6 players + 1 convener (Cheryl) convened to experiment with open-source getKanban v2.0.
H & K played their own board with standard rules starting from Day 10 to about Day 12 before moving on to other sessions.
Cheryl worked through the basic rules and "played" the scripted Day 9 with 3 folks who missed both of yesterday's sessions and wanted an introduction. We discussed WIP limits, creating a pull system, and the impact of the starting queues on cycle time. They took the URL to download the open-source game kit and then moved on to other sessions.
Stuart worked through a complete game experiment on his own and finished through Day 21 before lunch. He wrote up a summary of his game choices and outcomes:
Used Cheryl's "potential subscribers minus cycle time == actual subscribers" rule.
Start: set aside two testers, instead of one, and played with only one tester.
2 devs per day rotated into Test, with 1 dev (a different person each day, because morale) remaining in Dev.
Day 12: changed the WIP limit in the Ready column from 6 to 3.
Day 13: ignored "Carlos".
Day 15: changed the other WIP limits to Design 2, Dev 3, Test 2.
Day 17: hired the extra tester, bringing total testers to 2.
Day 18: by this point the board looked good – light overall WIP and clear queues.
Day 20: just missed delivering the F2 trade show story so didn't get the 25 subscribers.
Day 21: cycle time reached a 3-6 day range (down from 8-10 at start).
Final gross revenue: $8,600.
Cheryl is most impressed that even with half as many testers, the outcome of the game was not much different from a typical game! Swarming Test to knock down the queue and reducing WIP to keep it from re-forming had the same impact on overall cycle time. Total throughput/revenue was affected by the lack of tester, but cycle time was able to be reduced just as effectively. Very cool finding!
Richard stopped by to talk about his and Renee's idea about removing columns on the game board. We haven't playtested it yet.