Dear Sir or Madam,
You took part at the ESMIT survey ("Evaluation of Software Metrics in the Design Phase and their Implication on CASE tools") and requested feedback. The survey was conducted from March 7th until May 15th. During this timeframe a total number of 142 persons responded. Before the final survey results we present to you some demographically data of all respondents:
42% developers, 21% students, 20% project managers, 5% designers, 4% testers, 2% researchers and 6% job description not applicable
Company size (employees):
38% >300, 25% <10, 19% 10-50, 10% 101-300, 8% 51-100
Experiences in software engineering (years):
32% 1-3, 23% >10, 19% 3-5, 14% 5-10, 12% <1
34% work in a company that is certified (69% ISO, 18% CMM, 13% others), 66% work in a company that is not certified
Most used programming language:
36% Java, 20% C++, 9% C, 9% Delphi, 7% ASP, 7% VB, 4% Perl, 2% PHP, 13% other
These data should serve as a base for comparing your own demographically data.
The most import result from the survey is that
- only 21% of all respondents collect metrics in the design phase.
- 47% collect software metrics in other development phases than design.
This result is influenced by the fact that except for the process metric COCOMO and the traditional product metric Lines Of Code (LOC) all metrics such as Halstead's Complexity, Cyclomatic Complexity, object-oriented metrics (Depth of Inheritance Tree, Weighted Method per Class, Coupling between Objects, Number of Children, Lack of Cohesion, Response for a Class) are unknown to a high degree.
Furthermore the survey reveals that CASE tools are rarely used. Asking for the application of design and metrics tools, we found out that 37% using design tools and only 18% metrics tools.
The most frequent design tools are Rational Rose, Visio, Togethersoft Control Center and Microsoft Visual Studio. The most frequently used metrics tools are self-made/in-house solutions, followed by Microsoft Excel, Function Point Workbench, Together Control Center, and Eclipse.
The survey results show that on the one hand
- 83% of the respondents agree that software metrics are useful for control.
- On the other hand 55% believe that software metrics are difficult to collect and
- 56% think that software metrics are difficult to analyse.
- The vast majority of 80% disagree with the statement that software metrics are not necessary.
This indicates that software engineers have a belief in software metrics, but also are aware that the application of software metrics comes along with difficulties. It can be positively seen that people generally think that software metrics are necessary.
The conclusion of this work is that the use of software metrics has been proved by science and even industry, although only a minority is applying software metrics successfully. A reluctance to software metrics exists among software engineers which is obviously due to a lack of education. The infrastructure for applying software metrics such as the Goal-Question-Metric model already exists and can be easily applied in a metrics plan. The understanding for software metrics has to be communicated to software engineers since it is them who are responsible for the quality of the final product.