ACCU London May 2009 Talk

Jeff Sutherland: Agile Software Development in the Enterprise

The May 2009 talk at the ACCU London branch was given by Jeff Sutherland the co-creator of Scrum. This was a hugely popular talk hosted at the JP Morgan premises allowing a greater number of attendees, but even so a waiting list was needed. I was one of the 70 or so people lucky enough to get a place.

As the title suggests the talk was aimed squarely at those people that work in larger organisations or distributed teams, perhaps where they are a little more sceptical about the scalability of agile development practices. The blurb on the ACCU web site also describes “extraordinary financial returns” in case anyone needed a further carrot to listen to one of the founders of the Agile Alliance.

Jeff started with a brief introduction to what Agile Development is, along with a look at the core values from the Agile Manifesto and some of the other Agile Methods such as eXtreme Programming. However, rather than just listing XP as another methodology, he actively promoted the use of the XP technical disciplines such as TDD and Pair Programming. This was highlighted further in a later slide that discussed a more symbiotic relationship between Scrum and XP.

He swiftly moved on to the meat of Scrum and spent some time discussing Velocity and how it is achieved. The key factors appeared to be having a skilled Product Owner that drives the ‘Ready’ queue to ensure the developers are sufficiently productive, and the features being ‘Done’ at a level accepted by the business. The first point rose the question of what exactly is meant by “Ready” and how you find someone that is skilled enough in the business to prepare items for developer consumption – which in the Investment Banking world probably means a trader. However, they don’t come cheap, but Jeff’s argument is that they would be worth their weight in gold as they are the lynchpin. This somewhat surprising answer became perfect fodder for the pub discussions later. The second point about items only being considered ‘Done’ when they pass the business acceptance tests, was driven home by a study that showed that 1 hour of immediate testing, when postponed, would grow to become 23 hours of testing later.

After another brief history lesson on how Scrum was invented, Jeff started to go through some case studies to show where they had achieved Hyper-productivity and in the case of one client where that had translated into a significant growth in revenue – 8 times. Along the way we learnt that 65% of requirements change and amusingly that 63% of features are unused, not that any correlation was suggested mind you. The subject then turned to how Scrum scales to large teams with thousands of developers and how even an un-coached team can increase its velocity by 35%, but a coached team can achieve an increase of 200 to 400%. Another favoured sound bite was “Never outsource to waterfall teams”.

The final two case studies were about failed projects and were used to show how Distributed Scrum, not only scales, but was also able to turn the projects around. Although the choice of “Lines of Code” was not greeted as the best metric for rating productivity (which Jeff agreed with) it did show where the project fortunes changed. Both studies were also analysed using “Function Points per Developer/Month” to compare the real projects with the expected velocity according to a previous smaller study which compared Scrum to Waterfall (17.8 vs. 2.0) – they both came in a shade over 15.

Not unsurprisingly there were a barrage of questions from the attendees and not nearly enough time to answer them all. The only recourse was to seek out a watering hole and discuss the somewhat compelling statistical information further over a few beers. Somehow we lost focus and before long a quest was on to find out if Brian Blessed had ever been in Star Trek...

Chris Oldwood

18 June 2009