10/3/2007
INL 660
leaders and other decision makers are seeking cost effective ways in carrying on software
testing. In that direction, many have advanced test automation as the most cost effective
way of carrying on software testing as opposed to manual testing. Furthermore, test tools
vendors seizes such moment to advertise their tools as solutions to reducing testing cost.
In a paper presented during the Economics of test automation: test cost and
manually tested and the trade-off between the two. The authors observe the need for cost
software testing by growing international competition, shrinking budgets, and the fast
paced development process. The paper critically examines the cost models for automated
software testing used by proponents of automated test to argue in favor of test automation
cost. The drive to come up with an alternative model of test automation is based on the
authors’ own experience in supplying consulting services and support and report from
categorically not the solution to reduce testing cost, but its aim is to “stimulate
approaches.” By outlining factors based on the alternative model the authors hope to
Ramler et al begin their article by emphasizing the cost factor involved in quality
assurance quoting studies that software testing accounts for more than 50% of total
budget spent on projects. Due to this high cost, the need for finding more cost effective
ways which automated testing proposes to solve cannot be over simplified. While testing
tools vendors and the cost model for software testing supports this preposition, the
authors counter such argument first by analyzing the model calling it “overly simplistic”
and providing an alternative one. The authors assert that the cost model for software
testing and the “universal formula” which support automated testing over manual testing
for cost effectiveness are flawed because they take into account only costs and leave out
benefits. They argue “cost and benefits are both required for an accurate analysis.”
Benefits in testing can be measured in terms of the defects they address or risk mitigation.
Other reasons given for the model’s flaws include the incomparability of automated and
manual testing since the two are really two different processes rather two ways of doing
the same thing. They further criticize the model because additional cost factors such as
the cost of abandoning automated test after changes in the functionality are not taken into
account, etc. its inability to consider entire project context, and that the model
mistakenly considers all test cases equally important. To counter the “overly simplistic”
model, the authors propose an alternative model based on opportunity cost. Borrowing
the concept of opportunity cost from Economics, the authors maintain that the decision as
to which testing alternative to adopt is a trade off between using automated testing or
manual testing. They likened economists’ production possibility frontier to the restricted
budget of testing to show the opportunity cost of test automation as measured in terms of
manual test execution. They argue that management should decide as to which testing
method will yield the maximum benefits. There is always going to be some automation
forgone for every manual test, “once the efficient points on the frontier have been
reached; the only way of getting more automated test cases is to reduce manual testing.”
They are convinced that the opportunity cost model includes both costs and benefits
which makes the analysis more rational. It approaches high cost of testing debacle not
from the standpoint of minimizing the cost of testing but maximizing its benefits. They
suggest that decision makers faced with the problems of either to manually test or
automate must build on the potential benefit of a test case so as to allow a realistic
automation in terms of making decision for a cost cutting alternative. First, as supported
by “Reconciling Manual and Automated Testing: the Auto Test Experience” (2007),
Leitner et al agree that both manual and automated testing strategies are distinct but
complementary since one may capture cases that the other might not find. So the claim
that automated testing is more cost effective and therefore should replace manual testing
to save cost is a wild one. The authors’ charge that the argument in favor of automated
testing leaves out benefits is in place. Managers and IT leaders need not decide to choose
one way of testing over the other because of cost alone rather the benefits of carrying on
one test over the other. Consideration on optimal benefits should be the driver of
choosing one test over the other. The opportunity cost models captures most if not all of
the factors ignored by the cost model. With both cost and benefits analyzed, managers
The paper has achieved the objective it set out to achieve, that is to address the
issue of when a test should be automated and the trade off between the two methods.
While the authors maybe convincing, the paper fell short of presenting real world data to
support their proposed alternative. As the authors have mentioned that the paper is only
aiming to stimulate discussion in a different direction, one issue that is left is that the
solution has not been tested in the real world. To be convincing, I would love for the
authors to compare two companies using both approaches and analyze the cost and
benefits.
Like the cost model, the opportunity cost model does not take into account all the
influencing factors. The Authors agreed that incorporating additional factors to the
opportunity cost model for completeness will add further complexities and thereby
reduce its practicality. For further studies, the authors plan to “include further influencing