Take cubes of 3-5 different sizes.
Take two boxes of some size. Name them A and B.
For each task in the backlog ask the developer to pick a cube of size proportional to the size of the task.
Start with a task that is obviously complex and big. Provide a short justification for why the task is complex and big. This will help members compare other tasks to it and choose appropriate cubes.
Keep adding the cubes to the box A.
Stop adding tasks to the sprint once either the A is full or none of the member is prepared to take up further tasks, whichever happens first.
There will be some place left where you can fit small cubes. Pick up some small task and add to the sprint.
Keep the boxes near your work area.
Ask developers to take out cubes from A as and when they finish the tasks and put them in B.
As the productivity of the team improves, as the members become more aware of the system and get used to it, they may start choosing smaller cubes for tasks for which they would have chosen bigger cubes before.
The cubes and the boxes have their volumes printed.
P = Total volume of cubes in B / Volume of B is the Productivity Factor.
E = Total Volume of cubes in A / volume of A is the Error Factor.
D = Volume per developer.
Goal is to keep improving P and D. E may not always be a concern (depends on your company/project).
Any ideas/suggestions?
A Maker's Blog 