Cycle time, can't wrap my brain around it
Per PSK,
Cycle time metrics are always given with finished work in a percent of time as I understand it.
If I have a 3 devs, and we complete our work 80% of the time in 6 days how exactly do you use that information to better the team? Is it even beneficial to try to forecast with cycle time or is it strictly to be used in retro to try to improve.
If I have 100 stories left to complete on a team with 3 team members and 80% of the time we finish work in 6 days or less, how would you calculate that forecast or is it too incomplete of a picture?
Thanks for any help understanding this.
If I have a 3 devs, and we complete our work 80% of the time in 6 days how exactly do you use that information to better the team? Is it even beneficial to try to forecast with cycle time or is it strictly to be used in retro to try to improve.
Your team is trying to meet a Sprint Goal, and an understanding of cycle time could prove invaluable to their plans for key backlog items. When should we start them, with a reasonable expectation of their being finished?
Bear in mind that improving cycle time could actually be a Sprint Goal itself.
what I am struggling with is the math to produce that "reasonable expectation" in that scenario.
If we finish our work 80% of the time in 6 days or less but there are 100 items left to complete the feature how does one come up with that reasonable expectation.
Or in a sprint goal scenario lets say we have 10 items to finish to meet the sprint goal and we finish our work 80% of the time in 6 days or less...
6 days x 10stories - 60 days? / 3 devs = 20 days there is a 80% chance we will be completed? Is that kind of the sprit of it?
I am about the worst arithmetician in the world but yes, that is the spirit of it. I suspect the wisest approach however may involve the running of simulations using collected historical data.
I wrote a couple of blog posts on such matters:
https://www.scrum.org/resources/blog/monte-carlo-forecasting-scrum
https://www.scrum.org/resources/blog/flow-optimization-sprint-boundary
I think you are asking about probabilistic forecasting, which uses a range and a probability. You can use your data and mote carlo simulation to forecast a Sprint and a release.
Here is an additional blog explaining it: https://www.scrum.org/resources/blog/create-faster-and-more-accurate-forecasts-using-probabilities
I'm with @Ian on being bad at math. However, these books by Daniel S. Vacanti really helped me understand how to use metrics like Cycle Time for forecasting.
- Actionable agile metrics for Predictability
- When will it be done?
Thanks everyone
At some point I hoping to get the actionable agile plug in to help me do these forecasts and monte carlo sims.
Focused objective (the company) has some free spreadsheets to look at.
Cycle time and the related scatter chart is for 1 item (the next thing coming in will be complete in 8 days or less 80% of the time.
When you look at groups of items or over a time period, now you need some sort of simulation - that's monte carlo.
It's like saying - there is a 70% chance of the next person walking in this room is going to be under 6 foot tall. That's cycle. Asking "what's the chance of the next 10 people walking in the room being under 6 foot tall - that's where the simulation comes in.
It took me a while to get this under my hat also.