A few months ago I co-authored a white paper (with V. Lee Henson) on the…
I don’t always follow the same story splitting approach when I need to split a story. It has become intuitive for me so I might not be able to write about everything I do or what goes through my mind or how I know. But I can put here what comes to mind at the moment:
Look at your acceptance criteria. There is often some aspect of business value in each acceptance criteria that can be split out into a separate story that is valuable to the Product Owner.
Consider the tasks that need to be done. Can any of them be deferred (to a later sprint)? (And no, testing is not a task that can be deferred to a later sprint.) If so, then consider whether any of them are separately valuable to the Product Owner. If so, perhaps that would be a good story to split out.
If there are lots of unknowns, if it’s a 13 point story because of unanswered questions, make a list of the questions and uncertainties. For each, ask whether it’s a Business Analyst (BA) to-do or a Tech to-do. Also ask for each whether it’s easy and should be considered “grooming”. If it’s significant enough and technical, maybe you should split that out as a Research Spike. Then make an assumption about the likely outcome of the spike, or the desired outcome of the spike, note the assumption in the original story, and reestimate the original story given the assumption.
Look in the story description for conjunctions since and’s and or’s are a clue that the story may be doing too much. Consider whether you can split the story along the lines of the conjunctions.
Other Story Splitting ideas:
- Workflow steps: Identify specific steps that a user takes to accomplish the specific workflow, and then implement the work flow in incremental stages
- Business Rule Variations
- Happy path versus error paths
- Simple approach versus more and more complex approaches
- Variations in data entry methods or sources
- Support simple data 1st, then more complex data later
- Variations in output formatting, simple first, then complex
- Defer some system quality (an “ility”). Estimate or interpolate first, do real-time later. Support a speedier response later.
- Split out parts of CRUD. Do you really really really need to be able to Delete if you can Update or Deactivate? Do you really really really need to Update if you can Create and Delete? Sure, you may need those functions, but you don’t have to have them all in the same sprint or in the same story.
Some of the phrases in the above list may be direct quotes or paraphrases from Dean Leffingwell’s book “Agile Software Requirements”.