Learn is the last step of the SAFe for DevOps Health Radar, and in many ways it is the most important one. This is where we make the hard decisions about where to invest, where to stop, and how to continuously improve everything we do. In this video, I walk through what the Learn step involves and why it is the key to building the right thing right.
The Journey to Learn#
Before we reach the Learn step, we have gone through the entire continuous delivery pipeline. It starts with bright ideas from the customer or the business. We extract the hypothesis, put it into an epic hypothesis statement, collaborate and research to find the real customer need, architect the minimal amount needed, and break the epic into prioritized features on a backlog.
Then we develop user stories, commit code, build deployable artifacts, test end-to-end, deploy to staging, and deploy to production with the feature toggle off. We verify in production, monitor, and respond to incidents as a whole team. When the business says the time is right, we switch the feature toggle on. We stabilize, and then we measure whether our hypothesis holds.
Now, with all that measurement data, we enter the Learn step.
Pivot or Persevere#
With the knowledge gathered from the Measure step, we can make a pivot-or-persevere decision. Should we invest more or less in a feature or in an epic?
This is critical because one of the main problems in today’s software engineering is that we have so many bright ideas but not enough resources to implement them all. We need to identify the epics and features that bring the most business value and invest in those. Just because we have built something does not mean we need to keep it, maintain it, or invest more money into it. We need to ignore the sunk costs and focus on what brings the most value for our customers.
“Just because we have built it does not mean we need to keep a feature or invest more money into it. We need to ignore the sunk costs.”
This connects to the lean startup cycle that runs through the entire SAFe DevOps Health Radar. In the Hypothesize step, we created the hypothesis. In the Measure step, we gathered data. Now in the Learn step, we make the decision: is the hypothesis proven and we invest more, or is it disproven and we stop or pivot to a new epic?
Value Stream Mapping#
In the Learn step, we also analyze our value stream. We conduct a value stream mapping workshop. Here is how it works in short:
- Identify all steps needed to bring an idea from creation into production.
- Identify all people needed in each process step.
- Measure the process time: the actual value-adding work time in each step.
- Measure the lead time: the total elapsed time from one process step to the end of the next.
- Measure the percentage complete and accurate (%C&A): how often the output of a step can be used by the next step without rework.
With this data, you can clearly see your current value stream, identify bottlenecks, and focus on removing them. Next, you define your future value stream: where you want to be. You remove bottlenecks, streamline the flow, and eliminate unnecessary process steps. Then you identify the steps needed to get from current to future state.
Every three months, you revisit your value stream and assess how it has improved. This continuous analysis ensures ongoing progress.
Relentless Improvement#
The Learn step is about relentless improvement. We constantly maintain and improve our continuous delivery pipelines to enhance our ability to test hypotheses. This improvement happens through several practices:
Retrospectives: Every two weeks, the whole team holds a retrospective. What went well? What can we improve? The team creates backlog items for improvement actions and tackles them in the next sprint.
Value stream analysis: Continuously analyze the value stream, identify bottlenecks, go to the root cause, and resolve them systematically.
Incident post-mortems: With every incident, conduct a post-mortem analysis. Identify what can be done better, and continuously improve so that these incidents are prevented in the future.
The Maturity Levels#
The SAFe DevOps Health Radar provides a maturity assessment for the Learn step:
- Sit: Features are never evaluated post-release.
- Crawl: Features are sometimes evaluated using subjective information or unilateral opinions.
- Walk: Hypotheses are evaluated using objective measures, but actions are heavily influenced by corporate politics.
- Run: Hypotheses are objectively evaluated. Pivot-or-persevere decisions are made without mercy or guilt.
- Fly: Continuous learning and experimentation are ingrained in the DNA of the organization.
Where does your team fall on this scale? The goal is to move toward objective, data-driven decisions that are free from politics and sunk-cost thinking.
What Learn Produces#
The output of the Learn step is significant:
Continuous improvement: You get better and better with every iteration. Each sprint brings new insights and new improvements to your pipeline and processes.
A clear view of the value stream: Everyone working in the value stream understands how value flows through the system and where the bottlenecks are. This shared understanding aligns IT and business.
Hard decisions based on data: You have the information needed to decide whether to invest more in an epic or feature, or to stop. These decisions free up resources that can be redirected to the ideas that bring the most value.
Focus on the right things: By stopping work on invalidated hypotheses, you lower work in process and enable your teams to concentrate on what truly matters.
Key Takeaways#
- Make pivot-or-persevere decisions based on data. Ignore sunk costs and focus on what brings the most value to customers.
- Map your value stream regularly. Every three months, revisit the value stream to identify bottlenecks and track improvement progress.
- Practice relentless improvement. Hold retrospectives every sprint, analyze incidents with post-mortems, and continuously improve your delivery pipeline.
- Aim for objective evaluation. Move away from politics and subjective opinions toward data-driven decisions about features and epics.
- Learn completes the loop. The insights from Learn flow back into Continuous Exploration, creating a continuous cycle of hypothesis, build, measure, and learn.
- Free up resources by stopping. Not every idea deserves continued investment. Stopping work on disproven hypotheses is just as valuable as continuing work on proven ones.