Decision-making & prioritisation
Cynefin Framework
A sense-making framework that sorts situations into five domains by the relationship between cause and effect, so leaders can match how they decide to the kind of problem in front of them rather than to the kind they are most comfortable managing.
Also known as Cynefin sense-making framework, Snowden's five domains. First set out by Dave Snowden, developed with Cynthia F. Kurtz and popularised with Mary E. Boone in 1999; the primary source is cited in full below.
Where this is contested
Attribution to Snowden is secure, but the framework has been a moving target: the 2003 paper named the ordered domains Known and Knowable, Snowden and Boone (2007) renamed them Simple and Complicated, Simple became Obvious around 2014 and then Clear from about 2020, and the central domain Disorder is now Confusion. Kurtz's role in the 2003 formulation is sometimes under-credited in later accounts.
- Format
- Structural model
- Level
- Team · Business unit · Corporate
- Best for
- Evaluate options · Assess risk
- Decision stage
- Diagnose · Decide
- Difficulty
- Advanced
- Time to apply
- An hour to place a live problem with a leadership team; sustained value comes from changing how the organisation triages work over months.
Plate · The model
The components
Confusion
The central state of not knowing which domain applies. Formerly called Disorder, it is where most contested decisions actually sit, and its danger is that people resolve it by defaulting to the domain of their own comfort rather than by examining the situation.
Signals of strength
Contradictory advice from equally credible people · Arguments about method that masquerade as arguments about substance · Leaders retreating to their preferred style: process people demand procedure, experts demand analysis, commanders demand action · The situation has been categorised quickly and nobody can say on what evidence
Clear
The domain of best practice, formerly Simple and then Obvious. Cause and effect are evident to any reasonable observer, the right answer exists, and the decision model is sense, categorise, respond. This is where standardisation, checklists and automation belong.
Signals of strength
Stable, repeatable situations with low variation · The correct response is undisputed once the situation is identified · Work can be proceduralised and delegated safely · Efficiency, not judgement, is the constraint
Complicated
The domain of good practice and experts. Cause and effect exist but require analysis or expertise to see, several legitimate answers may be available, and the decision model is sense, analyse, respond. The risk is entrained thinking: experts defending the analysis they were trained to produce.
Signals of strength
Diagnosis requires investigation but converges with expertise · More than one defensible solution, with real trade-offs between them · Predictability is good once the analysis is done · Experts disagree on the best route, not on whether a route exists
Complex
The domain of emergent practice. Cause and effect can only be perceived in retrospect, because the system's many interacting agents change it as they act. The decision model is probe, sense, respond: run safe-to-fail experiments, amplify what works, damp what does not. Analysis alone cannot get ahead of the system.
Signals of strength
Plausible retrospective explanations that fail to predict the next occurrence · Many interacting agents, feedback loops and unintended consequences · Expert analyses that keep finding different root causes · Small interventions sometimes produce large effects, and vice versa
Chaotic
The domain of novel practice. No cause-and-effect relationship is perceivable, events are in freefall, and searching for right answers is a category error. The decision model is act, sense, respond: stabilise first, establish order, then work out where you are. Chaos is also an opportunity for decisive innovation, but nobody should stay here long.
Signals of strength
A genuine emergency with no time for analysis · High turbulence, broken communication, decisions needed immediately · Acting to establish any stability beats optimising the action · The priority is stemming the crisis, then moving the situation to another domain
When it earns its keep
- You need to decide how to approach a problem before deciding what to do about it, and suspect the default method is a matter of habit rather than fit.
- Best-practice programmes keep failing against problems like culture change or market transformation, and you want language for why analysis and planning are not working.
- You run a mixed portfolio of work, some of which should be standardised and automated while some needs expertise and some needs experimentation, and the organisation currently treats it all the same way.
- You are designing incident or crisis response and need to distinguish situations that demand immediate stabilising action from those that reward patient probing.
And when it doesn't
- You want a static classification of your organisation's problems to file and forget. Snowden is emphatic that Cynefin is a sense-making device, and situations move between domains.
- The decision itself is small and well understood. Convening a domain discussion for a routine choice is ceremony, and Clear-domain work needs no framework to tell you so.
- You need to prioritise or score options within a single domain. Cynefin tells you what kind of response fits; it does not rank candidate responses against each other.
- You are tempted to use it as a maturity model in which everything should migrate towards Clear. Complexity is a property of the situation, and treating simplification as progress is exactly the complacency the framework warns about.
How to run it
Before starting, gather the inputs the analysis depends on:
- A concrete description of the situation, broken into parts if necessary, since one programme usually spans several domains at once.
- Evidence about the cause-and-effect relationship: can outcomes be predicted, can experts converge through analysis, or does coherence only appear in hindsight?
- Multiple perspectives in the room, because individuals reliably place situations in the domain their training prefers.
- For anything provisionally complex, the capacity to run genuinely safe-to-fail experiments, including permission for some of them to fail.
- Honesty about what is actually known versus what is assumed because it has always been assumed.
- 1
Start from Confusion
Assume you do not yet know which domain applies; that state is the centre of the framework, and the danger is leaving it by reflex rather than by evidence. Break the situation into components, since a single initiative typically contains Clear, Complicated and Complex strands that need different handling.
- 2
Test the cause-and-effect relationship
For each strand ask what kind of knowability you face. If the link between action and outcome is self-evident to any competent observer, it is Clear. If it can be established by analysis or expertise, Complicated. If it can only be perceived in retrospect, Complex. If no link is perceivable and events are in freefall, Chaotic.
- 3
Apply the domain's decision model
Clear: sense, categorise, respond with best practice. Complicated: sense, analyse, respond with good practice chosen from several legitimate options. Complex: probe with safe-to-fail experiments, sense what emerges, respond by amplifying what works and damping what does not. Chaotic: act to stabilise first, then sense, then respond.
- 4
Manage the boundaries
The boundaries matter as much as the domains. The most dangerous is the cliff between Clear and Chaotic: entrenched best practice breeds complacency, and systems that assume order fail catastrophically rather than gracefully. Ask which of your stable processes are one surprise away from freefall.
- 5
Watch for movement
Situations migrate. A complex problem yields patterns that become complicated expertise and eventually clear procedure; a chaotic crisis, once stabilised, becomes complicated forensics. Revisit placements as evidence arrives, and treat disagreement about placement as information rather than noise.
Reading the result
A placement of each strand of a situation into one of the five domains, with the evidence for the placement, the matching decision model for each strand, and a watch-list of boundary risks, above all the cliff between complacent order and chaos.
- Placement prescribes method, not importance. A Clear item can be existential and a Complex one trivial; the domain tells you how to decide, never how much to care.
- Read the boundaries as seriously as the domains. The Clear-Chaotic boundary is a cliff, and the organisations that fall off it are the ones proudest of their best practice.
- Disagreement about where something sits is data. If your engineers call it Complicated and your frontline calls it Complex, run the cheap experiment before commissioning the expensive analysis.
A worked example
An IT managed services firm sorts its ticket queue and its crises
A UK managed services provider with 40 staff supports around 200 SME clients. After a bruising quarter that included a botched cloud migration and a ransomware incident at a client, the service director uses Cynefin to work out why the firm's runbook culture served some situations well and failed badly in others.
- Confusion
- The Monday-morning triage queue itself. An alert storm from a client network might be a failed switch, a misbehaving update or the first sign of an intrusion, yet the ticketing system forces a category within sixty seconds, so everything was being triaged as Clear by default. The fix is procedural humility: triage now asks 'do we know what this is?' before any runbook is assigned, and escalates genuine unknowns instead of guessing.
- Clear
- Password resets, licence renewals, patch schedules, starter and leaver processes. Cause and effect are known to everyone, best practice exists and the firm's margin depends on keeping this work proceduralised and increasingly automated. Correctly handled already; the framework's contribution is to stop Clear methods leaking into the other domains.
- Complicated
- Migrating a 60-seat client from on-premise Exchange to Microsoft 365. Analysable in advance by experts, with several defensible designs. The botched migration failed precisely here: a junior engineer treated it as Clear and followed a generic checklist that ignored the client's ageing line-of-business application. The work needed sense, analyse, respond, meaning a senior design review, not a runbook.
- Complex
- Persistent intermittent performance complaints at a hybrid-cloud client, where network, applications, user behaviour and two ISPs interact. Three months of root-cause analysis kept 'finding' causes that did not survive the following week, a classic sign of retrospective coherence. The firm switches to probe, sense, respond: small safe-to-fail changes such as a QoS adjustment at one site and staggered sync schedules, measured for a fortnight, amplified or reversed on evidence.
- Chaotic
- The ransomware outbreak at a client. No perceivable cause and effect in the moment and no time to find one. The correct model was act, sense, respond: isolate network segments, cut VPN links, invoke the incident plan, communicate on a fixed cadence. Once stabilised, the situation deliberately moved domains: forensics became Complicated work, and rebuilding the client's trust became Complex work.
The read. The firm's economics rely on pushing as much work as possible into Clear, but its reputation is made and lost in the other domains. The practical output is a triage question set that legitimises saying 'we do not know yet', reclassification of migrations as Complicated with mandatory senior design review, an experiment log for complex problems in place of endless root-cause hunts, and a rehearsed act-first protocol for chaos. The sharpest lesson is the cliff: the migration failure was not bad luck, it was Clear-domain complacency applied at the edge of the Chaotic boundary.
Pitfalls
- Using Cynefin as a categorisation matrix with four static boxes. Snowden distinguishes sense-making frameworks, where the data precedes the framework, from categorisation ones, and the central domain plus the boundary dynamics are the point.
- Treating Clear as the goal state and driving everything towards standardisation, which is how organisations walk off the cliff into chaos.
- Letting people place situations by professional preference: engineers find Complicated everywhere, agilists find Complex, and executives under pressure find Chaotic.
- Skipping Confusion and forcing an immediate placement, which simply launders the default habit through the framework's vocabulary.
- Running patient safe-to-fail experiments during a genuine emergency, or its mirror image, commanding decisive action against a complex problem that will absorb and punish it.
What the critics say
Independent applications report that domain placement is highly interpretive. Researchers who used the framework to analyse qualitative data found allocation between domains difficult and subjective, with fuzzy boundaries, which limits reproducibility when different analysts assess the same situation.
McLeod, J. and Childs, S. (2013) 'The Cynefin framework: A tool for analyzing qualitative data in information science?', Library & Information Science Research, 35(4), pp. 299-309.
The framework's development has been consultant-led, with repeated renaming of domains and limited peer-reviewed empirical validation beyond the founding papers. Supporters read the revisions as refinement; critics read them as a moving target that makes the framework hard to test or falsify.
Cynefin prescribes a decision style per domain but offers little method inside each. 'Run safe-to-fail experiments' or 'apply good practice' are headings rather than procedures, and practitioners need substantial additional apparatus, from experiment design to expert elicitation, to act on a placement.
A gap acknowledged in practice literature around Snowden, D. J. and Boone, M. E. (2007) 'A Leader's Framework for Decision Making', Harvard Business Review, November 2007.
Sources and further reading
- Kurtz, C. F. and Snowden, D. J. (2003) 'The New Dynamics of Strategy: Sense-making in a Complex and Complicated World', IBM Systems Journal, 42(3), pp. 462-483. ↗
- Snowden, D. J. and Boone, M. E. (2007) 'A Leader's Framework for Decision Making', Harvard Business Review, 85(11), November 2007. ↗
- Snowden, D. J. (2002) 'Complex Acts of Knowing: Paradox and Descriptive Self-awareness', Journal of Knowledge Management, 6(2), pp. 100-111.
- The Cynefin Co, 'Cynefin papers: a summary'. ↗