Closing the Loop

As developers we should be experts in feedback: we want to learn about the consequences of our code changes as quickly as possible. For instance:

  • The editor can tell if the new syntax is correct.
  • Unit tests can tell if basic logic and interactions are as expected.
  • Integration and Acceptance tests can tell how the change will affect the rest of the system.
  • Continuous integration can tell if changes coming from different developers will work together.
  • Build pipeline can tell if the changed system can be deployed and run after each change.
  • Production monitoring can tell if the system is healthy and alert us otherwise.

Those are great techniques for building things right, but they don’t tell if we are building the right things.

For that we should be trying harder to close the feedback loop of Planning, Building, Measuring and Learning. So far I’ve noticed the following approaches have the tendency to help:

  • Make sure requirements are created from meaningful conversations (Impact Mapping)
  • Include relevant data gathering into each feature (How to Measure Anything) and make business key performance indicators (KPI) part of the system monitoring
  • Use Retrospectives to reflect on the data collected and general KPIs.
  • Question any feature about to be built if it doesn’t carry a clear purpose.

Unfortunately the most important feedback loop in software development is not purely technical, thus developers should have a more active role trying to make sure the loop is always closed.

I’d be interest to hear what other people are trying on that front.

255 Words

2014-12-01 18:00 -0200