Opinion: todo - doing - done is too limited

November 24th, 2020

Like mentioned in Can we use an Andon Cord in programming?

  • What type of learning could we do?

  • Which problems can we swarm?

  • How can we raise the bar to increase problems?


We usually don't consider the entire value stream (Model: from idea till thank you)

And we certainly don't visualise it!


In Toyota they play with buffer size (aka Kanban's WIP limit) to create problems.

Maybe if we broaden the scope beyond todo-doing-done, we can do the same.


Which steps are there?

  • Problems

  • Options

  • Possible Experiments

  • Committed experiments (aka todo)

  • implementing (aka doing)

  • implemented & deployed (aka done)

  • data gathered

  • review

  • decide next actions

  • "thank you"


(Inspiration Video: PopcornFlow - Claudio Perrone)



We can visualise this value stream.


Quality issues

Every time we need to "go upstream" because of a "quality issue" with a decision, we can swarm that problem.

(assuming Model: decisions are the unit of work in software development)


Flow issues

We can also add WIP limits on each of these steps.

Keep in mind Quote: purpose of kanban, by achieving single piece flow.

So we can slowly keep reducing WIP limit everywhere until we reach WIP=1.

Maybe even WIP=0?

And when this starts causing flow issues, we can swarm them.