Everyone in the team states their goal for the retrospective, i.e. what they want out of the meeting. Examples of what participants might say:
I'm happy if we get 1 good action item
I want to talk about our argument about unit tests and agree on how we'll do it in the future
I'll consider this retro a success, if we come up with a plan to tidy up $obscureModule
[You can check if these goals were met if you close with activity #14.]
[The Meet - Core Protocol, which inspired this activity, also describes 'Alignment Checks': Whenever someone thinks the retrospective is not meeting people's needs they can ask for an Alignment Check. Then everyone says a number from 0 to 10 which reflects how much they are getting what they want. The person with the lowest number takes over to get nearer to what they want.]
Set up a 'retrospective mailbox' at the beginning of the iteration. Whenever something significant happens or someone has an idea for improvement, they write it down and 'post' it. (Alternatively the 'mailbox' can be a visible place. This can spark discussion during the iteration.) Go through the notes and discuss them. A mailbox is great for long iterations and forgetful teams.
This activity is helpful when a team is facing an ambiguous, volatile, uncertain or complex set of problems and has many suggested action items to choose from.
Draw a Landscape Diagram, i.e. an x-axis labeled 'Certainty about approach' and a y-axis labeled 'Agreement on issue'. Both go from low certainty / agreement in their mutual origin to high towards the top / right. For each action item ask 'How much agreement do we have that solving this problem would have a great beneficial impact? How certain are we about the first steps toward a solution?' Place the note on the diagram accordingly. When all actions are placed, shortly discuss the 'map' you created. Which actions will give the greatest benefit in the next iteration? Which are more long term?
Choose 2 actions from the simple / ordered area of the map or 1 action from the complex area.