Wednesday, February 8, 2017

Choosing an adequate sprint length

New Scrum teams must answer the question: "What sprint length should we choose?" Being inexperienced in Scrum, beginners often address that question to me. I have an opinion - which does not matter in itself. Let me give you the reasons why I have formed my opinion:

I advise new teams to start with 1 week sprints, because:

  1. You from new habits quicker when you have the ceremonies every week.
  2. You learn to deliver smaller batches when you MUST get some thing done within a week.
  3. The ceremonies are shorter and feel less invasive.
  4. You'll do more experimentation with the process.
  5. You get faster learning cycles.

There is another reason that may be more compelling, depending on the reasons for choosing Scrum in the first place: Team effectiveness.

Retrospectives are the Scrum ceremony where process changes are discussed. I call them "experiments", because the team is changing something they hope will work out - not something they have experience with.

If you do 1 significant process experiment per Sprint, denoting that Significant means a 5% difference in results (although this is hard to quantify), assuming half of your experiments succeed, within 1 year:

  • If you do 1 Retrospective every 4 weeks, you will be about 35% better. That will hardly be visible.
  • If you do weekly Retros, it will be about 200% better. That is a massive leap which everyone inside and around the team will notice.

Where do you want to be after your first year?


  1. I agree that shorter sprints encourage/enforce a steeper learning curve. On the downside I would see that it puts proportionately more initial stress on the team (which is not neccessarily bad) and it is more demanding with regard to the initial story splitting which I observe to be one of the hardest things to learn for a team.

    1. Agreed Rainer.
      This comes back to the old XP proverb: "If it hurts, do it more often".

      For teams that have issues with slicing, I've put some slides together (in the download section), although I agree slicing definitely needs some experimental learning.