What method does the GreenHopper team use to do estimation?

Jeremy Largman

It'd be great to have insight into how the GreenHopper team approaches Agile development.

Nicholas Muldoon

First, let me stress that every team at Atlassian operates differently. You'll see the operations side of things (Sales, Support, Marketing, IT) using a kanban approach while most development teams stick to scrum. There is no standard iteration length or estimation method.

At present the GreenHopper team uses story points for estimation and is doing weekly sprints (Wed - Tues). One of the engineers will usually give a rough estimate for a new story (around 30% of the backlog is estimated). The rough estimate helps me when prioritising work.

Once that issue lands in a sprint I will add acceptance criteria and take it to a sprint planning session. The team has a discussion about the story, clears up any queries and conducts planning poker to get a more accurate estimate. Quite often the acceptance criteria is amended during this session in response to the discussion.

Release planning is currently done on a story mapping wall internally. As epics are broken down they are added to the backlog. Given the size of the team we are only working on one release at a time. Sprints may incorporate stories that are released in a minor release (5.4.1) but are not enabled (ie dark features) for all users until a later major release (5.5). Some other products at Atlassian have a bugfix and a feature team enabling them to work on major releases alongside minor releases.

Late last year we went full tilt kanban. While we didn't estimate individual stories one of the engineers and I would try to ensure all stories were around the same size for consistency.

Our approach really depends on when you catch us as we chop and change the approach fairly often to experiment based on what our customers are exploring.