Feedback loops and learning

Introduction
One of the main reasons for trying to have short immediate feedback loops in what we do (not only in development) is for the increased opportunity of learning. Other important points are risk and waste, but this post is about learning.

Humans learn best when we get immediate feedback and results while we are performing a task. On the other side, the longer it takes from we act until we get a response, the less we are able to learn from the experience. Another important aspect of learning, is that we learn better when we can isolate and learn small steps at a time. Lastly we learn more from errors and failure than from success.

Feedback lag time impact
The first point with learning being affected by time from action to result is directly related to the feedback loop and why it is important to get the feedback loops as short as possible to ensure as little time as possible passes from an assumptions was made, an action or a decision taken on those assumptions and the validation, result and feedback on these. In development of software there are a numerous set of assumptions and work that goes into any feature, from those who define the needs, to those who implement and test these new features.

Baby step learning
The larger the batch of change and action that is performed before result and feedback, the less we are likely to learn from the work. This again plays into why it is important to try to minimize the feedback loops, so we can learn in smaller steps.

Safe to fail, learn by error
To understand why the last point of learning through failure and error is related to short feedback loops, one has to take into consideration waste and risk. In complex problem domains it may be a better strategy to allow for  safe failure rather than trying to ensure fail safety. With long iterations and long feedback loops, the waste and risk of failure becomes too high and failure becomes unacceptable. On the other hand if the risk and waste involved in a failure is small, it can more easily be tolerated and even encouraging some experimentation may be possible. Also since the failure is likely to be smaller, the impact low, it is easier to be open about such failures and have these communicated to a broader set of people in the organization for the learning to be distributed.