Book Review: The Economics of Software Quality 83
First time accepted submitter BenLinders writes "The Economics of Software Quality provides solutions to quantify software quality, helping you to manage software development and maintenance. It contains software quality data that you can use to build a business case to improve the quality of your software, and decide upon processes and techniques that can help to implement the needed improvements in your organization." Read below for the rest of Ben's review.
Quantifying software quality is not an easy thing. Several measurements exist, for instance estimating and tracking the number of defects that are found (both within development/maintenance and from customers), measuring software quality with static analysis tools (complexity, fan in/fan out), or measuring the effectiveness of software development methods and techniques (like inspections, test, and Cost of Poor Quality). This book covers software quality factors that influence the quality of software products as perceived (and believed!) by customers. An extensive list of factors is provided, where the authors have selected those factors that they consider most significant to achieve quality.The Economics of Software Quality | |
author | Capers Jones and Olivier Bonsignour |
pages | 587 |
publisher | Addison-Wesley |
rating | 8/10 |
reviewer | Ben Linders |
ISBN | 978-0-13-258220-9 |
summary | To build your Business Case for Software Quality Improvement |
Many software development processes and techniques are covered in this book, from a quality and economic point of view. This also includes agile methods, where a body of data is available about the effects of agile techniques like user stories, Test Driven Design, Scrum Sessions, Measuring Technical Debt, and Pair Programming. For instance, about agile user stories the book states "... the user story method seems to be concise and fairly trouble-free, User stories average below 0.5 pages per function point and seem to contain fewer than 0.5 defects per function point`. This kind of information can be very helpful to build a business case for using agile methods in your organization.
Most of the data on software quality that the book provides is in "Defects per Function Point". A backfiring table is also provided, to translate language statements/lines to function points. So if you are not using function point, but programming in Java, Ruby, C++ or any other popular programming language, the data can still be used.
There is a full chapter covering defect prevention. Methods like Reuse, Formal Inspections and Quality Function Deployment are the most effective in preventing defects, and also techniques like Root Cause Analysis and PSP/TSP are claimed to be very effective. Given that the top ten techniques reduce defects with 40% — 85%, makes it interesting for many organizations to investigate the business case to improve the quality of their products, using these methods and techniques.
Additional information is provided on how to measure the effects on quality from a given method or technique. The book also provides warning for quality measurements that can be unreliable. An example is measuring cost-per-defect. When the quality of your development activities increases, for instance by improving requirements practices and implementing defect prevention for design and coding, the number of defects that testing finds will go down. Since test case preparation is a fixed cost, the cost per defect for testing will go up when the software has fewer defects. This makes such a measurement potentially unreliable. I believe that the main benefits will come when you can reduce your testing activities, based upon measurements that quantify the quality of your products before testing starts. Techniques like risk based testing can also reduce your testing hours, thus saving time and money on tests that are not needed.
Defects measurements and tracking are used in more then 55% of the military and defense software applications (using CMMI, TSP, QFD, etc), but in less then 15% of IT, commercial, web or embedded applications. Given their prevention effectiveness of -35%, and removal effectiveness of 25%, it is still surprising to me that this is not used more often. The data needed for these kinds of measurements is usually available in the defect management systems, though some addition effort is needed to classify defects and to do Root Cause Analysis. The benefits of using these kinds of measurements, combined with estimations of the expected quality at release, to decide and steer software development and prevent defects during the development and before release are significant.
The book also gets into methods to quantify structural quality issues that are not exactly "defects" but have an important impact – "Technical Debt" being one of these methods of quantification. These kind of measurements help to manage the quality of your code base, being able to see the impact on quality from changes, and take action to get quality back on the desired level when needed.
Reviews and inspections are very effective ways to remove defects before testing. Several techniques are described, both informal and formal techniques. Several of them are also usable within agile methods, supporting teams in developing better quality software. Applying these techniques effectively requires training, and arrangements within your company that enable employees to use them. The book makes clear that if you want to reduce post release defects and lower your maintenance costs, the work needs to start with early software development activities, like using better techniques for managing requirements, software modeling and design, reviews and inspections, and automatic code analysis. Testing alone is not sufficient to improve quality, and is also very costly.
The relationship between quality and risks is also explored. Many major software problems are related to the quality of the software products, e.g. outages, data loss, security issues or regulatory non compliances. Investigating such issues, for instance with Audit or Root Cause Analyses, and taking action to prevent similar problems in the future can be essential for your business. Measuring the losses and estimating potential benefits from preventive actions helps you to select the right improvements, and acquire commitment and funding to implement them.
The capabilities and skills of the staff that develops the software have significant impact on the quality. The benefits of training, skill development, and sharing of experiences to develop a learning organization can be huge. Software methods like Agile and RUP include mechanisms to continuously evaluate, learn and improve the capabilities of your staff. E.g. using retrospectives and scrum boards, to identify and follow up with improvement actions.
Overall the book covers the economic perspective of quality. The information provided can be overwhelming for some readers. If you need to improve your product quality, and are limited in time and money to do it, this book helps you to select effective quality methods and techniques, and to measure and track your progress when implementing improvements.
Ben Linders is a specialist in quality, process improvement and organizational development.
You can purchase The Economics of Software Quality from amazon.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.
Agile programming is a lie (Score:4, Informative)
There, I said it.
"This programming philosophy will allow you to develop high quality software really quickly, and on the cheap" is the equivalent of a politician promising to fix every problem in the country with no sacrifices required, and put chocolate milk in all the water fountains to boot.
It's always the old thing: fast, cheap, or quick--pick any two.
Re:Agile programming is a lie (Score:4, Informative)
Aren't fast and quick the same thing?
don't forget the organization itself (Score:5, Informative)
While it's popular to focus on code metrics (defect count, test-based metrics, etc.), when it comes to how to improve software quality, don't forget that it's strongly related to organizational characteristics. Whether you look at it via Conway's Law [wikipedia.org], via Fred Brooks's analysis [amazon.com], or via recent empirical research [pdf] [umd.edu], it's pretty clear that software developed in an organization isn't independent of the organization itself, and sometimes the way the company is structured/managed is the problem.
Peopel dont buy quality, they buy Brand (Score:2, Informative)
"If it compiles, ship it. If it sells, patch it."
"Nobody ever got fired for buying IBM."
THAT is the software sales model. That, and having people who know braindead CTOs who buy from their golf buddies, creating a giant naked Emperor.
Re:Agile programming is a lie (Score:3, Informative)
Should be:
Fast, Cheap, or Good - pick two.
Re:Actually sounds interesting... (Score:3, Informative)
Re:Agile programming is a lie (Score:4, Informative)
Should be:
Fast, Cheap, or Good - pick two.
And expect at most one.
Bullshit. Agile programming is a manifesto. (Score:2, Informative)
The Agile Manifesto doesn't make any promises about quality, speed or price. The only lie here is that your claim that it does.
www.agilemanifesto.org
Manifesto for Agile Software Development
We are uncovering better ways of developing
software by doing it and helping others do it.
Through this work we have come to value:
Individuals and interactions over processes and tools
Working software over comprehensive documentation
Customer collaboration over contract negotiation
Responding to change over following a plan
That is, while there is value in the items on
the right, we value the items on the left more.