From Waterfall to Weekly Releases: A Case Study in using Evo and Kanban
In 2003, we had a major problem to solve - our products had far too many open field defects, and the bug arrival rate was moe than the closure rate. We tried to fix using our process which involved shipping quarterly service packs, but that was not only elongating the lead time, it was also not very amenable to changes. The process for customer specials (specific features, etc.) was not any better, and invariably it led to exec-level escalations just to get some deal-blocking customer escalations into the service packs mid-way in the quarter.
In 2004-05, we experimented with a pull system that limited the work in progress and created a more smoother flow of value. The result of this system was that we were able to significantly reduce the defect backlog, and were able to bring down cadence of features and bugs to a weekly cadence. The experiment was so successful that in about 6 quarters, we had fixed most of the field defects (brought down individual product's defect backlog to single digit) and we had to disband the team as there was no work left for them!
We were inspired by Tom Gilb's "Evo" method and experimented with it to create a weekly cadence. However, we found that the nature of field defects and customer specials was stochastic in nature and didn't lend itself very well to a timeboxed framework like scrum. However, there were no off-the-shelf methods available back then that were a viable alternative. Hence, we experimented with various methods and blended-in elements form various methods to create out our very first kanban by limiting work in progress.
In this talk, I will explain our first kanban experiment, and also compare and contrast with the later-day Lean Kanban by David Anderson.
Outline/Structure of the Experience Report
1. What is Lean Kanban
2. Case study
1. Understand the concept of pull, WIP, visualization and flow
2. Being able to apply these concepts to a real-life project