Cynefin

Goal

Cynefin helps a group choose a way of acting by first establishing the type of system it faces. It points to the family of methods that suits a situation and leaves the choice of the precise method to its users.

The framework rests on the principle of bounded applicability: few solutions hold in every context, many hold in a specific context. A practice that succeeds in one domain can fail in another.

Kurtz and Snowden locate the framework's value in its effect on how its users make sense of a situation and decide. They call it phenomenological: it describes how people perceive a situation. Two groups can therefore place the same project differently. The group rebuilds the framework each time it uses it.

The technique produces a shared map: the elements of a situation placed in the domains, the boundaries the group has drawn and a common vocabulary to discuss them. The decision it supports concerns how to treat each part of the problem: apply a rule, consult an expert, experiment or act at once.

The word cynefin is Welsh. Its authors' wiki renders it as "a place of your multiple belongings".

Usage

When to use it

• Scoping an unclear initiative: separate the parts to specify from the parts to discover by experiment.
• Disagreement on how to run the project: detailed plan or iterations; placing the elements settles the question element by element.
• Before freezing requirements: spot the complex elements, where a specification goes stale before delivery.
• Analysing a failure: check whether the team handled a complex situation with methods made for order.
• Several management levels involved: comparing their maps reveals divergent readings of the same situation.
• Team spread over several sites: the group can run the contextualisation on a shared whiteboard if all participants work on it at the same time.

When not to use it

• Choice between options already known: Cynefin does not compare options; use a decision matrix or decision analysis.
• Crisis under way: act first under the organisation's crisis arrangements; run the contextualisation once stability returns.

Description

Order, complexity, chaos

Cynefin distinguishes three types of system. In an ordered system, constraints make outcomes predictable. In a complex system, constraints make action possible without producing linear causality. In a chaotic system, no constraint acts. Order divides into the clear domain and the complicated domain. Confusion occupies the centre.

The domains

Clear. According to Snowden and Boone, the right answer here is often self-evident and undisputed. Kurtz and Snowden consider structured techniques mandatory in this domain.

Complicated. A right answer exists; an expert or an analysis finds it. According to Snowden and Boone, several right answers can coexist. The analysis then compares several valid options. Kurtz and Snowden place the greatest danger of entrained patterns of thinking here: a single wrong assumption leads to a false conclusion that is hard to isolate.

Complex. The relation between cause and effect appears only in retrospect: the authors speak of retrospective coherence. Kurtz and Snowden warn that a procedure drawn from a stabilised pattern will then meet other patterns it is not prepared for. According to them, narrative techniques are particularly effective here. Storytelling is one of them.

Chaotic. The group acts first to restore a minimum of order, then senses the effect of its action. According to Kurtz and Snowden, applying a best practice probably precipitated the chaos.

Confusion. Kurtz and Snowden called this central domain disorder. It gathers the situations on which decision makers clash because they look at them from different points of view. Those who prefer stable order want to set rules. Experts want to study and gather data. Politicians want to widen their contacts. Dictators want absolute control. Each pulls the question towards the domain where they feel strongest. Snowden and Boone give the way out of confusion: break the situation into parts and place each one in one of the other four domains. By shrinking this domain together, the group moves closer to consensus. The wiki warns against staying in confusion without knowing it.

The names changed as the framework evolved. The 2003 paper speaks of known, knowable, complex, chaos and disorder. The 2007 article by Snowden and Boone calls the first domain simple. The names in use are those of The Cynefin Company's wiki in its February 2021 version: clear, complicated, complex, chaotic and confusion. The same wiki adds four liminal zones, where a situation passes from one domain to another.

A map without value axes

The Cynefin drawing looks like a four-box matrix. A 2×2 matrix carries two value axes. One of its boxes is the best. Cynefin has no axes: Kurtz and Snowden state that no domain is more desirable than another. The wiki lists the quadrant reading among the misconceptions about the framework. The group draws the boundaries from the elements it has placed.

Running the contextualisation

Kurtz and Snowden use the term contextualisation for the procedure in which a group derives the domains from its own examples.

1. Collect the elements. The group produces many elements on one theme through structured brainstorming. Each element is concrete: a community, an event, an anecdote, a belief. The facilitator varies the make-up of the groups to keep everyone from thinking along the same habitual lines.
2. Prepare, if time allows. A story base gathered beforehand feeds the collection. Kurtz and Snowden also cite the composite fable, which condenses several stories into one. Participants compose it from many anecdotes and a fable template. The group then draws elements from it: characters, events, situations. For alternative history, the group works back from the present to the past. It notes the turning points and what they could have produced. A turning point is a moment when a small event caused a large change.
3. Set the extremes. The group discusses the corners of the space before the domains: the corner where everyone knows the right answer, the one where an expert should know it, the one where the situation becomes clear only in retrospect, the one where no right answer exists. It first chooses the elements that best illustrate each extreme. These elements build the framework.
4. Place all the elements. The group places the other elements on the wall relative to the extremes without drawing any boundary. Kurtz and Snowden speak of dynamic placement: each corner pulls the elements towards it like an elastic band. Unlike the affinity map, where participants place their notes in silence, the group talks during placement.
5. Draw the boundaries. The group draws lines around the elements that clearly belong to a single domain. A central zone of confusion remains. It is sometimes large. The group then pulls the boundaries in towards the centre. It splits any element that straddles a line into two more precise elements. This step requires consensus.
6. Read each domain. The group examines the characteristics of each domain and each boundary for the situation under study. It then decides how to treat the elements of each.

Kurtz and Snowden expect two effects from contextualisation. The first is surprise: the group sees patterns that overturn its beliefs. The second: the group learns to describe its own situation in its own words. The sources set no duration, no group size and no reporting template.

The group runs the procedure in person with hexagonal sticky notes on a wall. Participants cluster the hexagons into honeycombs. Square notes push them to sort into categories. Kurtz and Snowden accept an electronic medium if the work stays collective. They require large power gaps to be reduced before the session: a participant who stays silent in front of their superiors hides the gap without reducing it. When the gap cannot be reduced, each management level runs its own contextualisation. The group then compares the maps.

Acting in the complex: safe-to-fail probes

In the complex domain, the group acts through safe-to-fail probes: small experiments whose failure stays bearable and which make emergent patterns visible.

The group launches several probes in parallel that rest on contradictory hypotheses. The wiki draws a rule from this: a portfolio in which no probe fails is too cautious. The group designs each probe to learn even from its failure. It amplifies what works and dampens what fails. Before launching a probe, it answers the wiki's questions: what can we change, where can we observe the effect, where can we amplify or dampen?

Movements between domains

A situation changes domain over time. According to the wiki, the most stable path starts in the complex: probes bring out a pattern, the group checks it in a liminal zone, then moves it into the complicated. The move to the clear comes last, when stability is proven and the change carries little risk. The wiki describes this move as difficult and costly to undo.

The wiki describes the boundary between the clear and the chaotic as a cliff: excessive confidence in rigid constraints makes the organisation fall from the clear into chaos without seeing it coming. The 2003 paper calls this fall collapse. The way back then requires acting before understanding.

Among the movements in the 2003 paper, imposition brings the chaotic back to the clear by force. Exploration leads from the complicated to the complex. Swarming leads from the chaotic to the complex.

What each domain changes for the analyst

The Cynefin sources address decision makers. The mapping between domains and analysis work is ours.

Mistakes that distort its use

Treating Cynefin as a method. The framework designates the type of response. The group must still choose and run a method: a prototype, an expert analysis, a procedure. The session ends when each element carries a method.

Placing once and for all. A map drawn up at project launch rarely still holds six months later. The group revisits it at each milestone and whenever an element surprises it.

Confusing complicated and complex. In everyday language the two words are interchangeable. In Cynefin, the complicated yields to analysis and the complex resists it. An analyst who treats a complex element as complicated commissions a study that predicts nothing: Kurtz and Snowden write that the methods of the ordered domains do not work in the complex.

Relying on a single expert. An analyst who hands every question to the same specialist inherits that specialist's assumptions. A second opinion or a review of the assumptions limits this risk.

Ignoring confusion. A group that believes the question is clear while each member pulls it towards their own domain stays in confusion without knowing it. The group exposes these disagreements by drawing the boundaries in session.

AI considerations

For collection, a language model extracts concrete elements from transcribed interviews, incident tickets or project reports: a dated event, an anecdote, a stated belief. It proposes a list for the group to review and flags duplicates. For the complex domain, it formulates opposing hypotheses on the same element that widen the probe portfolio beyond the team's own ideas.

Placement stays with the group. Cynefin describes a perception: an AI that sorts the elements into domains removes the discussion that produces the surprise and the common vocabulary. A language model also tends to propose generic best practices and thereby to pull every element towards order. The anecdotes contain names, conflicts and customer data. Before passing them to an online service, the analyst anonymises them in line with the Federal Act on Data Protection (FADP) and the organisation's rules.

Examples

A cantonal tax administration is preparing a new online tax return portal. A contextualisation session brings together the tax department, the cantonal IT office, two fiduciary firms and representatives of the municipalities. The figure shows the map at the end of the boundaries step.

The element "Abandoning the paper form" straddled a line. The group split it: the transfer of the paper form fields to the portal belongs to the clear, the users' move away from paper belongs to the complex. "Go-live date" stays in confusion: management sees it as clear, IT sees it as complicated. This disagreement opens the next session.

Visualisations

The group draws the map without axes or grid: each domain is a region whose boundary the group traces itself. The confusion zone stays visible until the group has placed each of its elements. The two halves of a split element carry the same mark. The facilitator photographs the wall after placement, then after the boundaries are drawn. When several management levels have each run their own contextualisation, the facilitator displays their maps side by side.

Cost

Tooling

A wall, hexagonal sticky notes and markers are enough: Kurtz and Snowden describe the technique on this medium. For a distributed group, a shared whiteboard such as Miro or Mural replaces the wall if the participants work on it at the same time. A spreadsheet holds the list of elements before the session and their domain after it. A shared folder is enough to hold the story base.

Sources

• C. F. Kurtz and D. J. Snowden, The new dynamics of strategy: Sense-making in a complex and complicated world, IBM Systems Journal 42(3), 2003, pp. 462-483: the reference description of the domains, the response sequences, contextualisation and the movements between domains.
• D. J. Snowden and M. E. Boone, A Leader's Framework for Decision Making, Harvard Business Review, November 2007: the domain names of the time, the signs that distinguish the simple domain from the complicated and the way out of confusion.
• The Cynefin Company, Cynefin.io wiki, page Cynefin: the origin of the framework, the principle of bounded applicability and the misconceptions about it.
• The Cynefin Company, Cynefin.io wiki, page Cynefin Domains: the current domain names, the three types of system and the cliff between the clear and the chaotic.
• The Cynefin Company, Cynefin.io wiki, page Cynefin Dynamics: the movements between domains and the most stable path.
• The Cynefin Company, Cynefin.io wiki, page Safe to fail probes: how to run safe-to-fail probes and the questions to ask before launching them.