How to use estuarine mapping to navigate change in complex systems

Estuarine mapping is Dave Snowden’s evolving technique for working out what can actually change in a complex system, and what can’t, so that action starts from the present reality rather than an idealised future. It maps the people, rules, habits and resources at play (actants, in the framework’s language) against how much energy each would take to shift and how quickly that shift could happen, then works towards a portfolio of small interventions rather than a single grand plan. Snowden presents it explicitly as a work in progress, and it’s been revised in public more than once since it first appeared.

Use it when a change effort keeps stalling against something nobody can quite name, when a strategy process has produced a vision nobody disputes and a route nobody can agree on, or when you suspect the organisation already knows what needs to shift and has simply never had a structured way to say so.

Why this, why now

Most strategic planning still assumes a stable enough world that you can set a destination and work backwards from it. That assumption breaks down more often than it holds, and Cynefin is one of the ways complexity practitioners have found for diagnosing when it does. What Cynefin has always left open is the next step. Once you know you’re in complex territory, where genuinely novel situations respond to probing rather than planning, what do you actually do on Monday morning?

Estuarine mapping is Snowden’s attempt at an answer, under development since around 2015 and published in “first edition” form in October 2022, a label he’s kept rather than dropped, since he’s continued revising the typology and extending the metaphor in the years after. It picks up roughly where Cynefin’s diagnosis leaves off, though it stands on its own as a method: a way of mapping what’s fixed, what’s fluid, and what’s within reach, before committing resource to changing any of it. One claim Snowden makes for it is scale independence: the same components, mapped the same way, are meant to work whether you’re looking at a corporate change programme, a software project, or a military campaign, which is a fairly unusual thing for a single method to claim and worth testing against your own work rather than taking on trust.

The estuary as metaphor

An estuary is the stretch where a river loses its grip and the sea starts to win, twice a day, at every tide. Fresh water and salt water don’t blend cleanly, they layer and swirl and shift with wind, rainfall and the lunar cycle, so the same channel can run fresh in a wet week and brackish in a dry one. Nothing about an estuary sits still for long, except the parts that do. The bedrock, the higher banks, the shape of the wider bay, these were laid down over geological time and will outlast anyone currently working with them. Everything downstream of that is provisional: mudflats that dry out and flood twice daily, sandbanks that can migrate metres in a single storm, saltmarsh that stabilises a channel for a decade and then loses it to a new course entirely.

That combination, a small amount of genuinely fixed ground holding a much larger area of continuous, uneven change, is roughly what Snowden is pointing at. Organisations often work in a similar way. There’s usually a bedrock layer, regulation, deep infrastructure, culture built up over years, that isn’t moving regardless of how much energy is thrown at it. And there’s a much larger zone that looks solid on an org chart but behaves more like mudflat in practice: routines, informal norms, working relationships, resourcing decisions, reshaping themselves continuously whether anyone is managing that or not. Treating mudflat as bedrock, insisting a routine is fixed when it’s actually renegotiated every week, can waste as much energy as treating bedrock as mudflat, pushing at something that was never going to move on that timeline.

Estuaries are also among the more biologically productive habitats there are, largely because of that churn. The mixing of fresh and salt water and the constant redeposit of sediment create conditions few other environments offer. It’s worth sitting with that comparison: the zone of flux isn’t just a hazard to be managed around, it’s often where the most is actually happening, and where an intervention has decent odds of taking hold.

Tidal timing carries its own lesson too. A boat crosses a mudflat, or works a shifting channel, only at specific points in the tide, not because the destination changes but because the conditions for reaching it do. Estuarine mapping borrows that discipline: knowing where to intervene matters less without also knowing when conditions are actually right to attempt it.

Where the framework comes from

The framework’s roots go back to a field guide Snowden co-wrote with Alessandro Rancati for the European Commission’s Joint Research Centre in 2021, itself a response to the pandemic. That guide never uses the word estuarine, but the technique it teaches, building a landscape from narratives rather than from data or expert opinion, deconstructing short accounts people gave directly into roles, resources, boundaries and processes, then mapping the resulting agents and constraints, is recognisably the seed of what estuarine mapping became. The guide’s own constraint typology gets most of the way there too, it already distinguishes rigid, flexible and permeable constraints, alongside a category it calls dark constraints, effects that are visible without a nameable cause, described there, as later, through the analogy of dark matter. The typology of six types the framework eventually settled on reads less like a break from that earlier thinking than a tidying of it. One principle from that earlier work is worth keeping in view: disintermediation, the insistence on building the picture from what people directly experienced rather than from theory imposed from outside.

The later framework also borrows some of its underlying language from Constructor Theory, a piece of theoretical physics associated with David Deutsch and Chiara Marletto, which tries to describe physical laws as statements about what’s possible and what’s impossible rather than as predictions about specific motion or behaviour. That’s also where the word counterfactual comes from: in this context it means something ruled out, for now, by what’s actually possible, not merely undesirable or unplanned. A constructor, in Deutsch and Marletto’s version, is something capable of producing a repeatable transformation without itself being changed in the process. Snowden departs from that on one point and says so directly: in his version a constructor can change as it acts, provided it keeps a recognisable continuity of identity over time, a deliberate loosening of the physics for the sake of something that has to describe people and habits, not just particles. It’s an unusual place for a leadership and strategy framework to borrow from, and worth naming rather than smoothing over, since it’s part of why the terminology can read as more technical than the average change model.

Mapping the elements

The simplest version: At its most basic, Estuarine Mapping asks you to identify what is shaping the situation, estimate how difficult and how slow each thing would be to change, and then distinguish between what is effectively fixed, what is too volatile to rely on, and what is realistically movable. Everything that follows adds precision to those three moves.

The more complex version: Estuarine mapping starts by identifying what Snowden calls actants, anyone or anything with a genuine effect on the system you’re mapping. That splits into three broad kinds:

• Constraints. Things that limit or shape what’s possible, covered in more detail below.
• Constructors. Things that produce a repeatable transformation, a process, a ritual, a piece of habitual practice, a role someone occupies.
• Actors. People occupying a position in the system, whose behaviour matters as much as any formal rule around them.

Most useful actants turn out to be constraints, and they get looked at through two overlapping lenses. One is whether a constraint is governing or enabling, whether it works by containing something (a rule, a boundary, a limit) or by connecting things that wouldn’t otherwise interact (a shared platform, a forum, a relationship). The other lens matters more for what follows: a typology of six types, split into two families.

Robust constraints fail catastrophically once pushed past their limit:

• Rigid. Holds firm until it’s pushed past its limit, then breaks all at once rather than bending. Snowden’s own example is a sea wall or a dyke, brilliant right up until the point it isn’t. A regulatory deadline is the organisational equivalent, so is a fixed annual budget that can’t be renegotiated once the year is underway.
• Elastic. Flexes under pressure and springs back once the pressure eases, rather than staying changed. Overtime worked during a busy period is a familiar example, so is a process a team quietly loosens under deadline pressure and just as quietly restores afterwards.
• Tether. Holds something in place from a distance, often invisibly, rather than through a boundary anyone sees day to day. Snowden compares it to a climbing rope, slack and unnoticed until the moment it’s needed, then suddenly the thing holding you. A funding condition agreed years ago that still limits what’s possible now is the organisational version, so is an old contractual clause that nobody currently in the room actually negotiated.

Resilient constraints tend to fail more gracefully:

• Permeable. Lets some things through and blocks others rather than being simply open or closed, and tends to fail by clogging rather than breaking outright. An approval process that clears routine requests quickly but bottlenecks on anything unusual is a common example.
• Phase shift. Holds steady until a threshold is crossed, at which point the whole system settles into a genuinely different pattern rather than snapping back to where it was. Snowden points to Roe v Wade as an example, decades of a particular legal reality, then a single ruling and an entirely different one, with no way back to the old pattern by the same route. Team morale that absorbs setback after setback for months, then tips into disengagement that doesn’t recover once one change too many lands, is the organisational version of water turning to ice.
• Dark. A constraint whose effect is visible even though its cause resists being named, the organisational equivalent of dark matter. “That’s just how things are done here” is usually a dark constraint talking, and it tends to surface through people’s stories rather than through direct questioning.

Constructors get less formal treatment in the framework, and Snowden is explicit that his fuller list of them (thirteen types at last count) is still just a list rather than a settled typology, one more sign of a method still being worked out in public. A handful of examples carry the idea well enough:

• Rituals and processes. Something that produces the same outcome each time it runs, because the process itself carries the work rather than any one person’s judgement on the day. A recurring meeting or an onboarding sequence works this way, structure standing in for memory. A daily stand-up is the everyday version, so is a shift handover that reliably passes on what the next person needs to know regardless of who’s on either end of it.
• Roles and rules. A person or a written rule that shapes behaviour consistently whenever it’s invoked, doing the same job however many times it’s called on. A duty manager is this kind of constructor, so is an escalation policy that routes a problem to the same place no matter who raises it. What separates either from a constraint is that they produce something, a decision, a response, rather than simply block or permit.
• Artefacts and habits. A physical or behavioural pattern embedded consistently enough to function like machinery, doing its job whether or not anyone’s paying attention to it. A shared template is one example, so is a habitual way of opening a meeting that reliably sets the same tone before anyone’s said anything about the agenda. These can drift into something closer to a rigid constraint than a constructor once they stop producing anything and simply persist because nobody’s stopped to ask why, worth watching for.

Continuity of identity matters here too, the loosening from constructor theory mentioned earlier. A retrospective can change format, length, even facilitator, and still count as the same constructor, provided it keeps producing the same kind of transformation, a shared read on what to change next. What it can’t survive is drifting into something people simply comply with, unchanged in form but no longer producing anything, at which point it’s stopped functioning as a constructor at all.

Once the actants are named, the actual mapping starts, and it’s simpler in practice than the terminology suggests. Draw two axes on a large sheet of paper or a whiteboard. The vertical axis is energy: how much money, time, attention or political capital a change would genuinely cost, running from low at the bottom to high at the top. The horizontal axis is time: how long that change would realistically take to land, running from soon on the left to distant on the right. Time doesn’t get marked out evenly, because the gap between next week and next month matters more than the gap between next year and the year after, so the scale compresses as it moves right, roughly hours, then days, then weeks, then months, then years, each band wider than the last.

Then place each actant on the grid as a single mark, purely on the basis of what it would cost and how long it would take to shift, nothing else. A habit that only survives because the one person who insists on it happens to be in the room is cheap and fast to change, so it sits low down and close to the left. A legal reporting requirement sits high up and far to the right. Most of what gets placed lands somewhere in between, which is rather the point: the map only earns its keep once a room full of people has argued over where things actually belong, not once someone’s filled it in alone.

Once the grid is reasonably full, two lines get drawn across it by negotiation rather than calculation, splitting it into three zones.

• The counterfactual zone. The ground beyond what’s realistically achievable within your planning horizon, the border named for the same counterfactual sense described above. That doesn’t mean permanently fixed, only fixed on the timescale you’re actually working to. A statutory reporting requirement sits here, so does a reporting line that’s survived three reorganisations already, so does the layout of a building nobody has budget to alter. Treating any of these as live options is where a lot of change energy quietly disappears, campaigning against a legal requirement or trying to talk a building into a different shape wastes effort better spent elsewhere. Naming something as counterfactual isn’t giving up on it, it’s choosing not to spend this round of effort there.
• The volatile zone. Snowden’s original name for this was the vulnerable zone, later dropped for the unwanted implication that vulnerability is inherently bad. Actants here are so cheap and quick to shift that change happening there doesn’t tell you much on its own, a seating plan rearranged every few weeks, whether a particular meeting starts on time this week, which channel a team happens to be using for updates this month. If something in this zone has just changed by itself, that’s rarely a result worth claiming, more often it’s a sign of instability, or a sign the thing was never a serious constraint to begin with. What you do about it depends on impact rather than on the movement itself: an actant whose impact is trivial, drifting around in the volatile zone, is safe to ignore, that’s just noise. One whose impact would actually matter, behaving the same way, is worth active containment, not because it’s important in the abstract, but because something that shifts this easily and matters this much is exactly where an unwelcome surprise tends to come from.
• The liminal zone. The ground between the other two, and the one the whole method is really built for. Actants here would cost a plausible, if not trivial, amount of energy, and would take a realistic amount of time, weeks or months rather than hours or years, to shift. An unenforced policy that everyone quietly works around belongs here, expensive enough to fix properly that it won’t happen by accident, cheap enough that a modest effort, well aimed, could genuinely move it. So does a working relationship between two teams that’s currently poor but not structurally doomed. This is where a portfolio of small interventions gets aimed, and where the earlier argument about what belongs where does most of its work, getting a room to agree that something sits in the liminal zone, rather than being either untouchable or trivial, is often the moment a stuck change conversation actually starts moving.

Once the map is drawn, the question changes. A conventional strategy conversation starts by asking what the organisation wants to achieve. Estuarine mapping asks a narrower, more useful question first: given this map, where is movement actually possible? Each actant then gets one of a small number of responses, chosen for its position on the grid rather than its importance in the abstract:

• Act now. For actants sitting comfortably in the liminal zone, where the cost and timescale are both realistic, this is where a direct intervention is aimed. If the actant is already functioning as a constructor, producing a wanted outcome rather than just blocking things, the choice is usually to maintain it as it is or amend it. If it’s acting instead as a pure barrier, the choice leans towards loosening or destroying it. Where the map turns up a genuine gap, a useful constructor that plainly doesn’t exist yet, the answer isn’t to design one from first principles and roll it out, it’s to run what Cynefin practitioners call a safe to fail experiment, a small, reversible attempt at creating one, built so that failure teaches you something rather than just costing you something. An unenforced policy that everyone quietly works around is a typical candidate, expensive enough to fix properly that it won’t happen by accident, cheap enough that a modest effort could genuinely shift it. The question worth asking the room: what’s the smallest version of this change we could start this week, without waiting for permission we don’t actually need?
• Monitor. For actants near a boundary, close to tipping from liminal into volatile or counterfactual, watching for movement matters more than acting yet. A working relationship between two teams that’s begun to improve, but hasn’t held long enough to trust, fits here, so does a habit that’s started to loosen without anyone deliberately pushing it. The question worth asking: what would we expect to see over the next month if this really is shifting, and what would tell us we’d read it wrong?
• Escalate. For actants that matter but sit beyond what a team can move on its own, raising the constraint to whoever holds the energy or authority to shift it is the honest response, rather than pretending it’s actionable locally. A resourcing decision set two levels up that’s quietly shaping what a team can attempt is a common example, workarounds accumulate for years around constraints like this because nobody has ever put the case to the person who could actually change it. The question worth asking: whose decision would this actually need to become, and what exactly have we told them about it so far?
• Design around. For actants confirmed as counterfactual, genuinely fixed for now, the response is to design the intervention to work despite the constraint rather than against it. A legal requirement or a piece of physical infrastructure that isn’t changing this year regardless of the case for it belongs here. The question worth asking: given this isn’t moving, what would we build instead, working with it rather than waiting for it to give way?

There’s a fifth kind of move worth naming separately, one that doesn’t act directly on any single actant at all. Rather than intervening in the actant itself, you intervene on its position on the grid, lowering the energy a future change would cost, or extending how long something can hold before it needs revisiting again. A team that writes an unofficial workaround into a shared playbook hasn’t fixed the underlying constraint, but has made it cheaper to revisit: next time, nobody has to reconstruct tribal knowledge from scratch before the conversation about fixing it can even start. The move works in the other direction too. A practice that took real effort to establish, sitting in the liminal zone, easy to lose the moment its champion moves on, can be deliberately pushed towards the counterfactual end instead, built into onboarding, written into policy, given a formal home in governance, precisely so it becomes harder to erode by accident. This kind of move is often worth more than any single action on the list above, since it reshapes the map itself rather than solving one problem in isolation. Snowden has a useful image for the posture this asks of you: an ecologist who happens to use engineering, rather than an engineer who treats the ecosystem as something to be conquered. The aim isn’t to force a particular outcome into existence, it’s to shift the conditions so that the outcomes you want cost less energy than the ones you don’t, and let the system find its own way toward them. The question worth asking: if we did nothing else this quarter except make the next version of this problem cheaper to solve, or this practice harder to lose, what would we actually build?

None of this is meant to produce one plan. A portfolio built from these responses will look deliberately scattered on paper, three or four small interventions running in parallel across different parts of the liminal zone, a couple of things being watched rather than acted on, one constraint that’s been formally escalated and is now someone else’s problem to move. That’s by design, a single large initiative aimed at the whole landscape at once tends to reproduce exactly the kind of plan estuarine mapping exists to get away from. And the map itself has a shelf life. It describes the present, not a forecast, so a genuine change in the landscape, a constraint that quietly shifts from counterfactual to liminal, a monitored actant that tips into volatile, is reason enough to redraw it rather than treat the original version as still authoritative. Revisiting the map on a set rhythm, tied to whatever the monitored actants suggested, keeps it honest.

The tension between rigour and accessibility

How deep you use the tool depends on what’s actually available, people, time, and how fast the organisation moves as a matter of habit.

With just an afternoon and a handful of people, pick one stuck question rather than the whole organisation, gather eight or ten short stories about it, and sketch the grid together on the spot. Skip the jargon; nobody needs to hear a technical term to say what’s shifting and what isn’t. Give the map a month, then put it back on the table. Use it to help orientate yourself.

Given more time and a wider group, gather stories over several weeks instead, from people who wouldn’t normally end up in the same room, and use the full six-way typology rather than a quick call on what’s fixed. That’s where the real disagreements tend to surface. Either way, the map is always provisional, never a finished product.

Tips for running it

• Let people explore their stories: Don’t rush them to the point. The detail that seems like a tangent, what someone was wearing, who else was in the room, is often where the actual constraint shows up. Cutting people off to get to “the point” usually just gets you the point they’d already rehearsed.
• Let the disagreement about what’s fixed run: Don’t drive for consensus. Two people insisting the same policy is both immovable and easy to change isn’t a problem to resolve quickly, it’s the map telling you something real. Let them argue it out in front of the group rather than smoothing it over.
• Ask who saw it differently: Don’t let the majority account speak for the room. Most groups converge on the same story fast, and the version that wins isn’t always the most accurate one, just the loudest or the first told. A direct question, who noticed something else, usually surfaces what the room would otherwise have buried.
• Expect the first pass to be too big and vague: Don’t settle for it. “Communication is broken” isn’t a story, it’s a conclusion. Push once or twice more, what happened, who was there, what did they actually say, until it’s specific enough to place somewhere on the grid.
• Let it stay messy for a while: Don’t sort it into the grid too soon. The instinct to tidy things into categories kicks in almost immediately, and it’s usually too early. Give the room time to just describe what’s there before anyone starts deciding what it is.

Five questions for leaders

You don’t need the grid, the axes or the full typology to start using this thinking. These are the questions I’d put to a leadership team before any of that, partly because the answers usually tell you whether the fuller exercise is worth the room’s time at all:

1. What are we currently treating as fixed that’s actually just expensive or slow to change, not genuinely impossible?
2. Where have we mistaken easy, constant movement for real progress?
3. What disagreement about what’s really fixed have we been smoothing over rather than letting run?
4. What’s the smallest thing we could actually move this month, rather than the biggest thing we’d like to?
5. When did we last update our sense of what’s possible, rather than still working from an old map?

Close

This isn’t a framework that hands you an answer. It hands you a clearer look at what’s actually possible right now, and a few small things worth trying because of it. That’s a smaller promise than most strategy tools make, and a more honest one. Dave Snowden is still developing this, years after initially publishing it. That is worth taking at face value: try it small before you trust it large, and expect to redraw your map more than once.

Sources

Snowden, D. and Rancati, A. (2021) Managing complexity (and chaos) in times of crisis: a field guide for decision makers inspired by the Cynefin framework. Luxembourg: Publications Office of the European Union. doi:10.2760/353.

Snowden, D. (2022) ‘Estuarine mapping first edition’, The Cynefin Co, 7 October. Available at: https://thecynefin.co/estuarine-mapping/ (Accessed: 6 September 2026).

Cessan, V. (n.d.) ‘My experiences with Estuarine Mapping’. Available at: https://www.viktorcessan.com/my-experiences-with-estuarine-mapping/ (Accessed: 6 September 2026).

Bartlett, M. (n.d.) ‘Estuarine mapping, part one’, Laws of CX. Available at: https://lawsofcx.substack.com/p/estuarine-mapping-part-one (Accessed: 6 September 2026).