This paper has been a long time coming. The first draft of this was written…
One of the big wins of any agile method you pick… Scrum, Kanban, Extreme Programming, AUP… is that you get a built-in ability to balance capacity against demand. This natural throttling of work helps us better manage expectations between the customer and the team. It allows the team to work at a sustainable pace, dramatically reduce thrashing, and helps everyone get more predictable over time.
- Scrum has us meet with our Product Owner once every couple of weeks during sprint planning. The Product Owner gets to decide what we build. The team gets to decide how and how much, based on their historical velocity.
- Kanban introduces visual controls and explicit work in process limits to make sure that teams don’t confuse being busy with actually creating value. We can’t start more work, until the other work we’ve already started has left the queue.
When we get in the business of comparing the merits of the various agile approaches, we all kind of assume that limiting work in process is a good thing… it’s just a matter of how of best to accomplish that goal. That said, I’ve actually encountered a few organizations over the past few months that don’t value limiting work in process… they value saying ‘yes’ and then letting the chips fall where they may.
It is politically safer to say ‘yes’ now, give it your best shot, and then deal with the consequences of missed dates, and blown expectations, later after everyone tried really hard to make it all work. In my opinion, this is one of the most challenging barriers to adopting agile… and one of the most corrosive factors to actually sustaining agile in an organization after it has been transformed.
At the team level, people usually understand the problem, and are open, but don’t believe their managers will ever let them throttle back. The higher you go up the food chain, the less connected people are to the day-to-day work, and the more connected they are to the desires of the business. At that level, the deal has been sold and the work just has to get done. It’s up to the lower levels of the organization to figure out how to make it happen.
The irony is that the more we throw at the teams, the more we pressure them to deliver, the less they get done, the more defects get built into the system, and the more technical debt gets accumulated. Ultimately our ability to build working software grinds to a halt and our relationship with the business degrades to the point of open hostility. It’s a vicious cycle that someone has to step up and break.
Any agile method will give you some way of understanding the problem and some tools and techniques to do something about it. We might choose to use a prioritized product backlog and a Scrum board, or maybe a Kanban board with explicit work in process limits. Either way, if we can back off long enough to let the teams stop thrashing, establish a steady cadence of delivery, and a stable throughput… we have a chance to see the problem and do something about it.
Our only other option is to ignore the problem and hope it get’s better. In my opinion, hope isn’t a great business strategy, and ignoring the problem isn’t likely to make it go away. Doing something about your problems though requires taking risks… it requires courage to stand in the face of overwhelming pressure and saying ‘no’… or at least ‘not now’. But really though… what other option do you have?