Last week at the lean service summit, there was an interesting presentation by John Seddon from Vanguard consulting about combining a systems thinking approach with lean principles. One of the things he brought up was the approach some improvement schemes approach change. Processes such as Six Sigma (and in IT, the Capability Maturity Model) start with planning, and analyse the current situation only after planning. Six Sigma for instance calls its steps DMAIC (Design, Measure, Analyze, Improve, Control). Looking at where we are (Measure and Analyze) comes after the map has already been made (Design). This is a bit strange - how can you plan the road ahead, if you don't know where you are?

Managing change is better done by starting at check, so John Seddon suggests a change cycle that consists of only three steps: Check Plan Do. I can relate to this, because I have experienced the power of this cycle in various activities:

• Weekly planning games work like this Check what has been done last week and how that affects the customers' situation, Plan what to Do in the next week. Very simple, but also very powerful, as the software evolves together with the customers' situation.
• Retrospectives fulfill the same role: Check the current state of your process by looking back on the past project (or iteration), Plan new ways to add value and eliminate waste, by looking at what we can do better next time and Do by having the retrospective right in between your current and next project, and using the results directly in the next project.
• A systems thinking techique, that of creating a causal loop diagram (also known as Diagram of Effects) also follows this pattern. Check being telling a story and drawing observable values and the effects they have on each other, Plan finding intervention points within the effects or changing the system by introducing other observable values and Do enacting the interventions or changing the system.

It is important not to forget the Do step. I heard someone criticize Systems Thinking last week, he said he prefers Systems Doing. I think it is crucial to take immediate action after the planning has been done and go through the check-plan-do cycle in short iterations, that way up-to-date information is used in steering your system, so the goal can be achieved with less errors.

Playing the blindfold game at the last xp-nl proved this again - if you steer every 20 seconds, you won't get the blindfolded person anywhere near the target. With 5 second iterations, this is much more doable.