by Matthias Schmid
I am passionate about lean agile development processes, because I saw several failed projects in the past doing scrum.
A few examples about Scrum gone wrong:
- Miss the sprint goal every time: Inflation of story point values
- Translate story points into time: developers doing the work of the project manager
- Translate story points into money: developers doing the work of the purchase department
- Scaled agile framework (SAFe): justification that the business does not need to change anything
Thankfully you can be agile (=flexible) without scrum, for example by doing Extreme Programming.
Inspired by Allen Holub, I'd like to discuss a much more lean process that does not even need estimates:
Feedback from the session was that indeed some companies do scrum successfully and are happy with it. Scrum and even more #NoEstimates only work when the whole company is agile, especially startups.
A story point should resemble complexity, not a time or money equivalent.
- Split larger stories into smaller stories
- Do the most important story first
- Reduce scope
- Lesser important stories won’t make it into the product
- You must not live scrum by the book, you can customize
- Sprints help some teams to get accountability of the team
- Sprints help some teams to split stories to fit into 2-week sprints.
- Story points should not be visible to the customer
- You can get a better understanding of a story with Planning poker
- Estimating and splitting stories helps to gain understanding
- For larger projects, do a sprint commitment for each subteam, not just an overall commitment
- Teams shall not exceed ~8 people, 20 people are too much!
- Scrum with a sprint length of one day comes close to Extreme Programming.
- Some developers feel motivated by meeting the sprint commitment
- Sprint velocity can be used for salary negotiation / forecasting
- Switch to kanban if appropriate
To be honest, I was surprised, that scrum works for some companies, because I had several bad experiences with it. I was about that XP without estimates is a genius idea, however it may seem like a radical approach.