Cynefin Framework

Why it matters

Different kinds of problems need fundamentally different responses, and the most expensive mistake you can make is to misread which kind you’re facing. Some situations have a right answer you can look up; some have one an expert can work out; some have no knowable answer in advance and can only be felt out by experiment; and some are on fire and need action before understanding. Match your response to the kind of problem and things go well. Bring a checklist to a problem that has no checklist, and the very thoroughness of your planning becomes the thing that misleads you.

For example: a subscription business sees churn spike, and its growth team reaches for the playbook that has always worked — run the win-back email sequence, tune the discount, A/B test the landing page. Nothing moves. The reason isn’t a weak playbook; it’s that churn this time is emerging from something the playbook can’t touch — a shift in what customers value, rippling through word-of-mouth in ways no one can diagram in advance. This is not a problem you analyze your way through; it’s one you have to probe — try several small, different things, watch what the system does, and amplify whatever works. The team kept running the known play harder, and the apparent rigor of “we ran the standard process” hid that they were answering the wrong kind of question.

What it reveals. Which kind of situation you’re in, judged by how visible the link between cause and effect actually is — obvious, expert-discoverable, emergent-only-in-hindsight, or absent — and therefore which style of response will work.
How it changes the read. You stop asking “what’s the best practice here?” and start asking “what kind of problem is this — and does it even have a best practice, or do I have to experiment my way into one?”
When to foreground it. A proven playbook suddenly failing in a new context; a team split between “make a detailed plan” and “just run experiments”; experts confidently disagreeing; a crisis that’s made the old rules useless.
What you’d miss without it. The signature, costly error of treating a complex situation as merely complicated — pouring in analysis and planning that produce precise-looking forecasts a fundamentally emergent system will ignore.
Where it misleads. The domains aren’t fixed labels on situations — a stable, simple operation can collapse into chaos through complacency, and boundaries shift under you; and it’s tempting to file a problem under whichever domain your favorite tools happen to fit rather than where it actually sits.

How it works

Picture a manager who has been, by every measure, excellent. She ran a warehouse, and she ran it beautifully: clear procedures, the right checklist for each situation, best practices applied with discipline. Promote her to lead a fast-growing product organization and watch the same instincts quietly fail. She asks for a detailed plan and a forecast; she benchmarks the best practice; she analyzes. And the organization keeps surprising her — a reorg that should have boosted morale tanks it, a feature everyone loved lands with a thud, a rumor reshapes the team’s behavior overnight. She is working harder and reading the situation more carefully than ever, and it is not helping. The problem is not her competence. It is that she is treating one kind of problem as if it were another kind entirely.

The insight that untangles this comes from David Snowden, who argued that situations sort into a handful of fundamentally different kinds, distinguished by one thing: how visible the relationship between cause and effect is. In the clear kind, cause and effect are obvious to anyone — a misconfigured setting, a routine billing question — and the right move is simply to recognize the pattern and apply the known best practice. In the complicated kind, the link is real but hidden, discoverable only with expertise — a strange engine fault, a thorny legal question — and the right move is to bring in an expert to analyze before acting; there may be several good answers, not one. In the complex kind, cause and effect can only be connected in hindsight — an organization’s culture, a market’s response to something genuinely new, a cascade through a living system — and here analysis-before-action is a trap, because the connections don’t exist to be found yet; you have to run small, safe experiments, see how the system responds, and amplify what works. And in the chaotic kind, there is no discernible link at all — an active breach, a disaster unfolding — and the only right move is to act first to staunch the bleeding and create some stability, then sense and respond once you’re no longer in free fall.

The reason this is worth more than a tidy taxonomy is the specific, predictable, expensive mistake it catches: treating a complex problem as if it were merely complicated. The two feel similar from the inside — both are hard, both resist the obvious answer — but they demand opposite responses. The complicated problem rewards more analysis and a careful plan; the complex problem punishes them, because the careful plan produces a comforting illusion of precision while the real, emergent dynamics carry on generating outcomes no plan anticipated. Our manager keeps commissioning analyses and plans for a complex organization because that is what worked in her complicated and clear past — and the better and more thorough her plans look, the more confidently they point the wrong way. The escape isn’t to plan harder; it’s to recognize the domain and switch modes: probe, sense, respond.

Snowden gave the framework a Welsh word, Cynefin (roughly kuh-NEV-in) — a word for the place of your multiple belongings, the idea that we are shaped by many contexts we’re only half aware of. There’s a fifth region too, disorder: the state of not yet knowing which of the four kinds you’re in, which is where most people actually start and where the real risk lives, because from disorder we default to the response style we’re most comfortable with rather than the one the situation needs. The whole discipline, then, is a single honest question asked before you reach for any tool: what kind of problem is this, really — and does it have an answer I can find, an answer I can only grow, or no answer at all until I act?

Framework & implementation

Origin and evidence

The framework is David Snowden’s, developed at IBM in the early 2000s. Its foundational statement is Kurtz & Snowden’s “The New Dynamics of Strategy: Sense-Making in a Complex and Complicated World” (IBM Systems Journal, 2003), which sets out the domains and the sense-making stance; Snowden’s “Complex Acts of Knowing” (Journal of Knowledge Management, 2002) develops the epistemological underpinnings (the distinction between the knowable and the genuinely complex). The framework reached a wide audience through Snowden & Boone’s “A Leader’s Framework for Decision Making” (Harvard Business Review, 2007), the canonical popular treatment with its leadership case studies. Cynefin draws on complexity science (complex adaptive systems, emergence, the limits of prediction in living systems) and on naturalistic decision-making research; it is a sense-making framework rather than a predictive model — its claim is not that it forecasts outcomes but that it tells you which mode of engagement an outcome’s underlying causal structure permits. The word itself is Welsh, chosen by Snowden (a Welshman) for its sense of the multiple, half-conscious contexts that shape us.

Applications and common uses

Cynefin is a working tool wherever the first question is what kind of problem is this?

Leadership and management. The native use: recognizing when a situation calls for best-practice execution, expert analysis, emergent experimentation, or crisis action — and not running the wrong one of the four.
Software incidents and operations. Triaging an outage by cause-effect visibility — a known misconfiguration (clear), a subtle regression needing an expert (complicated), an emergent cascade across services (complex), an active breach (chaotic, act first).
Strategy and product. Telling apart markets that reward detailed planning from genuinely new categories that reward small probes and fast learning — and resisting the board’s demand for a confident plan where only experiments will do.
Crisis response. Recognizing the chaotic domain and acting to establish stability before attempting diagnosis, then moving deliberately into the complex domain where structured response can begin.
Orientation in an unfamiliar domain. Reading which parts of a new field are settled and rule-governed versus genuinely emergent and contested — the use that brings it into Terrain Mapping.

In every case the payoff is the same: the response style is matched to the situation’s actual causal structure, and the costly default — analyzing and planning an emergent problem into a false sense of control — is caught before it sets the strategy.

Failure modes and when not to use it

The lens’s characteristic ways of going wrong are catalogued in its Common Failure Modes:

Complex-as-Complicated. Applying expert analysis and detailed planning to an emergent system. The tell: the planning produces apparent precision that does not predict outcomes. Switch to probe-sense-respond, and expect the right response to feel less rigorous than the situation seems to demand.
Domain-comfort selection. Defaulting to the domain whose response style you prefer, regardless of the situation’s actual cause-effect visibility. The tell: the classification conveniently matches the toolkit you already have. Classify by the situation, not by the available tools.
Frozen-in-Disorder. Paralysis because the situation won’t classify cleanly. The tell: no response is initiated at all. Break the situation into parts, classify each, and move on a tentative classification while watching for evidence you got it wrong.

When not to reach for it. When the situation is unambiguously of one kind — a plainly routine task, a plainly technical fault — the four-domain apparatus is overhead; just do the obvious thing. Cynefin is a categorization and response-selection aid, not a solution: it tells you which mode of engagement fits, but it does not execute the probe, perform the expert analysis, or run the crisis response — that work still has to be done. And the domains are not permanent labels; treating a classification as fixed misses the boundary transitions (complacency tipping clear into chaotic) that are often the most important thing the framework has to say.

Terrain Mapping — the analysis this lens rides in; surveys an unfamiliar domain and tags each concept known/contested/open, where Cynefin reads what kind of cause-and-effect each region has.
Pareto Principle — the sibling orientation lens from the same territory: where Cynefin classifies the kind of terrain, Pareto finds the vital few concepts that orient you fastest within it.
The Map is Not the Territory — the companion always-loaded model: a reminder that the domain classification is itself a map, to be checked against how the situation actually behaves.
OODA Loop — the kindred orientation idea: Boyd’s “orient” step is the same move as reading which kind of situation you’re in before you act.