Beyond the two week iteration
On the nature of software delivery management
Continuous improvement over agile methodologies
The two-week iteration cycle is not good enough
25 releases
Few releases mean
- Frustration
- Stress
- Gaming requirements
- Release trifle
- Many features, one release
- QA headaches
- Big consequences if a release slips
- If one thing slips, everything slips
- Value realisation depends on your weakest link
What is the magical quality of the fortnight?
Dual-value stream fallacy
Would you take two-weeks to fix a bug in production?
Can you help us?
Yes
Let's do it now
Continuous Delivery?
continuous delivery
simon hildrew
Release when you want to
Release when you have to
My experience at the Guardian
Project one: Java monolith
- Core to the business
- Manually deployed once a fortnight
- Unloved and scary
We can't do it
- We can't develop fast enough
- We can't coordinate changes without time
- Many releases are riskier than fewer
- We can't deploy without disruption
- Users don't like inconsistency
- Partial features have no value
- What if something goes wrong?
We can do it
- We can build tools to control our deployments
- We can change the way we work
- We can separate releases and deployments
- We can deliver value faster
- Everyone can chill out
The Plan
- Fortnightly
- Weekly
- Daily
- AM/PM
- As we need
The most important thing we've done this year
Project two: dating site
- Three big milestones due
- Early delivery == additional revenue
- A background of an 18 month delivery
We can't do
- Users need consistency
- Analytics needs consistency
- No-one knows how to do it
- The business doesn't get it
We can do it
- Feature switches
- Audience segmentation
- Component-based transition
- Page-based transition
We did it
But it was inefficient
I didn't believe we could do it
But this team is great
My inexorable rise to greatness
We can't do it
- We can't release on Friday
- [Do not merge] pull requests
- We can't support our changes
Top change tips
Organisations don't change unless they are in crisis
People fear loss (or blame) more than they value reward
Getting people to change
- Articulate a vision
- Rewards need to be shared
- Benefits need to be realised
- Momentum must be preserved
Attitudes may change...
but automation is forever
Reconsider the release
- Deployment
- Release
- Value
- Feature
- Pull request
- Change
In conclusion
Two-week iterations are not good enough
Release when you want to
Release when you have to
Going beyond the two-week iteration means understanding your business
Only you can do this
You can do it
Please don't be afraid
Thank you
- Twitter/Github: rrees
- Blog: rrees.me
- developers.theguardian.com
- www.theguardian.com/workforus
Beyond the two-week iteration
By Robert Rees
Beyond the two-week iteration
- 998