The other day I had an interesting thought. I’ m not sure what precipitated it exactly, but there were several things that led me to this idea I’ve been mulling in my head. Perhaps it was Alistair Coburn’s keynote at Agile 2009 where he said that Agile as a distinct entity was gone; if it was once an iceberg, it has since melted and is now just part of the ocean. It could have been Jeff Sutherland’s presentation where he points out that 84% of IT organizations are self-reporting to use Scrum. Or perhaps it was simply working with a current client when they were asking for my help to come up with very clear guidelines about the number of acceptance criteria that should be allowed for a single story. Anyway, it struck me: as Scrum grows in popularity, are we institutionalizing it? The choice or words is intentional, as I see it as a double edged sword. Agreeing on a standard for Scrum – or any Agile practice for that matter -  can both allow us spread it faster, but it can also serve as a straight jacket. This seems to me to be the key challenge of scaling an Agile technique like Scrum: our very effort to “standardize” it for a large audience can fundamentally make it less flexible and undermine its true value. Now there are probably different levels upon which we could discuss this institutionalization. I won’t tread into the field of certifications, exams and other standards, but rather look to my own experience with several large organizations, where I have personally heard the siren call of institutionalization.

Within the large organization, I can appreciate the desire. Indeed, there certainly are some benefits around transparency, consistency and institutional knowledge to be had when we get everyone to agree to a standard. This is a careful balancing act, as the benefits may prove elusive and the risk is very real. First with transparency, while working code is the best indicator of progress, as teams grow into the hundreds, that can become too much “product” for an executive to effectively review and understand. This closely dovetails with consistency, where teams need to have common understandings of how to exchange stories, what “done” really means, and even what technical standards they will follow.

Technical standards can be quite pernicious, as enterprise licensing agreements, technology stacks and other architectural decisions can place very real requirements on how teams operate. We do like to think of each team as a trailblazer, but at some point, if you have three web development teams, one chooses ASP.NET, one chooses Java and the third chooses PHP, it doesn’t take an expert to see there is wasted energy. Lastly we come to the idea of institutional knowledge. As a consultant I can certainly appreciate that companies have a desire to “do it on their own” and apply hard earned lessons in one project to future ones.

It is not that these are invalid desires or that some benefit can’t be reached from standardizing, or as I’ve been calling it “institutionalizing” a Scrum implementation. Rather, it’sthat we need to understand the cost of each decision we make and weight it against the benefit. I think back to one of my earliest experiences in project management. An aspiring PMP, I was kicking off a project and I took out my PMBOK to find out what tools I could use to charter and initiate a project. From analyzing stakeholders to managing risk, I’m pretty sure if there was a checklist or template, I used it. At the time I didn’t understand exactly how each one applied, so I compensated by using every single one. Surely, this is not Agile, nor how one would advocate something like the PMBOK be used. I took a list of possible tools and considered them a mandatory set of things to do. Just like that, a supporting resource became a requirement and my project was about following lists rather than delivering value.

I see a similar challenge if an organization begins to offer too many guidelines. So what is the right limit for acceptance criteria in a story? Well, in my case, I took a 3×5 note card and figured out how many sentences I could fit on one side. This brought me to the recommendation of 7, but does it mean if I only have 2 that my story is wrong? Of course not. A better measure would if it clearly articulates what the story needs to do, is agreed upon by the team and they can commit to do it in a sprint. Then again, even user stories are not explicitly part of Scrum, so are we now mandating that all project requirements be formatted in this construct? Just because something makes sense in one domain, doesn’t mean it should be applied elsewhere.

Indeed, using a framework for empirical feedback like Scrum provides one with infinite ability to inspect & adapt. Of course, a common challenge with standardization is that, we have two groups of people: one group who is defining the standard and a second group who is trying to use it. Thus we have one group using the process and getting feedback, but a different team is supposed to make adaptations to the process, but is not experiencing that feedback directly. This is the paradox I alluded to before. In order to get a large organization to use a highly flexible & adaptive process, we must take away some of that flexibility by taking decisions out of the hands of individual teams and delegating them to centralized committees. When this is done, does the value proposition remain intact? How do we go about confronting making consistent & transparent Scrum throughout a very large organization without fundamentally undermining the nature of this framework?

I know many people have talked about this topic and I will certainly have more to say about it, but for now I think the key is to figure out how to standardize on the principles and keep maximum flexibility, but I can see this blog post is already too long so let me leave it with that until I continue this thought later. I’d love to hear your thoughts in the meantime.