Kanban metrics
complete
Vidas
complete
Vidas
We've completed first batch of metrics as PoC through layers. After we gather data we will run the second iteration which is metrics per column (or similar alternative).
Vidas
in progress
Vidas
planned
T
Tom Davidson
when as in a Monte Carlo fed histogram?
Vidas
Tom Davidson: Monte Carlo is one of the possible techniques to solve "when" question. But I cannot confirm that it will be the one chosen, we are investigating proprietary math/statistical solutions as well.
T
Tom Davidson
Vidas: thanks, my orgs will not be interested in proprietary or black box analytics.
One of the powerful things of kanban is it's timelessness. Flow metrics are not hip and new, but timeless and proven.
Why do something different?
Maybe expose a Prometheus endpoint that we could scrape for our own analytics?
Vidas
Tom Davidson: very valid point regarding black-box analytics, non-transparent method will be met with skepticism.
We are not yet committed to any technique. Monte Carlo has it's own flaws by focusing only on throughput and quatifying items as almost equal pieces. We do not intended to invent something out of the blue, we still want to use flow metrics.
Goal here is to be more accurate and tangible, rarely an average user would get value from Monte Carlo, at least how it is being done nowadays in other Kanban tools. If we fail to find a better way, we go with Monte Carlo
T
Tom Davidson
Vidas: work items of different sizes do not matter for kanban or flow metrics and if teamhood goes down that path of story points and faux kanban then Im out.
The predictably of a flow system is all about probabilities which monte carlo is perfect for. 85% confidence based on historical performance that we will hit this milestone by such and such date is immensity more valuable to predictability than sprints, sizing, and deadlines (ie SCUM bull). With kanban and flow metrics we use data instead of human judgment to make forecasts - sizing will do more harm than good because it poisons the forecast with human judgment.
Vidas
Tom Davidson: thank you for caring and sharing. I will dedicate time to get wiser before we reach some sort of conclusion. And I will share the idea here prior to even making a comittmet to deliver a specific product change.
Vidas
We have started with Cycle time and Throughput, we will show these metrics on the board level. Then we will use them for our own concept of layers. Lastly, monte carlo simulation will follow.