Framework, Method, or Practice? The Snowden Distinction

Everyone sells a framework. Scrum calls itself “a lightweight framework.” SAFe and LeSS do the same. AME3 describes itself as a meta-framework. The word has become a badge that every method wants to wear.

Dave Snowden, the creator of Cynefin, uses the word far more strictly. Under his definition, Cynefin is a framework and Scrum is not. This is not word games. The distinction changes how you choose your methods, how you read your own organization, and what you should expect a “framework” to do for you. It also clarifies what AME3 actually is.

What Snowden Means by “Framework”

For Snowden, a framework does not tell you what to do. It helps you make sense of your situation so that you can decide what to do. It sits above your methods and points you at the right one. The community wiki that his organization maintains states it plainly:

Cynefin is at its heart a decision support framework, not a method or model. It is a way of determining what method or approach you should adopt and, critically, when you should change it.

Notice the two jobs in that sentence: choose the method, and know when to switch. A framework in this sense is domain-independent. It works whether you are building software, running a hospital, or planning a military operation, because it operates on the nature of the problem, not the content of the solution.

The most cited source carries the word in its title. Snowden and Boone published A Leader’s Framework for Decision Making in Harvard Business Review in 2007. The framework sorts a leader’s situation into contexts defined by the relationship between cause and effect, so that the leader can diagnose the context before acting. Diagnosis first, action second. That ordering is the whole point.

Framework Is Not a Model

The cleanest way to understand a framework is to contrast it with a categorization model. Such a model puts its categories first: they exist before you arrive, and you sort your data into them. A sense-making framework works the other way: the meaning emerges from your data. Snowden compressed this into a single antithesis in his 2010 series The Origins of Cynefin:

a sense-making framework (data precedes framework) and a categorisation model (framework precedes data).

In a model, the boxes are drawn before you arrive. The familiar two-by-two matrix is the archetype: four quadrants, an implied “good” corner, and your job is to locate yourself and move up and to the right. Snowden made exactly this critique in the original 2003 paper that introduced Cynefin:

We make a strong distinction here between sense-making frameworks and categorization frameworks.

He resists calling Cynefin a model for this reason. In The Origins of Cynefin he goes further and calls it a meta-model: “it’s not a model, it’s a framework but really it’s a meta-model, a way of understanding models and approaches.” The Stacey Matrix is a useful companion here: its own author disowned the diagram precisely because people read a sense-making tool as a categorization grid.

Cynefin names its domains, which invites a fair objection: is that not a grid after all? The difference is in the using. You are not placed in a domain in advance. You work out which one you are in from the situation, the boundaries are drawn from the evidence, and no domain is the desirable corner to move toward. The categories serve the sense-making rather than precede it.

A note on attribution, since the elegant “data precedes framework” line travels widely: it comes from Snowden’s 2010 blog series, not from the 2003 paper. The 2003 paper makes the same distinction in different words.

Framework Is Not a Method

If a framework helps you choose, a method is one of the things it chooses. A method is a domain-specific sequence of activities: a recipe you follow to get a repeatable result. Snowden is openly hostile to the inflation of methods into methodologies. In his 2021 series Re-wilding, he argues for a toolkit of methods broken into recombinable components rather than “the overwhelming and highly energy inefficient approach of an overarching framework or methodology.” His charge against packaged methods that overreach is precise: they are “contextually appropriate codification which then makes the mistake of claiming universality.”

His sharpest image for the difference is the recipe and the chef:

In universities we are training recipe book users and assessing whether they can reproduce the recipe. We are not training chefs who can achieve a huge amount without a recipe.

A recipe is a method you reproduce. A chef exercises the judgment that tells you which recipe fits tonight’s ingredients, or whether to abandon the recipe entirely. The framework is what makes you a chef. The framework sits above the methods and chooses among them.

Practice: What Each Domain Demands

Underneath methods sit practices: the things people actually do. Snowden’s key insight is that the kind of practice that is valid depends on the domain you are in. The same framework that classifies your context also tells you what kind of practice to trust.

“Best practice” is only legitimate in the Clear domain, where cause and effect are obvious and repeatable. Push it into the Complex domain and it becomes a trap, because a complex system does not repeat. This is why a method built for one domain fails when it claims universality across all of them.

Where does Scrum sit? On the reading Cynefin invites, in the liminal band between Complex and Complicated: the boundary zone where a domain is shifting from one to the next. Scrum is an iterative technique that takes a complex problem and drives it, increment by increment, toward good practice. That is a real and valuable job. It is also a specific one. Scrum is a practice-bearing method for a particular kind of terrain, not a device that tells you which terrain you are on.

The Taxonomy at a Glance

So, Is Scrum a Framework?

There are two honest answers, because the word carries two meanings.

In the everyday sense, a framework is a supporting structure: scaffolding you build within. Choose Your Own Game uses exactly this image, a framework as a supporting structure and not a ready-to-move-in house. By that meaning, Scrum is a framework, and the Scrum Guide calling it “a lightweight framework” is fair. It gives you beams and cornerstones and leaves the rooms for you to fill.

In Snowden’s strict sense, a framework is the meta-level chooser, and Scrum does not qualify. Scrum is a method: a bundle of practices tuned for a particular context. It does not tell you when not to use it. That is the missing job. Snowden’s critique of scaling frameworks lands on the same nerve. Writing about SAFe in 2014, he described what happens when a method is forced to be the whole model:

SCRUM as an approach was emasculated in a small box to the bottom right of a hugely overcomplicated linear model.

Scrum treated as an approach, boxed inside someone’s model of the world. Not the framework. The framework is whatever told you Scrum belonged there in the first place.

The distinction resolves into a single question each answers. Scrum answers: how do we work in this context? A framework answers: which way of working does this context call for?

In Practice: The Framework That Was Really a Method

Consider Marcus, a CTO at a mid-sized financial services company. Under pressure to move faster and to “do AI,” he did what many capable leaders do: he picked the method with the best track record and rolled it out everywhere. Every team would run Scrum. Product development, the regulatory reporting group, the mainframe maintenance crew, the platform team keeping the payment rails alive. One method, applied uniformly, called “our framework.”

It worked in some places and quietly failed in others. The new product teams thrived on short cycles and fast feedback. The regulatory group, whose work was well understood and audited against fixed rules, spent its Sprints performing ceremonies that produced no new learning, because there was nothing to learn: the work was Clear, and it needed best practice and flow, not empirical discovery. The maintenance crew, firefighting unpredictable incidents, found the two-week rhythm actively in the way. Morale dropped in exactly the teams that had been running smoothly before.

The mistake was not choosing Scrum. Scrum was right for some of those teams. The mistake was treating a method as a framework: assuming that one way of working was the answer for every context, and having no device to tell the contexts apart.

A framework would have asked the prior question. What kind of problem is each of these teams facing? Cynefin separates the Clear from the Complicated from the Complex, and each calls for a different practice. Read that way, the regulatory group wanted stable process, the maintenance crew wanted rapid response and flow, and only the product teams were in the complex terrain where Scrum earns its keep. The framework is what selects the method per context. Marcus had skipped it and reached straight for the method. (Choose Your Own Game walks through how to make that selection deliberately across the stages of a product’s evolution.)

The Word Has Moved

Step back from Snowden for a moment and admit a stubborn fact. In common usage, “framework” has drifted well past a supporting structure. To most people in the industry it now means the packaged answer you adopt to fix the organization: Scrum, SAFe, the recipe that promises to solve everything. That is how the word is used, and a well-sourced definition does not change it. Usage owns a word, not its footnotes.

Language does not hold still. It drifts with use, the way anything alive does. A word is a small evolving system, shaped by the people who use it and indifferent to what its originators intended. This is familiar ground for this book. Evolution is not something you argue with. Don’t fight evolution, lead it.

For a word, leading it means conceding the term and keeping the distinction anyway. Call Scrum a framework if that is the common tongue. Just do not forget the level above it, the one that tells you whether Scrum is the right recipe for the problem in front of you. Lose the word if you must. Do not lose the difference.

Where AME3 Lands

This is where the meta-framework claim earns its name. AME3 takes the popular word and does the strict-sense job underneath. It does not prescribe a single way of working. The “M” stands for meta-framework: a framework that integrates other frameworks and methods rather than replacing them. Under Snowden’s strict definition, that is precisely what makes AME3 a framework in his sense rather than a dressed-up method. It does the two jobs a framework must do: choose the method, and change it when reality shifts.

The mechanism is built into AME3’s design. The System Lead selects and adapts the methods inside an Arena, based on what the Ambition and the evolutionary stage of the Arena Product demand. Empirical Control is the sensing loop that reads the context before committing. The Tournament and the Match are the built-in points where you check whether the method you chose is still the right one, and switch if it is not. Sense the domain, choose the method, inspect, adapt. That is a framework doing its job. AME3 prescribes the sensing-and-switching mechanism, not the delivery method each Arena runs.

Scrum tells you how to play. Cynefin helps you read the game. AME3 turns that reading into method selection.

Connection to AME3

References

• Kurtz, C. F., & Snowden, D. J. (2003). The New Dynamics of Strategy: Sense-making in a Complex and Complicated World. IBM Systems Journal, 42(3), 462–483. DOI 10.1147/sj.423.0462
• Snowden, D. J., & Boone, M. E. (2007). A Leader’s Framework for Decision Making. Harvard Business Review, 85(11), 68–76. Reprint R0711C. hbr.org/2007/11/a-leaders-framework-for-decision-making
• Snowden, D. J. (2010). The Origins of Cynefin (blog series, Parts 4–6). cognitive-edge.com
• Snowden, D. J. (2014). SAFe: the infantilism of management. thecynefin.co
• Snowden, D. J. (2021). Re-wilding. thecynefin.co
• Cynefin community wiki: Cynefin, Cynefin Domains, The recipe book user and the chef. cynefin.io/wiki