Transcript 7-Run rule

Quality Management Tools
Apr-16
393SYS
1
1 Modern Quality Management

Modern quality management





Apr-16
requires customer satisfaction
prefers prevention to inspection
recognizes management responsibility for quality
Noteworthy quality experts include Deming, Juran,
Crosby, Ishikawa, Taguchi, and Feigenbaum
Quality; Who’s who
393SYS
2
Quality Experts






Deming was famous for his work in rebuilding Japan and his
14 points
Juran wrote the Quality Control Handbook and 10 steps to
quality improvement
Crosby wrote Quality is Free and suggested that organizations
strive for zero defects
Ishikawa developed the concept of quality circles and using
fishbone diagrams
Taguchi developed methods for optimizing the process of
engineering experimentation
Feigenbaum developed the concept of total quality control
Apr-16
393SYS
3
Fishbone or Ishikawa Diagram
Apr-16
393SYS
4
Malcolm Baldrige Award and
ISO 9000



The Malcolm Baldrige Quality Award was started in 1987 to
recognize companies with world-class quality
ISO 9000 provides minimum requirements for an organization to
meet their quality certification standards
HKMA Quality Award
http://www.hkma.org.hk/qa/award.htm
Apr-16
393SYS
5
2. Quality Planning



It is important to design in quality and communicate
important factors that directly contribute to meeting the
customer’s requirements
Design of experiments helps identify which variables
have the most influence on the overall outcome of a
process
Many scope aspects of IT projects affect quality like
functionality, features, system outputs, performance,
reliability, and maintainability
Apr-16
393SYS
6
Quality Assurance




Quality assurance includes all the activities related to
satisfying the relevant quality standards for a project
Another goal of quality assurance is continuous
quality improvement
Benchmarking can be used to generate ideas for
quality improvements
Quality audits help identify lessons learned that can
improve performance on current or future projects
Apr-16
393SYS
7
Quality Control

What it is:


The main outputs of quality control are




acceptance decisions
rework
process adjustments
Some tools and techniques include




Apr-16
***
pareto analysis
statistical sampling
quality control charts
testing
393SYS
8
3. Pareto Analysis



Apr-16
Pareto analysis involves identifying the vital few
contributors that account for the most quality
problems in a system
Also called the 80-20 rule, meaning that 80% of
problems are often due to 20% of the causes. It is
the fundamental postulates underlie the rational for
the Statistical SW Quality Assurance.
Pareto diagrams are histograms that help identify
and prioritize problem areas.
393SYS
9
80% of the contribution comes from
20% of the contributors:





Apr-16
80% of the engineering is consumed by 20% of the
requirements
80% of the software cost is consumed by 20% of the
components
80% of the errors are caused by 20% of the
components
80% of software scrap and rework is caused by 20%
of the errors
80% of the progress is made by 20% of the people.
…
393SYS
10
Sample Pareto Diagram
Apr-16
393SYS
11
Statistical SQA


Apr-16
SQA & Traceability as example on SQA & FTR notes
L14
p209, Pressman.
393SYS
12
Statistical Sampling and Standard Deviation



Statistical sampling involves choosing part of a population of interest
for inspection
The size of a sample depends on how representative you want the
sample to be
Sample size formula:
Sample size = .25 X (certainty Factor/acceptable error)2
Apr-16
393SYS
13
Commonly Used Certainty Factors
Desired Certainty
Certainty Factor
95%
1.960
90%
1.645
80%
1.281
95% certainty: Sample size = 0.25 X (1.960/.05)2 = 384
90% certainty: Sample size = 0.25 X (1.645/.10)2= 68
80% certainty: Sample size = 0.25 X (1.281/.20)2 = 10
Apr-16
393SYS
14
Standard Deviation



Standard deviation measures how much variation exists
in a distribution of data
A small standard deviation means that data cluster
closely around the middle of a distribution and there is
little variability among the data
A normal distribution is a bell-shaped curve that is
symmetrical about the mean or average value of a
population
Apr-16
393SYS
15
Normal Distribution and Standard
Deviation
Apr-16
393SYS
16
Sigma and Defective Units
Specification Range Percent of
Population
(in +/- Sigmas)
Defective Units
Per Billion
Within Range
1
68.27
317,300,000
2
95.45
45,400,000
3
99.73
2,700,000
4
99.9937
63,000
5
99.999943
57
6
99.9999998
2
Apr-16
393SYS
17
Quality Control Charts and the Seven
Run Rule


Apr-16
A control chart is a graphic display of data that illustrates
the results of a process over time. It helps prevent
defects and allows you to determine whether a process is
in control or out of control
The seven run rule states that if seven data points in a
row are all below the mean, above the mean, or
increasing or decreasing, then the process needs to be
examined for non-random problems
393SYS
18
Control Chart of 12” ruler
Apr-16
393SYS
19
Control Chart



Apr-16
contiu.
The output of a production process will fluctuate. The causes of
fluctuation can just be random or non-random due to
desirable/undesirable process change. Control charts graph and
measure process data against control limits. Control charts can
distinguish the random variation from assignable causes or nonrandom causes.
We cannot adjust random variation out of a process. Process
adjustments for random variation are neither necessary nor desirable.
This is over-adjustment or tempering, and it makes the process worse.
We can and must investigate assignable causes (or non-random
causes). Points outside the control limits are evidence of process
problems. Analyst must investigate every out of control point for an
assignable cause. They must record their findings and any corrective
actions. For example, a tool adjustment, or change in Formal Technical
Review format or worn tooling, may correct the problem.
393SYS
20
Pattern analyzing of Control Chart
7-Run rule
7-run-rule is used to filter out the random variation in a production
process. shows the ‘trends’ that are caused by the ‘assignable
causes’ or non-random causes that required investigation and
possible corrective action to be taken.
7-run-rule pattern:

seven points above mean value;

seven points below mean value;

seven points or all increasing ; or

seven points all decreasing
the patterns are indicators of non-random problems which can be
symptom of process out of control.
Apr-16
393SYS
21
To develop a Control Chart to
determine project stability





Apr-16
Plot individual metric values on a chart.
Compute the mean value for the metrics value and plot
the line.
Plot the Upper Control Limit and Lower Control Limit.
Compute a standard deviation as (Upper-control-limit mean)/3. Plot lines one and two standard deviation
above and below Am. If any of the standard deviation
lines is less than 0.0, it need not be plotted unless the
metric being evaluated takes on values that are less than
0.0.
The Std Dev.# is then plotted on the control chart.
393SYS
22
Statistical Testing
Statistical testing is a system testing process in which
the objective is to measure the reliability of a system
rather than to discover faults. Statistical testing can be
combined with reliability growth modeling. Predictions
of the final system reliability and when that will be
achieved can be made. As failures are discovered, the
underlying faults causing these failures are repaired so
that the reliability of the system can improve in the
course of the testing process
A lot of airlines manage dispatch reliability with
statistical testing approach.
Apr-16
393SYS
23
The Cost of Quality

The cost of quality is


Apr-16
the cost of conformance or delivering products that
meet requirements and fitness for use
the cost of nonconformance or taking responsibility
for failures or not meeting quality expectations
393SYS
24
Total Quality Cost


Price of nonconformance PONC
Price of Conformance (POC)




Apr-16
peer walkthroughs, inspections
development & impl. QMS, standards, training
setting up & running of quality program
Cost of Quality (COQ) = POC + PONC
393SYS
25
Costs Per Hour of Downtime Caused by
Software Defects
Business
Cost per Hour Downtime
Automated teller machines (medium-sized bank)
$14,500
Package shipping service
$28,250
Telephone ticket sales
$69,000
Catalog sales center
$90,000
Airline reservation center (small airline)
$89,500
Apr-16
393SYS
26
Five Cost Categories Related to Quality





Prevention cost: the cost of planning and executing a project so it is
error-free or within an acceptable error range
Appraisal cost: the cost of evaluating processes and their outputs to
ensure quality
Internal failure cost: cost incurred to correct an identified defect
before the customer receives the product
External failure cost: cost that relates to all errors not detected and
corrected before delivery to the customer
Measurement and test equipment costs: capital cost of equipment
used to perform prevention and appraisal activities
Apr-16
393SYS
27
Organization Influences, Workplace
Factors, and Quality




Apr-16
Study by DeMarco and Lister showed that organizational issues
had a much greater influence on programmer productivity than
the technical environment or programming languages
Programmer productivity varied by a factor of one to ten across
organizations, but only by 21% within the same organization
Study found no correlation between productivity and
programming language, years of experience, or salary
A dedicated workspace and a quiet work environment were key
factors to improving programmer productivity
393SYS
28