Currently there is a lot of discussion ongoing on the causes of failing agile teams. After watching these discussions for quite a while now, here is my view of the topic. The brief quint-essence is: If you’re doing crap, it doesn’t care whether you call it Agile or not – you’re just doing crap. If you’re doing a good job, it doesn’t care whether you call it Agile or not either – you’re just doing a good job. About a year ago I started to take a look on how Agile Software Development is done, especially under the circumstances of testing on Agile Teams. Since I realized that my company was struggling with delivery based on waterfall projects and non-cross-functional teams, Agile seemed to be a solution worthwhile to take a look on. Since introducing change in working habits is a large effort and I was not in a key position to introduce this, I started especially focused on the mind-set behind Agile Software Development and the sucess aspects of it. Incrementally I achieved to introduce changes based on the practices from XP – but for several reasons I was confident, that I could not introduce all aspects of it. Today I would say that I did not take the courage to just change stuff and come up with the success story afterwards.
The changes I started to introduce in my company where tiny. We were suffering from shell-scripted test automation, that was very interface-sensitive. A change in one of our software components could lead to a full regression test result disaster, our testing team being unable to give the stamp of approval on it safely. Manual testing was no practical option, since it would have lead to many weeks of efforts for that, but the bugfixes we had on our desk were critical for our customer. My team decided to reimplement test automation with a new approach and we decided to use FIT for it.
The next six weeks were spent focussing on the highest priority business use cases. We were able to build a new framework for testing on it. While I was reading The Art of Agile Development from James Shore and Shane Warden, I brought in many aspects of the practices and underlying principles and values into my team. Though we were not implementing the full set of practices, we were quite successful over the next few months and finally finished our energized efforts with a test framework, so that we were able to run our group during vacational time with just 1.5 persons, while we were unable to hold the pace one year before with 10 persons.
In the meantime one of my colleagues as well introduced a testing framework based on FIT but with a different approach. While my team used Java, his used Python. While my team focused on the readability of the tests from a business perspective reducing all visible test data to the necessary level of detail, his team put every little detail for the tests in there – whether they contributed to the expected results or not. From organizational point of view it was decided, that we needed to focus on just one programming for easier job advertisements. We came up with the solution to use Java as programming language, since most of our developers have either a Java background or are Java Developers. We started to move the previous Python classes to Java, without having the people introduced to Java beforehand. I did not that hard participate in the migration of the classes and see today that there is bunch of technical debt introduced. From my point of view there are two flaws that result from neglecting the craftsman aspect of our work: the table structure and the Java class code.
While I had read the book on FIT before starting to do anything, my colleague just started. Since he did not get the recommendations from people who have invented the framework and the problems they see on several topics, the tables are interconnected and while reading them I get a brain-twister. The second problem was, that we were forced to stick to the wrong table layout while getting the fixture code from Python to Java. As well since the people doing the work were new to Java, they did not come up with Pair Programming, Unit-Tests for the fixture code and no reusable design in the mind. Getting the toolset now ready for the next customer, will be an immense amount of work, if we agree to stick to the table layout, from my point of view.
The conclusion from that lesson for me is, that Robert C. Martin is of course right. With proper introduction to Java, JUnit and FIT the solution would be better today. We did not say, we were implementing Agile. We just focused on the parts of Agile, that helped us very much. The first approach from my side was a success, because we were simply doing a good job with future opportunities in mind and did just the necessary stuff skipping most of the YAGNI implementations. We were not doing Agile, we were not calling it Agile, we were just doing a good job. On the second step, we were not doing Agile, either. We were not calling it Agile, either. We were just doing a bad job, not paying attention to the skill-set of the people doing the implementation, not looking over the code before or even directly after submitting, no Pair Programming. We were doing a bad job, that’s it.
As stated intially, it does not matter, whether you do a good job and call it Agile or not, or whether you do crap and call it Agile or not. In either case you either succeed or fail, no matter what name you give that child.
- From Agile Development to the New Software Engineering (Column by Alistair Cockburn)
- How to Fail with Agile (article by Clinton Keith and Mike Cohn)
- Why can’t people deliver? (blog entry by Alistair Cockburn)
- Quintessence: The fifth element for the Agile Manifesto (blog entry by Robert C. Martin)
- Testers: The Hidden Resource (blog entry by Lisa Crispin)
- The Decline and Fall of Agile (blog entry by James Shore)
- Dirty Rotten ScrumDrels (blog entry by Robert C. Martin)