Overview of Testing - Software Testing: Testing Across the Entire Software Development Life Cycle - page 19

Testing principle 5: Make the highest risks the fi rst testing priority.
When faced with limited testing staff, limited testing tools, and limited time to
complete the testing (as most testing projects are), it is important to ensure that there
are suffi cient testing resources to address at least the top business risks. When test-
ing resources cannot cover the top business risks, proceeding with testing anyway
will give the system stakeholders the false expectation that the company will not be
torpedoed and sunk by software defects.
Testing principle 6: Make the most frequent business activities ( the 80/20 rule ) the
second testing priority.
Once you have the real business killers well within your testing sights, consider
the second priority to be the most frequent business activities. It is common industry
knowledge that 80% of any daily business activity is provided by 20% of the busi-
ness system functions, transactions, or workfl ow. This is known as the 80/20 rule .
So concentrate the testing on the 20% that really drives the business. Because the
scarcity of testing resources continues to be a concern, this approach provides the
most testing “bang for the buck.” The other 80% of the business system typically
represents the exception transactions that are invoked only when the most active
20% cannot solve a problem. An exception to this approach is a business activity
that occurs very seldom, but its testing importance is way beyond its indication by
frequency of use. The classic example of a sleeper business activity is a year-end
closing for a fi nancial system.
Testing principle 7: Statistical analyses of defect arrival patterns and other defect
characteristics are a very effective way to forecast testing completion.
To date, no one has reported the exhaustive testing of every aspect of any rea-
sonably complex business software system. So how does a tester know when the test-
ing is complete? A group of noted statisticians observed a striking parallel between
the defect arrival, or discovery patterns in software under development, and a family
of statistical models called the Weibull distribution. The good news is that the intel-
ligent use of these statistical models enables the tester to predict within 10%-20%
the total number of defects that should be discovered in a software implementation.
These models and their ability to predict human behavior (software development)
have been around for at least 20 years. The bad news is that we have not found any
signifi cantly better ways to develop software during the same 20 years, even though
programming languages have gone through multiple new and powerful paradigms.
Chapter 12 takes a closer look at these models and how they can assist the tester.
Testing principle 8: Test the system the way customers will use it.
This principle seems so intuitive; however, the authors see examples of software
every year that simply were not tested from the customer's perspective. The fol-
lowing is a case in point. A major retail chain of toy stores implemented a Web site
on the public Internet. Dr. Everett attempted to buy four toys for his grandchildren
on this toy store Internet Web site with catalog numbers in hand. Finding the toys
to purchase was very diffi cult and took over 45min to achieve. When he fi nally
found all four toys and placed them in his shopping cart, Dr. Everett was unable to