Dan Vacanti on Predictability and Kanban (Scaling Tech Podcast Ep2)

Jun 28, 2022 | Kanban

Daniel Vacanti is an expert Kanban thought leader, trainer and consultant, who is the author of “Actionable Agile Metrics for Predictability”, co-author of “The Kanban Guide”, and Co-Founder of ProKanban.org. He is also co-host of the “Drunk Agile” podcast. Dan joined David and Arin to discuss Predicability in Kanban. They covered a wide range of topics from why cycle time is really just a measurement of time to feedback, flow metrics in Kanban, and how the most effective teams often combine aspects of Scrum and Kanban.

Listen on Spotify
Listen on Apple Podcasts

Watch the video:
Show notes with links to jump ahead are below

Show Notes from Dan Vacanti on Predictability and Kanban
Timestamp links will open that part of the show in YouTube in a new window

  • 00:00 Introductions by Arin and David. How Kanban focuses on the metrics that matter, and Arin notes how comfortable Dan is with mixing the elements of Scrum and Kanban in order to create the most effective teams.
  • 03:02 Interview with Dan Vacanti begins, first order of business was sharing a drink together. Dan pours himself a Glendronach 23 year old Scotch Whisky. Arin pours a 15 year old Knob Creek, and David pours a Blanton’s.
  • 07:00 Drinks are done and we start with Dan’s story about how his background, and how he was brought into a company called Corbis which is where they worked on scaling many Kanban practices that had been working on an engineering team. Those practices became what we know as modern Kanban for software development. Dan later focused on flow metrics in Kanban and spotting opportunities for improvement.
  • 09:00 Dan gives us his short explanation of what Kanban is. It’s all about focusing on flow in order to optimize the delivery of customer value.
  • 10:30 Dan addresses what makes Kanban predictable, or actionable. Dan talks about how the practices aid in predictability. Dan notes that if you want to NOT be successful, then start 50 tasks at once. Kanban discourages that through focusing on flow.
  • 12:30 David talks about how he and Arin were previously scrum coaches and his initial impressions of Kanban after being burned out on Scrum. Dan and David talk about how Scrum is focused on the team – it’s not possible to do Scrum without a team. Whereas Kanban is more about improving throughput, and doesn’t focus on team practices.
  • 14:30 Dan comments that he doesn’t know how you do Scrum without Kanban, or how you do Kanban without Scrum. There are practices that complementary and he feels the lines between them should be blurred. David and Arin talk about blending Scrum and Kanban. Arin talks about his frustration in the past as a Scrum coach when teams would get too focused on the rituals of Scrum and forget about if you’re actually delivering value or not.
  • 17:15 Dan talks about how Kanban can also still suffer from a lack of rigor, there’s more to it than putting post it notes up on a board. Dan talks about learning from a wide range of methodologies.
  • 18:40 Arin asks Dan how he likes to combine Scrum and Kanban. Dan comments about how Scrum is good at telling you how to set up and tear down a sprint, but it doesn’t tell you anything about achieving flow between those boundaries. That’s where Kanban helps.
  • 20:00 David talks about how moving from Kanban to Scrum with some teams helped him see who on the team “liked to be blocked” in a Scrum process. Kanban helped expose systematic problems and team problems. Dan noted how Kanban won’t help you fix that, but it does give you the evidence that something is not right.
  • 22:10 Dan discusses Flow in the context of estimation and predictability. We are all very bad at estimation. Scrum uses story points typically, but Kanban flow metrics are different. Dan mentions the 4 flow metrics used in Kanban: Cycle Time, Work in Progress, Thorughput and Work Item Aging. Kanban focuses on these metrics because they are in a language that customers can understand, whereas a customer can never truly understand story points. Customers understand these Kanban metrics more easily and that makes them more effective. This allows us to get away from estimation all together, in favor of probabilistic forecasting instead.
  • 26:30 Dan talks about using these metrics to see opportunities for improvement. Predictability is just a probabilistic statement about the future, it’s not a deterministic forecast. Predictability does not mean what you think it means.
  • 28:45 Dan talks about the emphasis on “just in time” mindset of Lean, since over planning is very wasteful. We have to embrace and acknowledge that we don’t know everything, but we can still make progress. This is what drew Dan to Kanban. David comments how it’s healthy to have ambitious goals, but leaders have to be able to revise that constantly in the face of reality.
  • 31:30 Dan talks about cycle time, and how most people think that the reason we track that is so that we can weaponize it and use it against teams who are not getting enough down. Cycle time is really measuring that “time to feedback” because we don’t know if something is useful to our customers until we get it in his hands. Dan quotes his friend Prateek Singh, “The key is finding out how wrong you are, as quickly as possible.”
  • 33:00 Arin asks how you should respond to a manager who wants to know if you will get something done in budget or by the deadline, and talks about answering that with a form of risk management. Kanban will help inform how likely you are to make a deadline, and so you can consider how much risk of that deadline you are willing to take on. If the risk is too high, then looking at how you can reduce that risk. Upper management usually understands a conversation that is framed around risk.
  • 35:50 Dan talks about turning flow data into probabilistic statements about the future, and that the chart to start with is a cycle time scatter plot. This distribution of today shows you the distribution of results you will likely get. The challenge though is that you have to wait for an item to finish before you can measure it’s cycle time. So the hands down best chart is a Work Item Aging Chart, which shows you how long an item has been in progress. This can show you problems in a cycle time scatter plot before it is complete.
  • 39:00 Dan talks about why 85% is typically used as the threshold in scatter plots. It’s somewhat arbitrary, but represents a reasonable and realistic level of confidence. This threshold could change some based on the context of your environment, ie, a very highly regulated environment might want to go a bit higher, but something as low as 50% would be meaningless except in some prototyping work where predictability may not matter at all.
  • 42:15 David asks about dogmatism and camps in the agile movement, and what’s the message that Dan would give to a dogmatic Scrum team member, on how they can improve? Dan talks about how he met Arin in an organization where Arin was a scrum coach and Dan was a Kanban coach, but they were both very pragmatic and found a lot common in the approaches to value delivery.
  • 44:50 David asks the group, what their takeaways were from this conversation. Arin notes that Dan’s description of how “cycle time is the time to feedback” was a really important reminder. David really enjoyed the predictability and scatter plot conversation. Dan’s takeaway is that agile doesn’t have to be either/or Scrum or Kanban, it’s a false dichotomy to think you have to choose one or the other. Dan notes that he is looking forward to joining us again in the future to talk more about things like aging’s impact on predictability and more about probabilistic forecasting.

Links from Episode 2 – Dan Vacanti on Predictability and Kanban