We wanted to see through the jargon and analyse LeSS and SAFe against real significant patterns in large organisations.
We gave a presentation Scaling Agility Explored at XP2015 at XP2015 Helsinki last week.
Everyone did not like our message.
Yes, our presentation disagrees with what SAFe says. We tried to dig deeper than what is visible in surface of SAFe (and also LeSS).
We analyzed significant patterns in big organizations, and then matched LeSS and SAFe against those patterns. We did not try to analyze what SAFe says or plans in detail, but rather do pattern matching against our analysis.
In our experience, and by looking at the reasoning in the presentation, cross-functional feature teams are very very difficult to create. In our opinion promoting is not enough, but serious focus and structural changes are needed to create and sustain them.
Unfortunately the word bureaucratic has a negative tone. However, that is exactly the term that Ouchi is using. Of his three control mechanisms, market, clan and bureaucratic, SAFe big picture, program process with descriptions formultiple roles and responsibilities, match the definition of bureaucratic control.
In our opinion backlog is a queue, and large backlog is a large queue. Our conclusion was that technology-specialized organizations need to plan longer ahead, there is no way around this. The plan created in release planning is a queue and release planning is event that moves the queue in to the development system. Of course lazy evaluation is a good strategy to work with that queue, though difficult with many dependencies.
8-10 weeks is closer to to 3 months than 2 weeks. We did conclude only that planning in detail for 4*2 weeks increases the organizational batch size and creates queues. We did not say batches and queues are bad we just quoted Stefan Thomke and Donald Reinertsen who have analyzed in detail queues and batches.
I am sure we found some fresh perspective. I hope, that the discussion will go towards both improving the analysis we started, and help to improve any scaling framework.