So what is this White elephant in the Agile camp?
Well, I am thinking that all team based work-flows have components of process and practices. So, when it comes to Agile approaches why are some not more explicit about the process components? Let me explain.
In the webinar by Alan Shalloway, where he presented The Different Agile Approaches: First (XP, Scrum) and Second (Lean/Kanban) Generation Methods. On seeing the tabulated difference/similarities with Agile approaches (Scrum, XP, Kanban, Scrumban) a light bulb went on, especially when he was talking about TOC (Eli Goldratt).
So, for sometime I have been thinking based on my experience, when using Scrum approach in developing projects, and this doesn’t have to be limited to software projects only, that one looses sight of the critical path that is otherwise formally expressed if the project was to be planned out in detail based on the assumption that all requirements have been expressed, collected and analyzed. Now, I am not suggesting we go back to doing waterfall when it comes to software development just to gain what the critical path in the project is. However, what I am suggesting is that bottlenecks arise quite naturally in any team based work-flow, making team resources and/or the work area critical path and in the way of the flow. In fact this is common in any thing that exhibits network like behavior. So for instance you see bottlenecks on a daily basis on freeways or even at grocery check outs as capacity limits are reached at some points on the path. Based on this I realize the following:
In reality all team based work-flows have a process component and a practice component. So when it comes to getting things done (GTD, David Allen), the best approach is to deal with requests in small batches, tune the practice of work according to the flow experienced in the process with the purpose of maintaining a harmony between these two components. For software development the process steps are often just thought of as A.D.D.T, Analysis, Design, Develop, Test. Well actually the D.T process step, well the practice is intertwined especially if you adopt TDD, but lets just say there is a final acceptance test.
Now secretly but surely, lets all say it all together: This doesn’t make it waterfall development. As pointed out in Alan’s presentation that at the granular level you have to follow these steps and really its a matter of scale. Small batch, iterative development approach not waiting for all the project requirements have been nailed down first before coding is clearly the root for any Agile approach. Seriously though, its unfortunate that for quite some time the term waterfall is used as mud, often to harass scrum teams with. This is often by some layer of management and at times scrum masters, who are pushing for team velocity over say helping the team overcome systemic issues that an explicit process of steps would have helped uncover.
When it comes to the Kanban approach, and specifically WIP Limits. Well at the team level this expresses what is needed to get and maintain the practice in harmony with the process. In software development TDD, ATDD, CI & some of the XP practices are patterns that best help in creating a resonance between the practice and the process.
So what of orchestrating this at the enterprise level, which really means work of multiple teams where some have cross team dependencies, potentially dealing with multiple products and at times vying for constrained resources. I see each team has its own harmonic and the following questions came to mind:
[Q] How does one conduct such an orchestra?
- – For instance with Scrum, its stated that there ought to be an SOS but as far as I know nothing else is spelled out about SOS. It’s assumed that this follows the pattern of Scrum,
- . However, in absence of any specific guide and best practice I’ve witnessed SOS being used to air out an issue log, with out the same level of accountability, transparency and expectation being prescribed as it is at the individual scrum team level. In fact the way I see it SOS team should have a prioritized backlog (in my thinking issue log expresses one level of acceptance criteria) representing the product|sprint backlog. This in turn helps with the business decision-making, including the business of product releases at the enterprise level as well as aligning elements of the business operations (marketing, sales, operations, finance, support etc.) to enable customers can attain the best value. I would say the paper on
- better expresses how to orchestrate this with a top-down approach at the enterprise level.
[Q] What is the process at this level?
- – I can only think of the overall release process for the domain, in software environments this is some combination of (release engineering, support, marketing, sales, education …)
[Q] What are the patterns that can be called on to help reduce the impedance and create smooth flow at this level?
- – Can’t think what this would be right now, but recognize that there must be patterns to deal with bottlenecks at this level too. I think what ever it is, must include “effective collaboration”, meaning things that need to get everyone on the same page and hopefully on a
single glass pane
Finally, as for TOC this is well recognized in critical chain planning, a technique used in traditionally planned projects to balance critical resource on a critical path. This leads to the other observation, that I think Alan was expressing, and I have seen frustration in both software and technical operations teams when it comes to Scrum. Whether its Scrum or XP, the frameworks don’t prescribe or guide the team to pick a process path that is best suited for their domain. Even worse is that most practitioners (scrum masters, coaches etc.) don’t see this hole and thus fail to guide teams in addressing this during the adoption phase, often resulting in the cargo cult behaviour, or as I call it monkey see monkey do. The way I see it, process expresses work-flow, transition states and interfaces for hand-offs in the chain of adding value to the product. For instance in technical operations I’ve seen poor handoff’s, and in a way I am sure this implies poorly defined or agreed upon acceptance criteria. There is no doubt in my mind that this was but one symptom of poorly expressed and agreed upon work-flow process that the Agile practice, in this case trying to bolt on the same Scrum framework that the development teams were using, helped expose. Alas in the same environment the core principle of the Agile practice, one of continuous improvement or Kaizen really meant just tinkering rather then seeing the whole work chain and recognizing the constraints for what they were.
So what does this all mean? and why does it matter? Well for one, which ever Agile framework you pick make sure you explicitly specify the process component that you want teams to subscribe to. This will help mature teams to become that much more self organizing and less dependent on manager types to make decisions for them when they hit bottlenecks (constraints) or when the team needs to set right level of expectations with other teams within the organization.