Scott Ambler on the World of Agile Development
Scott Ambler is the lead methodologist for Agile at IBM, where he focuses on helping globally distributed teams adopt (and adapt) Agile practices. A prolific author with a half-dozen books to his credit (including Refactoring Databases: Evolutionary Database Design), Scott is also presenting at Agile 2009. Matt Heusser caught up with Scott just before the conference to ask him his views on IBM, transitioning, and what he see coming on the horizon.
Matt Heusser: You're currently the practice chief methodologist for Agile at IBM Rational. How did you get to that role, and what does it entail?
Scott Ambler: I didn’t attend the meeting when they were handing out roles, and that’s the one that I got stuck with!
Seriously though, it was my background in Agile and my years of working with customers to help them to understand better ways of developing software. I spend a lot of time helping executives at our customers understand Agile and the implications for their organization. I also help to drive some of our process development efforts.
Matt: What are you working on now, and what are you speaking about at Agile 2009?
Scott: I’m working on the Agile Process Maturity Model, a contextual framework for helping organizations and teams understand and then tailor agile processes and practices to meet their situation. At Agile 2009 I’ll be giving a talk entitled, “Agile by the Numbers: What People are Really Doing in Practice” which summarizes the juicy results from surveys that I’ve run over the years, plus some recent results that haven’t yet been published. There’s often a very big difference between what Agilists claim they’re doing and what they’re actually doing in practice, and until you get them into the “privacy” offered by anonymous surveys you don’t find these things out.
Matt: Talk to me about IBM and Agile. What does it mean for IBM's internal software groups?
Scott: IBM embarked on an official Agile and lean adoption program a couple of years ago although many people were doing stealth Agile adoption years before that. We’re finding improved quality, quicker time to market, and increased productivity among the benefits of doing so. Sue McKinney, our vice president of Development Transformation and Integration, presented some of our experiences last year at Agile 2008 and we continue to share that information with our customers around the world.
Matt: Is that different than for IBM consultants?
Scott: Life’s a bit different for the consultants because they’ve got a different set of constraints. Where people in a software group are constrained by the realities of developing and supporting commercial software, our consultants are constrained by the desires and environments of the customers that they work for.
Matt: How may software people work at IBM? How can you possibly advise all of them?
Scott: There’s over 25,000 people involved with product development in IBM Software Group (which is made up of Rational. Lotus, Information Management, Websphere, and Tivoli). There are over 190,000 people in IBM Global Services, many of whom are developers. Right now I think that there are about 380,000 employees in total.
Matt: There's a criticism in the Agile community that IBM's implementation of Agile is too light - simply timeboxed iterations with working features. How do you respond to that?
Scott: Having our approach criticized as being too light is a welcome change of pace, usually we’re accused of the exact opposite!
Seriously though, you need to adopt a process which is sufficient for the situation that you face. Unfortunately, the situation is often far more complex than what you hear about in the two-day Agile certification courses, and many organizations struggle as a result to understand how to apply Agile effectively.
Matt: Likewise, you've said that Agile is now mainstream, but are you concerned at all that what people are calling Agile is watered down compared to Agile's original, more noble ideas? I mean, yes it's Agile, but people sit in cubicles. Yes, it's Agile, but people aren't doing pair programming, or test driven development (TDD), or refactoring. At what point do you say "Um, er... what?"
Scott: If people are working in cubicles, wouldn’t you prefer that they be as effective as possible doing so? The point is: The goal isn’t to be Agile, it’s to be as effective as possible given the situation that you find yourself in.
Also, you’re not going be perfect from the very beginning, so process improvement needs to be seen as a journey. As far as criteria goes for determining if a team is Agile or not, I’ve suggested some. I have no doubt that a really good non-Agile team could pass these criteria, but if they’re just as effective then I’m not really all that concerned about what paradigm they’ve adopted.
Matt: You used to write a column for Software Development magazine, which was combined with Dr Dobbs, which later became a small supplement to Information Week. Have developer magazines "gone away" or will they come back? Either way, is that a good thing? And what will replace them?
Scott: The media landscape is clearly changing. The economics of magazine publishing have changed for many reasons, particularly in the computer industry, at the same time as the culture of the readership moved from a paper-based one to an online one. The magazines have to change with the times, and I think that online sources of information are here to stay.
Matt: Likewise, the SD conference was recently discontinued, or at least delayed. Where do you think the conference world is headed?
Scott: I was really disappointed to see the SD conferences cancelled as they really offered value by addressing a range of topics. I think that we’re going to see more point-specific conferences, Agile conferences, Ruby conferences... and vendor conferences for the next few years. Once the economy gets better I suspect that the SD conferences will come back.
Matt: You described yourself as a "former CMM"-er and have said, "I used to be one of those guys," meaning someone advocating a particular kind of defined, managed, measured, requirements-up-front software process. What changed your mind? Did you have some sort of a conversion experience?
Scott: There are some really great ideas in CMM/CMMI [Capability Maturity Model Integration] and with a pragmatic adoption you can be quite successful. The challenge is that many organizations choose an overly bureaucratic strategy and are very wasteful as a result.
Nothing in CMMI indicates that you need to be waterfall or non-Agile but for whatever reason organizations interpret it as such. I saw too many CMM/CMMI suffer from these problems so became jaded on it as a result. However, I also see a lot of rabid “anti-CMMI” people out there who very clearly have never taken the time to look at CMMI with an open mind. Agilists talk a lot about the importance of respect, so maybe they should respect that the people behind the CMMI are very intelligent and very experienced and have some valuable insights to share with us.
Matt: The list of ideas you've proposed or championed is impressive: Agile Model Driven Development (AMDD), the Enterprise Unified Process (EUP), the Agile Unified Process, and Agile Data are just a few. Those are innovative ideas, but they haven't exactly taken off in the community. Most of the articles on the subject were written by you and contain references to your other work. A few years later: How do you feel about those techniques? Are they relevant?
Scott: Over the years I’ve taken on some of the less-than-sexy topics in the IT community because that’s what a lot of organizations struggle with. Agile modeling, for example, addressed the issue of how do you approach modeling and documentation on Agile projects, a topic that wasn’t well understood at that time. Surveys that I’ve done over the years have shown a very high adoption rate of some of the practices of AMDD, such as initial requirements envisioning and initial architecture envisioning for example.
But it’s not “macho” to talk about doing initial modeling, unlike talking about technical practices such as TDD for example, so people don’t do it. People also struggle to recognize that they’re modeling; I’ve literally seen people stand in front of a whiteboard with a sketch of their system architecture on it, with a stack of index cards capturing their user stories beside it, and yet still claim that their project team does no modeling.
EUP addresses enterprise-level issues such as enterprise architecture, portfolio management, strategic reuse, and other issues which also aren’t popular within the developer-centric community. Nor are many organizations sophisticated enough to effectively address these issues, even though they desperately want to do so.
The Agile Unified Process (AUP) didn’t really get the attention that it deserved, I didn’t put the effort into supporting and evolving it that I should have, and as a result it sort of died on the vine. However, I do know of several organizations that have adopted it, so “died on the vine” may not be accurate. Every so often someone comes along and suggests that it should be moved into the Eclipse Process Framework, but I haven’t found the time to do so. It would likely take a couple of days of copying and pasting to get the job done, and my wife usually has a different opinion as to what I should be spending my free time doing.
Agile database techniques are in fact catching on. Many of them are followed regularly by Ruby-on-rails developers for example, and we’re seeing more and more tool support each year. The challenge with anything pertaining to database development is that programmers generally underestimate the complexity and skills required and have a tendency to do a rather poor job of it. Worse yet, data practitioners are very slow to change and have a very solid culture centered around serial development. I’m seeing more-and-more data professionals being relegated to operational data administration, clearly an important task in any organization, and out of development or enterprise architecture roles. This is unfortunate for them, for the most part, but likely a good thing for organizations because it will make it easier for them to adopt modern database development techniques. So the jury is still out.
Matt: The discussion about the Agile Process Maturity Model is exhaustive and controversial. Let's start fresh. Could you give us your elevator pitch?
Scott: The short story is that that the APMM is a contextual framework for effective adoption and tailoring of agile processes and practices to meet the unique challenges faced by a system delivery team. There are three maturity levels (four, if you count level zero, which I don’t) that build upon each other:
- Level 1 Agile processes, such as Scrum and Agile Modeling, address a portion of the development life cycle. Level 1 processes and practices are optimized for small, co-located teams and are the building blocks for more mature solutions.
- Level 2 processes, such as Dynamic System Development Method (DSDM) and Open Unified Process (OpenUP), go further by covering the full Agile “system delivery life cycle” from project inception all the way to transitioning the system into your production environment (or into the marketplace, as the case may be). Level 2 Agile processes are self governing within an appropriate governance framework, recognizing that teams are not completely free to make any decision that they want but instead are both constrained and enhanced by your organizational environment and goals. Level 2 life cycles are both risk and value driven; mature teams strive to reduce their business and technical risks as early as possible and thereby increase their chance of success by doing so.
- Level 3 processes, which are tailored forms of level 2 processes, are disciplined Agile delivery where one or more scaling factors are applicable. The scaling factors are team size, geographical distribution, regulatory compliance, environmental complexity, organizational distribution, and enterprise discipline.
There’s a whitepaper coming out in late July which goes into the details, and I’ll continue blogging about it in my Agility@Scale blog
Matt: How is APMM different than the CMM, which was a five-level maturity process to which, at least in part, Agile was a reaction?
Scott: I wouldn’t say that Agile was designed to fight against CMM/CMMI but instead the development community standing up and taking responsibility for defining their practices and processes. Unfortunately, there is a lot of misunderstanding surrounding CMMI in the Agile community (and, arguably, greater misunderstanding about Agile within the CMMI community for that matter).
First, CMMI does not prescribe a bureaucratic approach, nor does it even prescribe how to implement practices. Second, there are two CMMI strategies: the staged model where an organization is rated, and continuous model where individual process components are adopted to an appropriate extent. Third, many organizations are applying Agile in a CMMI compliant manner.
The bottom line is that APMM and CMMI are complementary but different strategies. If your goal is to have your organization’s process maturity rated then you should consider the staged model of CMMI. However, if your goal is to adopt and then tailor the right Agile processes and practices to address the situation that you find yourself in, then APMM provides direct guidance for that. The continuous model of CMMI can help this effort if you are so inclined to go that way. The APMM whitepaper goes into more detail on this issue. Sadly, I suspect that I’m fated to do a focused paper on the topic of APMM and CMMI; time will tell.
Matt: Does APMM need further definition? Is it "just a thought experiment"? Or is it ”done” development?
Scott: APMM, like everything else, will evolve over time. With the publication of the whitepaper I guess you could say that APMM version 1.00 is done.
Matt: What's your definition of a mature software process?
Scott: One that meets the needs of the situation at hand in an effective manner.
Matt: Say you're brought into a group straight out of 1995. They are doing simple “what / how / build / use” software development, releasing every 18 months. They want to know one technique to start using on Monday. What would you recommend?
Scott: It depends on the situation of course, but adoption of short iterations (one month or less), where you need to deliver potentially shippable software each iteration, would likely be my advice.
Matt: What's new and exciting in the world of Agile? What's the new “new thing?” What's the cutting edge?
Scott: I’m really excited about the promises of integrated tooling environments like Rational Team Concert (RTC) which supports distributed Agile development and reduces the need for toolsmithing. It also improves visibility to project stakeholders through its project workbench which displays metrics automatically generated from tool usage, thereby reducing overall project bureaucracy and freeing up the development team to focus on producing working software. At http://www.jazz.net you can see the project workbench of Erich Gamma’s RTC development team, seeing their burndown charts, build history, and other critical measures which support effective development governance.
Matt: Care to try to predict the next steps for the Agile movement in the next few years?
Scott: I think that we’ll see lean thinking become even more prevalent and scaling issues being addressed in a realistic manner. We really do need to do more outreach to other communities, including but not limited to the business analysis community, the data community, the business rule community, the systems community, and the enterprise modernization community. We need to actually apply the lean principle of optimizing the whole and look at the whole IT process picture (something that the EUP does quite well, I’d like to add).
Ideally I wish that the certification scams would disappear (and adding tests to existing schemes isn’t going to cut it), but unfortunately too many of our “thought leaders” find it easier to turn a blind eye instead of speaking out against them. I guess it’s easier to talk about courage than it actually is to show some.