacturer-driven quality approach

Quality has also been defined from several other perspectives. For example, quality may
have a different meaning to the engineer who designs the product, or to the
manufacturer involved in the production of a product. Thus quality can be defined from
the perspective of the manufacturer or the designer.

Following are some of the other ways quality has been defined:

• Transcendent: Quality is something that is intuitively understood but nearly

impossible to communicate, such as beauty or love.
• Product-based: Quality is found in the components and attributes of a product.
• User-based: If the product or service meets or exceeds customer’s expectations, it has
good quality.
• Manufacturing-based: If the product conforms to design specifications, it has good
quality.
• Value-based: If the product is perceived as providing good value for the price, it has
good quality.

Quality has also been defined as:

• Meeting or exceeding customer expectation

• Fitness for intended use
• Conformance to specifications
• Inversely proportional to variation
• Total customer service and satisfaction
• The degree or standard of excellence of something

Dimensions of quality

The dimensions of quality specify the characteristics the product or service should
possess in order to be high quality. Garvin has identified eight dimensions of quality
described here. These dimensions describe the product quality that is critical to
developing high quality products or services. The recognition of these dimensions by
the management and the selection of these dimensions along which the business will
compete is critical to business success.

1. Performance: Will the product do the job?

2. Features or added features: Does it have features beyond the basic performance
characteristics?

3. Reliability: Is it reliable? Will it last a long time?

4. Conformance: Does the product conform to the specifications? Is the product made
exactly as the design specified?

5. Serviceability: Can it be fixed easily and cost effectively?

6. Durability: Can the product tolerate stress without failure?

7. Aesthetics: Does it have sensory characteristics such as taste, feel, sound, look, and
smell?

8. Perceived quality: What is the customer’s opinion about the product or service? How
customers perceive the quality of the product or service?

Importance of quality

There is a close relationship between quality, profitability, and market share. Quality is
achieved through customer’s perception, therefore, organizations must understand
customer needs and expectations to meet and exceed them. Customer needs and
expectations can be achieved through quality improvement. Quality is important to the
consumers. In today’s highly competitive and global economy, a company cannot
survive and stay in business unless they are able to provide high quality products and
services. Figure 1.1 shows how improving quality can help organizations increase their
market share and increase profitability.

Quality is also important because the quality—both good and bad—costs money. There
is a cost involved with improving the quality of products and services, because poor
quality can significantly affect an organization’s competitiveness and market share.
Quality costs or the costs of quality (COQ) as having two components:

(1) costs of good quality (or the cost of conformance) and

(2) costs of poor quality (or the cost of non-conformance).

The focus of many quality programs is to reduce the cost of poor quality. Since the cost
of poor quality is significant, reducing this cost will lead to increased revenue and
improved productivity. A quality program should be focused on preventing poor quality.
A prevention system is focused on preventing the poor quality and is far superior to a
detection system that detects the defects and nonconformities in the products after they
are produced.

The major components of costs of good quality—prevention costs and appraisal costs,
and the costs of poor quality—internal failure and external failure costs, are explained
in Table 1.

Table 1 Quality costs

Prevention cost Appraisal costs

 Quality planning  Inspection and testing of incoming
 Product and process design material
 Process control  Inspection and testing of products
 New product review  Staffing inspectors and supervisors
 Manufacturing engineering tasks  Maintaining the accuracy of test
equipment
  Quality training  Maintaining test and inspection
 Vendor relations records
 Variability analyses  Performing audits and field tests
 Design reviews and
process/manufacturing planning
 Designing equipment and processes
to measure and control quality
Internal failure cost External failure cost
 Scrap  Returned products or rejected
 Rework services that must be inspected,
 Retest reworked or scrapped
 Repair  Customer complains
 Failure analysis  Cost of retesting, legal services,
 Down time settlements
 Loss in profit due to substandard  Other costs related to product
product or services liability
 Customer dissatisfaction

Systems and process

The quality methods and tools are applied to the systems and the processes that make
the systems. The system and the processes within the system are responsible for
creating the products or services. Therefore, it is important to understand the systems
and the processes.

A system usually consists of a group of interacting, interrelated, or interdependent

processes forming a complex whole. Thus a system is a collection of processes with a
specific mission or purpose. A process can be viewed as a part of a system. Refer Fig.1
for system and process, which known as SIPOC (Supplier, Input, Process, Output,
Customer) and Fig. 2 for an “Input-Output Process”:

Fig.1
Fig. 2

Statistical Quality control (SQC)

Statistical process control (SQC) is a method of quality control in which statistical

methods are employed. SQC is applied in order to monitor and control a process and its
output. Monitoring and controlling the process and its output ensures that it operates at
its full potential.

Statistical quality control is important because it uses statistical methods to monitor the
quality of a product/process output. This type of auditing maximizes manufacturing
productivity and minimizes errors associated with human judgement.

There are three categories in statistical quality control, and each of these categories is
effectively used in product quality evaluation.

Descriptive Statistics - In this SQC category, quality characteristics and relationships

are described. In order to achieve these, it is important to find out the mean, range,
standard deviation, and measure of the distribution of data of the sample. This we use in
analysis to evaluate the process performance (human or instruments).

Statistical Process Control - This category verifies if a process is functioning properly

or not. A random sample of the products is taken, inspected, and decided if these
manifest characteristics that fall into a predetermined range. The predetermined range
determines if the process is functioning or not. In laboratory we do it as “repeat
analysis” by same chemist and by multiple chemists.

Acceptance Sampling - Acceptance sampling determines if a whole lot of products is

acceptable based on the results of random inspection. In other words, acceptance
sampling determines if a product batch is acceptable or not. If the products are not
acceptable, then the whole batch is rejected. This we do in analysis in terms of CCV
during a batch analysis or as QC sample for conventional analysis.

Now we have understood what quality is and why to do quality control, we will focus to
use statistics for quality control. Please remember, we do not have to be a statistician,
but we have to know the basic tools which are needed to use to for Quality Control in
testing laboratory, and we will go step by step with examples considering our regular
work in laboratory.

Mean, Median and Mode

A Chloride sample of 20 mg/L analysed 9 times. Data obtained as follows:

A B C
Run No Value mg/l (A - mean) B2
1 18 -0.44 0.20
2 17 -1.44 2.09
3 21 2.56 6.53
4 17 -1.44 2.09
5 19 0.56 0.31
6 17 -1.44 2.09
7 21 2.56 6.53
8 19 0.56 0.31
9 17 -1.44 2.09
Total 22.22
Mean 18.44
Std. Dev. 1.667

If we arrange the data from lowest to highest then it looks like:

17, 17, 17, 17, 17, 18, 19, 19, 21, 21

Here the “18” is the middle one and both of this side are 4 readings. So, the median is 18.

We will discuss later with example, if the midpoint have two values.

————— 13/11/2019 12:14:18 PM ————————————————————

Welcome to Minitab, press F1 for help.

Sum of Value mg/l

Sum of Value mg/l = 166

Mean of Value mg/l

Mean of Value mg/l = 18.4444

Median of Value mg/l

Median of Value mg/l = 18

Standard Deviation of Value mg/l

Standard deviation of Value mg/l = 1.66667

Histogram of Value mg/l

Normal
Mean 18.44
4
StDev 1.667
N 9

3
Frequency

0
15 16 17 18 19 20 21 22
Value mg/l

From this histogram, we can see that for 20 mg/L known sample, we are getting results from 17 to
21 mg/L. Though the average is 18.44 mg/L, but chances to get 17 mg/L is more. We can conclude
that we have to do the RCA to find out the reason of this deviation from actual value.

So, mean tells us the average of results we are getting and how far it is from the actual value
(accuracy). But it does not tell everything. Median tells us the midpoint of different values, and it
also does not reflect the actual variation. Mode is telling the frequency of getting a specific value.
This all comes together in the histogram to give a graphical representation.
 A B C
Value
Run No mg/l (A - mean) B2
1 19 0.56 0.31
2 16 -2.44 5.98
3 23 4.56 20.75
4 15 -3.44 11.86
5 21 2.56 6.53
6 15 -3.44 11.86
7 22 3.56 12.64
8 18 -0.44 0.20
9 17 -1.44 2.09
Total 72.22
Mean 18.44
Variance = Total of C/(run no – 1)
Std. Std. dev. = Sq. Rt of Variance
Dev. 3.005

Now here you can see the mean is same as earlier data, but std. dev. is more. When you
see this data, you can conclude that as the mean is same, so both are as per. But when
we will see the histogram, we will get the difference. See below histogram:

Histogram of Value mg/l

Normal
Mean 18.44
2.0
StDev 3.005
N 9

1.5
Frequency

1.0

0.5

0.0
12 14 16 18 20 22 24
Value mg/l

Now you can see that the process is not in control and worse than earlier one.
This is the numerical methods of describing data. This can be divided into following
categories: (1) measures of central tendency or measures of location, (2) measures of
position, (3) measures of variation or dispersion, and (4) the measures of shape.

The important measures of central tendency are the mean and median. The mean or the
average is a statistical constant which enable us to comprehend the significance of the
whole. It provides an idea about the concentration of the data in the central part of the
distribution. The median is the value of the variable that divides the data into two equal
parts, such that half of the value lies above the median and the other half below it.

Calculating Median When the Number of Observations Is Odd If the number of

observations is odd, the median can be calculated by

• arranging the data in increasing order and

• locating the middle value after the values have been arranged in ascending order of
magnitude.

Suppose we have the following observations arranged in increasing order:

The number of observations is 7 (n = 7), which is odd; therefore, the middle value or the
median is the fourth value which is 9.6.

Suppose the following data are the annual incomes of eight employees (in thousands of
dollars) of a manufacturing company for the past year. Find the median.

The number of observations is: n = 8 (even), then, arrange the data in increasing order:
The location of the median for even observations is given by

(n + 1)/2 = (8 + 1)/2 = 4.5

Therefore, the median is the average of 4th and 5th values = (45 + 56) / 2 = 50.5

The measures of central tendency provide an idea about the concentration of the
observations about the central part of distribution. The measures of central tendency
(mean and median) are not sufficient to give us a complete description of the data. They
must be supported by other measures. These measures are the measures of variation or
measures of dispersion. They tell us about the variation or dispersion of the data values
around the average. Refer again the two histogram explained above.

If we measure only the mean of different data, we miss other important characteristics.
Mean, median, and mode tell us only part of what we need to know about the
characteristic of the data. In order to better analyse and understand the data, we must
also measure its dispersion; that is, its spread or variability.

Dispersion or the variation is often used to compare two or more sets of data. In
statistical quality control, one of the major objectives is to measure and reduce
variability. This is done by extracting data from the process at different stages or time
intervals, and analysing the data by different means in order to measure and reduce
variation.

Quality is inversely proportional to the variation. One way of improving quality is

reducing variation.

The variability in the data is measured using the following measures. These are known
as the measures of variation.

1. Range
2. Variance
3. Standard Deviation
4. Coefficient of Variation
5. Interquartile Range (IQR)

Out of these, the variance and the standard deviation are the most widely used. You
have already learnt how to calculate the standard deviation.

There are two important rules in statistics that relate the mean and standard deviation:
(1) the Chebyshev’s Rule and (2) Empirical Rule. These rules combine the mean and the
standard deviation in a data set and provide specific conclusions. The Empirical Rule
provides more definite results and applies to symmetrical or bell-shaped data. The
histogram, stem-and-leaf, and box plot discussed earlier provide information about the
shape (symmetry or lack in symmetry) in the data. The empirical rule applies to
symmetrical or bell-shaped distribution. This is also known as the normal distribution.
Unlike Chebyshev’s theorem that applies to any shape (skewed or symmetrical), the
empirical rule applies to symmetrical shape only. This rule states that if the data are
symmetrical or bell shaped:
 Approximately 68 percent of the observations will lie between the mean and ± 1
standard deviation.
 Approximately 95 percent of the observations will lie between the mean and ± 2
standard deviations.
 Approximately 99.7 percent of the observations will lie between the mean and ±
3 standard deviations.

For a population data, the mean and standard deviation are denoted by μ (read as mu)
and σ (sigma). Using these symbols, the earlier empirical rule can be stated as:

 μ ±1- will contain approximately 68 percent of the observations

 μ ± 2- will contain approximately 95 percent of the observations
 μ ± 3- will contain approximately 99.7 percent of the observations

Note: for a sample data, the mean and standard deviation are denoted by x and s.

For clearer understanding the data, now we will focus on percentile and quartile for
data set.