In complex systems, we cannot reliably predict how change will unfold. Plans that assume clear relationships between cause and effect can easily produce unexpected consequences. Safe-to-fail probes offer a different approach: make several small interventions, observe what happens, and adapt our actions as the situation evolves.
In a complex system, change rarely works so neatly. There are too many interactions, relationships, and feedback loops for us to predict with confidence how the system will respond to an intervention. Cause and effect may only become clear in retrospect.
Safe-to-fail probes offer a different approach. Rather than committing to one large intervention based on what we think will happen, we make several small interventions in parallel and observe how the system responds.
Some probes may produce promising results. Others may fail or reveal unexpected consequences. The purpose is not simply to discover what works, but to learn more about the system and use that learning to decide what to do next.
Change becomes an iterative process of probing, observing, making sense of what happens, and adapting our actions accordingly.
Dave Snowden‘s concept of “safe-to-fail” is an approach to working with the inherent uncertainty and unpredictability of complex systems.
Traditional approaches to change often assume that if we analyse a situation carefully enough, we can determine what needs to be done, create a plan, and implement it. This works well enough where cause and effect are reasonably clear and predictable.
In complex situations, however, we cannot know with confidence what the consequences of an intervention will be. People adapt, circumstances change, relationships interact, and actions can have unintended consequences.
Rather than asking, “What is the right solution and how do we implement it?”, we can ask, “What small things might we try, what happens when we try them, and what can we learn from the response?”
Safe-to-fail probes are small-scale interventions designed to help us explore these possibilities. They are “safe” in the sense that if they do not work as hoped, the consequences are contained and we can stop or modify them before serious harm is done.
Importantly, safe-to-fail does not simply mean starting small with a solution we have already chosen. In a genuinely complex situation, we do not know what will work. We therefore run several different probes in parallel, sometimes taking quite different approaches.
We then pay close attention to what happens. Where desirable patterns emerge, we can amplify them. Where undesirable patterns appear, we can dampen them or stop the probe. Unexpected outcomes are particularly valuable because they may reveal something about the system that we had not previously understood.
The process is therefore less about implementing a predetermined plan and more about probing, sensing, responding, and continually adapting as the situation changes.
- Improving collaboration across teams
Several teams are struggling to work together, but it is unclear why. Rather than introducing a new collaboration process across the organization, try several small interventions in parallel. One group holds informal cross-team conversations, another experiments with temporary staff exchanges, and another changes how joint decisions are made. Observe what happens and build on whatever seems to improve relationships and cooperation. - Encouraging people to speak more openly
People seem reluctant to raise concerns or challenge decisions, and the reasons are unclear. Try several small changes: anonymous questions in one meeting, a senior person speaking last in another, and small peer conversations elsewhere. Watch for changes in what people are willing to say and use what emerges to decide what to try next. - Responding to falling community participation
Participation in a local community has been declining, but there is no obvious explanation. Instead of designing a major engagement programme, try a few small initiatives: informal gatherings, conversations with people who have stopped participating, and giving members control of a small project. Look for unexpected responses and emerging patterns rather than judging everything against a predetermined solution. - Exploring the introduction of AI at work
An organization knows AI will affect how people work but cannot predict exactly how. Rather than developing a comprehensive AI transformation plan, allow several small groups to explore different uses within clear constraints. Bring them together regularly to share what they are discovering, including failures and surprises. Amplify useful practices and stop those creating problems.
A safe-to-fail probe is not simply a small-scale pilot of a solution we have already chosen. It is a way of exploring a complex situation when we cannot know in advance what will work.
Probe, sense and respond
Safe-to-fail probes change how we think about change itself.
Instead of trying to predict the future and then construct a detailed path towards it, we take small steps that allow us to learn more about the situation as it develops.
A simple way of thinking about the process is:
Probe → observe → make sense → amplify or dampen → probe again
This does not mean that we have no sense of direction or purpose. We may have a clear idea of what we hope to improve or what we want more or less of. What we cannot know in advance is precisely how to get there.
The path emerges through action and learning. Each intervention changes the situation to some degree, giving us new information and creating new possibilities. We respond to the situation as it now exists rather than continuing to follow a plan based on how we thought it would behave.
The term “experiment” can suggest a controlled process with a defined hypothesis and predetermined criteria for success or failure. A probe has a different emphasis. We intervene in a complex system partly to discover how it responds.
The purpose is therefore not simply to prove or disprove a hypothesis. It is to gain information, notice emerging patterns, discover unexpected consequences, and identify possibilities for further action.
The word “probe” also suggests a more humble stance towards complexity. We acknowledge that we cannot fully understand or control the system in advance. We try something small, pay attention to what happens, and adjust accordingly.
Safe-to-fail probes are therefore exploratory, iterative, and learning-oriented. They allow change to develop through repeated cycles of action, observation, and adaptation rather than through the implementation of a predetermined solution.
Safe-to-fail experiments | Jennifer Garvey Berger (source)
In complex situations, change cannot always be planned in advance. Safe-to-fail probes allow us to make several small interventions, observe how the system responds, and adapt as we learn. Rather than trying to predict and control change, we probe, sense, respond, and discover the next steps as we go.
Resources
- Safe to fail experiments
| NHS Improvement - Blog Post: Safe-fail probes by Dave Snowden
- Cynefin Wiki: Safe to fail probes
- Blog Post: Safe-to-Fail Requires Safety, Not Just Failure
Posts that link to this post
- Adaptive Funding for Innovation Diversifying investments to drive innovation
- Exploring Uncertainty – the Role of Catalytic and Safe-to-fail Probes Essential Tools for Understanding and Influencing Complex Systems
- Introduction: Rethinking Change Finding our way forward in a complex world
- Nurturing a Collaborative Culture Through Knowledge Cafés Using Knowledge Cafés to bridge silos and encourage open dialogue
- Nurturing an Innovation Culture Allowing emergence, not imposing change
- Safe-to-fail Conversations Using conversation as a probe when the way forward is uncertain
- The Birth of Uber Innovation Story: How everyday frustrations and collaboration shaped Uber’s global transformation
- The Gurteen KM Thinking Framework Knowledge Management
(Public Domain)Join one of my upcoming Knowledge Cafés or workshops and explore Conversational Leadership in practice.
Engage in thoughtful conversation, challenge assumptions, and connect with others working in knowledge, leadership, and organisational change.