“Oh, yes, we’re using Agile,” my client’s two CIOs told me.
Their company had engaged us to assess their IT practices and effectiveness. In addition to consolidating to one CIO (some recommendations are easier to figure out than others), we also suggested IT should adopt Agile to guide its application development efforts.
Their situation wasn’t all that unusual. The road to Agile is replete with forks. Travelers should avoid many of their tines.
My client had chosen a popular one: They were practicing, not Agile, but Haphazard.
Sadly, they ignored our 12-step App Dev recovery program, the two CIOs presided over a calamity, and eventually our former client became a division of a larger non-client.
Astonishingly, many business and IT leaders continue to misunderstand Agile, and that doesn’t include last week’s advice about applying Agile thinking to business situations beyond application development.
In some cases the misunderstanding is willful, grounded in distrust that Agile’s open-endedness can lead to useful results. If you share this distrust, a question: When you were a child, did you draw up a roadmap for your growth and development that would take you, step-by-step, through your adolescence, remaining education, personal relationships, offspring, careers, retirement, and demise?
Please say you didn’t. I recall, as a teaching assistant back in my grad school days, talking with pre-meds who discovered, in their junior or senior years, that they really didn’t want to become physicians after all, but felt trapped by the choices they’d already made.
I’ll leave the analogy there for you to explore at your leisure. Here, for your edification and amusement, are some other popular misunderstandings about Agile:
Agile isn’t Haphazard: With Haphazard development, project team members wake up each day figuring out what they should do to move the project forward. Or, nearly as bad, the project manager wakes up trying to figure this out.
No matter which Agile variant you use, the project is built around an organized list of Things the Application Should Do — the Backlog.
Agile testing isn’t haphazard, either: With Agile, testing is, if anything, more planned than with its Waterfall alternatives: Those developing a chunk of functionality — usually a module defined by a User Story — know when testing should start on their work. Even more important, every module includes, as part of its definition, a test plan.
Agile isn’t Scrum: It’s a squares and rectangles thing. All Scrum is Agile; not all Agile is Scrum.
Scrum, the most structured Agile variant, is especially popular among IT folk who, schooled in Waterfall development, need control structures and formal governance. There are, however, worthy alternatives.
Especially, look at Kanban, in which developers, when they finish work on a User Story, pull another one out of the Backlog, usually the one (1) with the highest priority, that (2) they’re qualified to develop. Alternatively the project manager assign a user story to them instead.
And, especially if you’re implementing a COTS (commercial, off-the-shelf) software solution, look at “Conference Room Pilot” (CRP) or its close relative, Acceptance Test Driven Development (ATDD). They’re good choices for packages because they’re built around business users trying to do their jobs using the package.
They try, that is, with a real-life business transaction. Wherever they can’t process it, they tell the developers locked in the conference room with them what the problem is. The developers, using the configuration tools built into the package, adjust it to make the problem go away. Rinse and repeat.
Agile disrespects architecture: Developers don’t just write code however they’d like. They develop in conformance to a well-defined application architecture — probably a microservices architecture, but what matters is that, early in the project, the team will, collaborating with the Architecture Review Board or equivalent, establish architectural standards to guide development.
Nor do developers modify the data model whenever they face a situation that calls for it. The project’s consulting data designer reviews the suggested data model change with the developer and, once the two have converged on the best approach, makes the changes happen.
An equivalent set of conversations happens with any other developer solutions that don’t fit the company’s architectural standards.
What doesn’t happen is the developer filling out a form to request a change, for review by some governance committee or other.
That would be decidedly not-agile, as a key Agile principle is that individuals and interactions are more important than processes and tools.
It would also, by the way, ignore a key KJR principle: That there is no process so glacial that it can’t be slowed even further by involving a committee.