Thursday, August 20, 2009

A further experiment in ScrumBan…

I’ve recently evolved my company process to include WIP limits, but before I share this story, I must first provide the following background.

When I started at this company 2.5 years ago, there was a large divide. I was trained in PMI, was a CSM, and I was a staunch advocate of XP/Scrum by the book. On the other side of the divide, the company had little process, little documentation, no source control, and made minimal attempts to plan work a day or two in advance. One could say this was a recipe for disaster, but this is a startup… and the point of hiring me was to improve as they doubled the size of the company.

Several things have happened since then:
  • retrospectives and weekly iterations were quickly adopted
  • source control is now in place
  • we put tools in place to manage our work (VersionOne)
  • standups have become common for both teams and projects
  • we’ve modified the 3 questions
  • we now use a ScrumBan inspired board
  • most importantly, I’ve learned that there is more to Agile than Scrum/XP… I’ve become an interested follower/borrower of Lean and Kanban practices. This is an important step for me as a coach since not all environments fit the Scrum/XP model. A mixed bag of tricks helps win the debate, especially when it is gamed against you!
So, what is the new problem of the day? I’ll call it the plan-to-budget problem. I’ve struggled for 2.5 years to find a way to get the company to plan work based on the available capacity. We’ve always had “BUT-we-need-to-do-ALL-of-it”-itis.

I’ve explained velocity and scrum, but this was viewed as too expensive to pursue. “Why should we estimate everything when we only care about certain parts of it meeting a specific deadline?” This is a good point. If only a small portion of our work is deadline driven, and the rest is “do as you can in this order”… then putting time into estimating and projecting everything might lead to waste. (Although this is contradictory to “I-want-all-of-it-by-next-month”-itis.)

Thus, I reduced focus and tried a top5 concept. Hey CEO, pick only 5 things that are important for this week, we’ll do our best with that set before touching other stuff. This worked well for a while since it forced us to admit we can’t get everything done and focused on fewer items. It reduced our “in flight” work… for a while. Two things derailed this approach.
  1. The CEO found ways to assign work to people outside the top5 set (most people can’t say no to the CEO/founder when approached).
  2. Not all items were the same size (scream together now: ESTIMATE!) so 1 large top5 item could clog the whole pipe.
We let go of this concept and evolved our breakdown further. I was happy to push for smaller stories and quicker closure. Appropriate sizing is a great stress reducer for the team, especially when you involve them in that process. But we slowly returned to our problem of demanding more work than we have capacity for.

This is where I started thinking about Kanban’s WIP limits. After all, this would fit our model well, and we already have the board in place. Unfortunately I knew that this would be bypassed just like the top5 limit. Our problem was a cultural problem, and I had to tackle it at the source.

So, we’ve started a new trial change in our process. After 3 hours of debate, I was able to convince management of the following (meaning they agree with these):
  • Giving one person too many things at once will slow down throughput
  • Swarming is a good way to get things done faster
  • Saying we have to do everything doesn’t mean it will get done
  • Not deciding on priority means it will get decided by the people doing the work
  • Our biggest issue might be over-committing (actually, over-requesting since nobody was necessarily committing)
The solution? Headshots! I made little headshots of each team member. We plan the queue of work and then prioritize it. Then the CEO points at a card on the board and says “I want that first”. We put 1-N heads on it based on availability, capability, and size of work. The CEO keeps pointing at items until we run out of headshots. If he points at big things, he picks less stuff; if he points to small things, he can pick more (Budgeting to capacity!) When a person frees up, they move to something else on the board. Everyone on the team might have a few things assigned to them in VersionOne, but they KNOW which item is most important (the one with their head on it on the board). Everyone on the team focuses on helping each other finish these items before starting new things. All of this is clearly transparent on the board! It's not rigid like scrum velocity or Kanban column limits, but it is a place to start and a way to prove value (a door to further improvement).

Summary: We aren’t doing true scrum, but some of the pieces. We aren’t doing true Kanban or Lean, but some of the pieces. We aren’t truly self-empowered, self-managed… but somewhat. I would be the first to declare we aren’t “agile” as defined by most in the community. BUT, I’m getting to the point where I believe we have a system that works for us (2.5 years ago, we were a demand-and-control company led by the CEO). It’s a hybrid of many agile approaches, driven by my teaching of agile philosophies, allowing this company of ~70 work better together. Hopefully this new change will help us overcome our capacity/planning issues, only time will tell.


  1. Thanks for sharing this, as I too see gaps & limitations in XP/Scrum. The biggest being that last day or so of Sprint when you realize a feature isn't going to fit.

    1. Do you hack it up?
    2. Do you hurredly finish it and introduce bugs?

    I think as some of the slower brick & mortars adopt Agile, the smaller leaner companies will have already made significant steps into Lean & Kanban adoption.

    It'll be a challenge but as we know XP / Scrum is not the silver bullet, so why not adopt other pieces that make sense right?

    Definitely keep us in the loop on how this progresses!

  2. David-

    I didn't see these as gaps in Scrum as much as I couldn't convince my current company to abide by Scrum estimation and iteration practices. In my mind, this was our fall-down, not scrums.

    To your point, this is something my last team dealt with a lot. As we matured in Scrum/XP, we learned that it was okay to redefine our sprint planning commitment. Instead of the team committing to a specific set of stories for the sprint, we were committing to a specific velocity and an "intended" list of stories. This was important for two reasons:
    - it allowed everyone (team and PO) to negotiate when things changed and we needed to be 'agile'
    - it gave us the ability to keep working through the end of the sprint and get started on the next one.

    In your case, your team should just work as normal. If it is not done, it doesn't get merged into source control and the build. It should just become a story that is re-estimated at planning and carried into the next sprint... depending on how much work was done, the estimate might go down.