Monday, September 15, 2008

Debunking some myths about Agile


I think I get quite a few comments on this blog and I publish most of them. Almost all of the ones I reject are spam/graffiti type, people looking to put a link to their site on my blog. Sometime I reply to a comment but not always, it depends on whats on my mind and how much time I have.

Last week a comment arrived on my Failures in Agile process entry from April 2007. Its worth pointing out that the comment has very little to do with what I wrote in the entry. Except for the fact that both my entry and the poster are concerned with “Agile failure” the content of the post seems to have little connection with what I wrote.

What’s interesting is that this post exists. Its a reminder that many people out there still believe a lot of myths about Agile development. The poster is right about some stuff but then just lists a bunch of myths with no supporting evidence or examples.

So lets take the post and dissect the myths:

• “Considering the reasons for failure you named someone may conclude there is no difference between Agile project failure and RUP project failure. In other words the software development approach obviously does not make difference.“

True: indeed I concluded this myself when I wrote: ”Failure of an Agile project looks a lot like the failure of a Waterfall or RUP project.“

The RUP folks increasingly claim that ”RUP is Agile”. There is long post on that subject but not today. If the RUP folks are right then you would expect RUP to fail like an Agile method because it is an Agile method.

There is a stream of thought that say that method doesn’t make much difference. Check out “Amethodical systems development: the deferred meaning of systems development methods” by Truex, Baskerville, and Travis if you want the full details.

It is widely felt that people, not process or method, are the key ingredient for successful software development. So how do you get good people? I think you need to grow them, develop your own developers. And I think the way you do this is through a learning culture. And I argue that Agile development has its roots in Organizational Learning so is better suited than RUP, SSADM, Waterfall, etc. to do this.

For my full argument: Buy the book, Changing Software Development.

• “The main bottleneck I came across is the failure to define good system requirements. In other words: garbage in - garbage out, principle applies here.”

True: Not sure it is always the main bottleneck but I think can be and it is always significant. This is one of the reasons I’ve devoted a lot of blog entried (and the odd conference presentation) to Product Management.

Although most Agile methods say little about requirements I think they do help. By improving deliver they free up management time and energy to think about that they actually want.

• “Agile wins more attention of management becacuse they discovered the way to shift the responsibility and blame for project failure to programmers...”

Not in my experience. I think managers have always blamed developers and developers have always blamed management.

But blame is not a very useful concept.

• “Downsides of agile:”

To the myths...

• “agile may only be used on small projects”
• “maximum 5 developers on project“

False: this myth started because the early writers on Agile hadn’t applied it to large projects when the wrote the first books.

I’m working with a large .com at the moment to introduce Agile to a large (50+) development team. Jutta Eckstein wrote about large Agile projects four years ago.

Software development always goes better with small teams. Since when did RUP, Waterfall, etc. etc. guarantee success? At the very least Agile is no worse than the next best methodology.

One of my personal mantras is: ”Inside every big project is a small project struggling to get out“. To find this project you need to a) be able to deliver, b) actually deliver business value.

• ”all developers must be higly skilled (no learning moments are allowed)”

First part: Highly skilled developers help any project go better. Agile is no different. (With the possible excpetion of Mongolian Hordes I’m not aware of any development technique that claims to do better with less skilled developers.)

Second part: False, Agile is all about learning, there are lots of learning moments. Again, buy the book, Changing Software Development - I’m not going to repeat myself.

• “developers must be in the same physical room with the project owner.“

False: there is a lot of experience with distributed development teams using Agile.
And once again, any development method goes better when teams are in the same room.

Frankly, distributing your development teams is always going to hinder your efforts. Putting them in one room is the most effective way. However sometimes you just can’t do it. And sometimes companies are prepared to take the productivity hit because the cost savings compensate for it.

This came up in the work I did on Conway’s Law.

• ”project owner must be available 24/7 since good programmers also work extra hours”

False: There is no law that says “good programmers work long hours” so the assumption is wrong. In fact, Agile emphases sustainability, you don’t get that by working people long hours.

Lets be clear: Agile does not mean working long hours. If you are working long hours you have a problem and you should fix that.

There is a problem with Karoshi (death by over work) in some cultures but this can’t be blamed on Agile or Lean. Yes it exists, yes it occurs at Toyota in Japan. No I have not heard of any stories of it occurring in Agile Software Development teams or at Lean enterprises like Dell or Amazon. Or even at Toyota outside Japan.

The original XP project, C3, did cause the near nervous breakdown of the first “Customer” - Business Analyst or Product Manager really which was a failure in C3. However, this does some way to supporting the view (held by myself and the poster) than knowing what you are doing (requirements) is important and thus you need more effort here.

To work someone to death or a breakdown is a problem that needs to be fixed.

• “agile does not support use of virutual teams or outsourcing constructions“

False: Where does this idea come from? I’d love to know so I can understand what to disprove!

OK, lets try...

”virtual teams“ - see above.
”outsourcing“ - ThoughtWorks has built a pretty big business as an Agile outsourcer.

• ”agile provides poor quality of software since documentation usually does not exist. If it does - it is written in "reverse engineering" style.”

False.

I have never seen any evidence that documentation leads to high quality software. Year ago I worked on rail privatisation, we had lots and lots of documentation, and then some more. The software was rubbish.

Agile provides high quality software because it simply wouldn’t work without quality. Look at TDD, pair programming, customer involvement, etc. etc. Its all about eliminating re-work, i.e. improving initial quality.

The tests produced by Agile developer are documentation, executable documentation at that but documentation all the same.

And reverse engineering documentation is actually the only sort worth having. Since design and requirements are emergent anything written in advance will change.

• “selling "agile project" as "fixed price & fixed scope" is mission imposible since scope of the project is not defined.“

False: it just means you need a different type of contract.

If Anonymous would like to reply to me please do, but please, next time you make some of these wild claims please explain your thinking or cite your evidence.