Safe to fail probes

Jump to navigation Jump to search

A complex system has no repeating relationships between cause and effect, is highly sensitive to small interventions and cannot be determined by outcome based targets, hence the need for experimentation; hence when dealing with complex systems there is the need for experimentation. Safe-fail Probes are small-scale experiments that approach issues from different angles, in small and safe-to-fail ways, the intent of which is to approach issues in small, contained ways to allow emergent possibilities to become more visible. The emphasis, then, is not on ensuring success or avoiding failure, but in allowing ideas that are not useful to fail in small, contained and tolerable ways. The ideas that do produce observable benefits can then be adopted and amplified when the complex system has shown the appropriate response to its stimulus. Where systems and the environments in which they exist become increasingly complex, what is known and what can be planned for becomes less certain – introducing and increasing organisational tolerance for failure is more crucial than ever.

Typical Uses

  • Safe-fail probes are usually designed to follow Model Creation by social construction.
  • Making decisions in situations of high uncertainty.
  • As an alternative to traditional strategic planning which place excessive emphasis on ideal future states.
  • Stimulating starting conditions for innovation.
  • Engender support and consensus for new initiatives to contribute to meaningful change.
  • They are always run in parallel around all coherent hypothesises


This activity can be carried out either with individuals, in pairs or in small groups. It is best carried out in small workgroup with relative sizes of three to four participants. The larger the number of groups and participants, the greater the variety of perspectives and iterations. If working with groups of people, carefully review the participants of your workshop to optimise cross-silo grouping of members. This will help produce the tacit outcome of promoting engagement and cooperation amongst people from different departments, ranks of authority and/or organisations

Have organised working spaces for each group; ideally working in circles with writing surfaces

Set some ground rules for the participants – on top of the specifications and guidelines laid out in the worksheet, it would also be helpful to create a simple guideline for people to assess their own proposals on. (eg. this could be as simple as agreeing to consider any idea that has a remote possibility of creating improvement; or to consider ideas that can only be carried out within a certain time period). This is not critical, but provides a simple guideline and consistent position to adopt when considering proposals for experiments.

Things You'll Need

Worksheets with certain guidelines and areas of concern can be prepared. These help give heuristics to the type of experiments participants should be thinking about. (e.g. What the probe is? Rationale behind the probe? How to put the experiment into action? How to observe success of experiment?) As these experiments need to be practical and actionable, specific guidelines should also be worked out and specified - such as the time-frame and budgetary limitations that the experiment should work within.

Examples of Actions Forms are at the links below.

- Action Form - Chaotic

- Action Form - Complex

- Action From - Complicated

- Action Form - Simple

Dos and Don'ts

  • Do encourage thinking outside of traditional boundaries. Safe-fail probes require and enhance the ability for exploration.
  • Do give interesting examples of oblique solutions that illustrate the importance of not thinking in direct, causal relationships.
  • Do make use of Safe-fail probes to promote cross-silo and cross-rank collaboration and sensitisation.
  • Do not give direct examples relating directly to the organisation, as this has the effect of patterning how participants start thinking about their problem. Oblique examples from different industries are best for conveying the principle behind this method.
  • Don't get involved with the content in any way, your views are not a part of this process.


Pre-stage? Link to other page?

Identify target areas - which (constraints?) have potential to experiment with, to learn about and shift the system?

  1. What can I change?
  2. (out of that set) Where can I monitor the impact of change?
  3. (out of that set) Where can I readily amplify success or dampen failure?
Design a portfolio of Safe-to-Fail probes

Choose (constraint/issue?) and design a portfolio of probes you could run in parallel to learn and perhaps shift the system with.

  • Think small & local
  • Think “What might happen if …?”
  • Think about radically repurposing a current product/skill for a new context
  • Think about tweaking constraints and “rules”
  • What hasn’t worked in the past but might work now?
  • Maximize variety i.e. many contradictory experiments at the same time – scan broadly.
  • Probes aim to provide insight and test business hypotheses, not provide 'solutions'. If some probes do not fail, you’re playing it too safe.
  • Design the experiment so that even if it fails you learn something about the system.
  • The definition of what is “safe-to-fail” depends on your specific context
  • Minimal interventions – small, local & recoverable - ensure survivability
Peer consulting - in trios, take turns to present your probe ideas to others for feedback. Provide each other with feedback and suggestions. Think about things like:
  • Bounded applicability - are the methods chosen appropriate to the context?
  • Are they applying complex facilitation principles?
  • What could be potential unintended consequences?
  • Are they considering dynamics and constraints?

In a virtual environment

The method has been conducted in virtual settings with success over a range of platforms.


Link to other articles on this wiki if they are relevant.


Specific articles can be referenced here

Blog posts

Link with commentary


Link to case articles here or third party material

DJS NOTE - the current examples in slides are not PARALLEL they are single