Lewin’s Change Management: An S-Tier Behavioral Designer’s Guide

Kurt Lewin died of a heart attack in February 1947, at fifty-six, a few months after the paper that would make him the father of change management appeared in the first issue of Human Relations.

He never saw what happened to it.

Somewhere in the middle of that thirty-six-page paper, under a subsection heading about group standards, sits a short passage naming three aspects of a successful change: unfreezing the present level, moving to a new level, and freezing group life on the new level. A few sentences, in a subsection, about things like cafeteria menus and factory output norms.

Eight decades later, that passage is taught as Lewin’s change management model, and it is the most-taught framework in organizational change. Every MBA syllabus. Every consulting deck. Every change-management certification on the market. It also carries a word Lewin never used, because he wrote “freezing” and someone else supplied “refreezing” three years after he was dead.

Here is the part that should bother you. Lewin’s actual position was that the thing you are being told to unfreeze was never frozen in the first place. He described the status quo as a quasi-stationary equilibrium and compared it to a river, flowing at a given velocity in a given direction, perfectly capable of holding a steady level while everything inside it moves. A river does not need melting. It needs its balance of forces changed.

That single correction rewrites the prescription, and it explains why so many change programs feel like shoving a boulder that shoves back. This guide takes the model apart: what Lewin actually wrote and how a subsection became a global framework, the diagnostic half of his work that almost nobody runs, the one empirical result he left behind that is worth more than the three steps combined, which Core Drive is really doing the resisting when an organization digs in, and the most-quoted statistic in the entire field, which turns out to have no source at all.

Speed Run Notes

• The provenance: Lewin never packaged a three-step model. A short passage in a 36-page 1947 paper names unfreezing, moving and freezing of group standards, and he wrote “freezing.” Festinger supplied “refreezing” in 1950.
• The correction that matters: he called the status quo a quasi-stationary equilibrium and compared it to a river holding a steady level while everything in it moves. Nothing is frozen, so nothing needs thawing.
• The prescription nobody uses: Lewin argued for reducing restraining forces ahead of adding driving forces, because extra push raises tension inside a system that is still opposed. Most change programs do the opposite.
• The Core Drive: what resists you is usually Core Drive 8 (CD8): Loss & Avoidance, sharpened by Core Drive 4 (CD4): Ownership & Possession. Another vision speech aims at the wrong half of the force field.
• The evidence he actually left: in his 1943 wartime food study, 23 of 44 housewives who discussed and decided as a group served the target food, against 4 of 41 who sat through an expert lecture. A social and ownership result, not an inspiration one.
• The statistic: “70% of change initiatives fail” traces to one sourceless sentence in a 2000 Harvard Business Review article. There is also no quantitative test of the three-step model itself.

Author Credibility: Yu-kai Chou

Yu-kai Chou created the Octalysis Framework after studying gamification since 2003 — years before the term entered mainstream vocabulary. As a Human-Systems Architect & Behavioral Designer, his framework has been applied by LEGO, Microsoft, Porsche, Coca-Cola, Salesforce, and MrBeast, impacting over 1.5 Billion Users.

Chou has taught the Octalysis methodology at Harvard, Stanford, Yale, Tesla, Google, BCG, and IDEO.

His work has been cited by Harvard, Stanford, MIT, Forbes, Wall Street Journal, Wired, US Department of Energy, NIST, NSF, NCBI, US Department of Education, ClinicalTrials.gov, and Google Scholar — with 3,700+ more academic publications. Explore his books here.

What Is Lewin’s Change Management Model?

Lewin’s change management model, as it is taught today, says that lasting change happens in three stages.

Unfreeze. Loosen the existing state. Create dissatisfaction with how things are, so people become willing to let go of the current way of working.

Change (Lewin’s word was moving). Shift to the new way. New processes, new behaviors, new structures, usually accompanied by training and support.

Refreeze (Lewin’s word was freezing). Lock the new state in place so the organization does not drift back. New norms, new reward systems, new defaults.

That is the version on the slide. It is clean, memorable, and teachable in ninety seconds, which is a large part of why it has outlived nearly every framework invented since.

The version Lewin argued for is stranger and more useful. He treated any social system as a field of forces in tension, and the current state as a balance point those forces happen to be holding. Change, in that reading, is a question about the forces rather than about the state. Which forces are pushing toward the new level? Which are pushing back? What happens to the system’s internal pressure depending on which side you touch?

That framing produces a specific and counterintuitive instruction about where to intervene, which the three-step summary drops entirely. We will get to it.

Two definitions worth pinning down before anything else, because the rest of this guide leans on them.

Quasi-stationary equilibrium. A level that stays recognizably constant while the process producing it keeps running. The water level of a river, not the shape of a rock.

Force field. The set of driving forces pushing a system toward a different level, and restraining forces pushing back, whose balance holds the current level in place.

The Two Pages the Whole Model Came From

The source is a single paper: Kurt Lewin, “Frontiers in Group Dynamics: Concept, Method and Reality in Social Science; Social Equilibria and Social Change,” published in the first issue of Human Relations in 1947, running from page 5 to page 41.

It is a dense paper about group standards, social equilibria, and the measurement problems of social science. Somewhere inside it sits a subsection with the heading “Changing as Three Steps: Unfreezing, Moving, and Freezing of Group Standards.”

The operative sentence describes a successful change as including three aspects: unfreezing (if necessary) the present level L1, moving to the new level L2, and freezing group life on the new level.

That is the seed. Everything else grew from it later: the ice cube, the three boxes with arrows between them, the certification modules.

He was already dead when the paper landed

Lewin died on 12 February 1947 of heart failure, at his home in Newtonville, Massachusetts. He was fifty-six.

He had founded the Research Center for Group Dynamics at MIT two years earlier and was running it at the time of his death. In 1948 the center relocated to the University of Michigan under Dorwin Cartwright, and in 1949 it merged with the Survey Research Center to become the Institute for Social Research, where it still operates.

So the author was never available to say what he did or did not mean by that subsection. Everyone who built on it was interpreting a dead man’s shorthand.

Who added the word “refreezing”

Lewin wrote freezing. The familiar word is refreezing, and the two are not interchangeable, because the prefix smuggles in an assumption: that the thing was frozen to begin with, then thawed, then set again.

Stephen Cummings, Todd Bridgman and Kenneth Brown traced the substitution in Human Relations in 2016. Their finding is that the word first appears in a 1950 conference paper by Leon Festinger and a co-author. Festinger had been one of Lewin’s own students at the Research Center for Group Dynamics.

One prefix, added by a student three years after the teacher’s death, and it reframed the entire model.

How a subsection became the canon

The same paper makes a stronger claim, and it is the one worth sitting with. Cummings and colleagues argue that the three-step model was largely invisible until the 1980s, when change management was consolidating into a discipline and needed a founding text. Lewin’s passage was available, brief enough to be summarized without contradiction, and attached to an eminent name. It got promoted from a paragraph to a foundation.

Their term for the result is a reconstruction with noble provenance. A field wanted an origin story, and this one fit.

This reading is contested, and honesty requires saying so plainly. Bernard Burnes, who has spent much of his career working on Lewin, published a direct rebuttal in the Journal of Applied Behavioral Science in 2020 arguing that Lewin did develop the three-step approach as a considered method grounded in field theory, and that the difference between Lewin’s version and later presentations is mostly nomenclature. His earlier 2004 reappraisal in the Journal of Management Studies makes the same case at book length: the three steps were one element of a much richer planned-change approach that also included field theory, group dynamics and action research, and critics who attack the three steps in isolation are attacking a stub.

Burnes and Cummings disagree about provenance. They agree about the thing that matters for anyone designing an actual change program: the three steps on their own are an amputated version of Lewin’s method. Whether the amputation happened in 1950, 1980, or in the retelling, the diagnostic apparatus is missing from the version you were taught.

The River Lewin Actually Described

Here is the sentence that should have become famous instead. Writing about social change in 1947, Lewin explained what he meant by a state of no social change:

by a state of “no social change” we do not refer to a stationary but to a quasi-stationary equilibrium; that is, to a state comparable to that of a river which flows with a given velocity in a given direction during a certain time interval. A social change is comparable to a change in the velocity or direction of that river.

A river. Moving the whole time. Holding a level anyway.

Elsewhere he put it more bluntly: the status quo in social life is a dynamic process that flows on while still keeping a recognizable form.

Now compare that to the metaphor the model is famous for. In The Challenge of Organizational Change (1992), Rosabeth Moss Kanter, Barry Stein and Todd Jick delivered what became the standard criticism, on page 10:

This quaintly linear and static conception – the organization as an ice cube – is so wildly inappropriate that it is difficult to see why it has not only survived but prospered.

The criticism is correct about ice cubes. The trouble is that Lewin never proposed one. He proposed a river, in print, and the ice cube was supplied by his critics to describe a model his popularizers had already flattened.

Thirty-plus years of “Lewin is too static for modern organizations” rests on an image Lewin explicitly ruled out.

Why the metaphor decides your strategy

This is not a history-of-ideas quibble. The two metaphors hand you completely different jobs.

If the organization is an ice cube, your job is thermal. Apply enough heat to melt resistance, pour the water into a new mold, then chill it. Heat, in practice, means urgency: the burning platform, the town hall, the survey results that prove the current state is unacceptable, the deadline. Every one of those is an input you add.

If the organization is a river, your job is hydraulic. The level you dislike is being actively produced, right now, by forces that are still running. Nothing needs melting because nothing has set. What holds the level is a live balance, and you change the level by changing the balance.

Ask a leadership team why a process has not changed and you will usually hear a variation of “people are resistant to change.” That is ice-cube thinking. It locates the problem in a substance that needs softening.

The river question is different and far more productive: what is currently holding this in place, and who benefits from it holding? Every stuck process in every organization is being held by something specific: a reporting line, an incentive, a legacy system, someone’s headcount, a fear that has never been said out loud. Those are the forces. They are usually nameable within an hour of asking properly.

Force Field Analysis: The Half Nobody Runs

Force field analysis is Lewin’s diagnostic method, and it is the half of his work that got left behind when the three steps went on tour.

The procedure is simple enough to run on a whiteboard. Draw a vertical line and label it with the current state. On the left, list every driving force pushing toward the change you want: competitive pressure, a new regulation, executive sponsorship, a customer complaint. On the right, list every restraining force pushing back: fear of looking incompetent on a new system, a bonus scheme that pays for the old behavior, a team lead whose expertise is tied up in the current process, sheer volume of existing work.

Score each force by strength. Now you have a picture of the equilibrium instead of a feeling about it.

The prescription almost nobody follows

Here is where Lewin’s argument gets sharp, and where the three-step summary drops the most valuable thing he wrote.

He observed that you can shift the equilibrium two ways: add forces in the direction you want, or diminish the forces opposing you. Both can move the level. They do not produce the same system.

Adding driving forces reaches the new level with high tension inside the system, because the restraining forces are all still there, now compressed harder. Reducing restraining forces reaches the same new level with low tension. And Lewin’s own view was that as a rule, the low-tension route is preferable to the high-pressure one, because high tension brings aggressiveness, emotionality, and less constructive behavior along with it.

Read that again against how change actually gets run.

The all-hands with the burning-platform slide is a driving force. The countdown to migration day is a driving force. So is the executive sponsor emailing about alignment, the dashboard turning red, and the bonus tied to adoption.

Nearly every lever a change program reaches for adds push. Almost nothing in a standard playbook removes restraint, because removing restraint is unglamorous work: killing a report nobody reads, rewriting a compensation rule, giving someone a graceful exit from a role their identity is welded to, admitting the new system is worse at one thing the old one did well.

None of that photographs well in a steering committee deck. All of it is what Lewin actually recommended.

Tension is a cost you pay later

The reason this matters operationally is that compressed restraining forces do not disappear. They wait.

You can absolutely force a level change with enough executive push. The system will sit at the new level for as long as the push continues, and the moment attention moves to the next initiative, the restraining forces are still there, still at full strength, and the level slides back. Lewin saw this in 1947 and named it: change toward a higher level of group performance is frequently short-lived, and after a shot in the arm, group life soon returns to the previous level.

Which is the sentence the whole three-step passage exists to solve. He was answering a specific question about why improvements decay. His answer was that permanence has to be designed in as an objective, not hoped for afterwards.

The version that survived turned that answer into a box labeled “refreeze,” to be completed in Q4.

The Evidence Lewin Actually Left

There is essentially no quantitative test of the three-step model as a model. Nobody has run a controlled comparison of organizations that followed unfreeze-move-freeze against organizations that did not, with defined outcome measures and the stages isolated. It has been applied thousands of times and used to narrate case studies after the fact, and it has never been made to make a prediction that could fail.

That is worth stating plainly, because the model is usually presented with the authority of a research finding.

Lewin did leave hard evidence. It is just about something else, and it is better.

The wartime food experiments

During the Second World War, meat was scarce and organ meats were not. The National Research Council’s Committee on Food Habits wanted American families to start eating kidneys, hearts and sweetbreads, and Lewin was asked to work out how.

The problem is a change-management problem in miniature. A population has an established behavior, a good external reason to change it, and no interest in changing it.

Lewin ran a comparison. Groups of housewives, Red Cross volunteers, in sessions of roughly fifteen. One arm heard a skilled nutritionist give a lecture: the health case, the war-effort case, the preparation methods, recipes to take home. The other arm got much the same information, then held a group discussion and reached a decision together about whether they would try serving the food.

Follow-up came about a week later.

In the group-decision arm, 23 of 44 women had served one of the target meats. In the lecture arm, 4 of 41. Fifty-two per cent against ten.

Same room, same information, same expert, roughly the same amount of time. The only variable was who did the deciding.

One footnote for anyone who goes looking. The figures most often quoted for this study are 32 per cent against 3 per cent, which come from Lewin’s own later summary of the work rather than from the 1943 report. They do not match the counts in the primary source. The direction and the size of the gap hold up either way, but if you are citing a number, cite the one from the report and know which is which.

What that result is actually about

The lecture arm is a driving force. It adds information, authority and moral pressure, which is everything a good change communication does, delivered by a professional.

The discussion arm changes who owns the decision.

The restraining forces in that living room were never informational. They were “my husband will complain,” “I have never cooked this and will do it badly,” “my mother would find this strange,” “nobody I know serves kidneys.” A lecture cannot touch any of those. A room of fifteen women in the same position, saying out loud that they will try it, dissolves most of them at once, because the social fact that made it strange has just changed.

Lewin’s most cited empirical result is a demonstration that reducing restraint beats adding push, run twenty-five years before anyone drew the three boxes.

It is not in the model. It is barely in the textbooks. It is the single most useful thing he found.

Gatekeepers, and where the leverage sits

The same research produced Lewin’s channel theory, which deserves more attention than it gets.

Food does not arrive on a table by accident. It travels through channels (the store, the garden, the icebox) and at the entrance to each channel sits a gate, controlled by a person Lewin called the gatekeeper. In 1940s American households that was almost always the housewife. Persuading the family was close to pointless. The food never entered the house unless one specific person put it in the basket.

The organizational translation is direct and slightly uncomfortable. Every change you are trying to make travels through channels with gates on them, and the gatekeeper is rarely the person named as the sponsor. It is the ops manager who decides which tickets get triaged, the finance analyst who owns the template everyone copies, the team lead whose opinion the engineers actually check. Change programs are typically aimed at the org chart. The gates are somewhere else.

Lewin’s term for a gatekeeper is worth keeping in your working vocabulary, because it forces a better question than the one on the RACI chart. Stop asking who approves this. Ask who can quietly refuse it.

The Octalysis Anatomy of Unfreezing

I built the Octalysis Framework to answer a question Lewin’s force field asks but does not answer: what are the forces made of?

Force field analysis tells you a restraining force is present and roughly how strong it is. It does not tell you which of the eight human motivations is generating it, and that determines what actually removes it. A restraint made of fear does not respond to the intervention that dissolves a restraint made of pride.

Restraining forces are usually Core Drive 8

Core Drive 8 (CD8): Loss & Avoidance is the motivation to prevent something bad from happening, and it is the dominant engine of organizational resistance.

Look at what people actually say when they push back on a change, once you get past the professional phrasing. I will be slower at this than I am now. My expertise stops counting. My headcount justification disappears. The thing I built gets deleted. I will look stupid in front of people I manage. Every one of those is a projected loss.

This is why the burning-platform speech so often backfires. It is an attempt to activate Core Drive 8: Loss & Avoidance at the organizational level, we will all lose if we do not move, aimed at people whose CD8 is already fully engaged at the personal level, pointed the other way. You are not adding motivation. You are adding a second threat to a person already managing one, and the personal loss is concrete while the organizational loss is abstract.

The design implication is uncomfortable and reliable. If the restraint is CD8, the intervention that works is making the feared loss not happen, visibly and in advance. Name the role that survives the change. Guarantee the ramp period where being slow is expected. Say out loud what happens to the person whose specialty is being automated, before they have to ask.

The second force is ownership, and it is not irrational

Core Drive 4 (CD4): Ownership & Possession is the motivation that comes from owning, controlling and improving something, and it sharpens every restraining force it touches.

People do not resist changes to processes. They resist changes to their processes. The spreadsheet somebody has maintained for six years is not a spreadsheet to them, and the research on psychological ownership is clear that this feeling attaches to things we have invested effort in regardless of who legally owns them. The IKEA effect is the same mechanism in a consumer setting: labor creates valuation.

Which reframes the food-habits result exactly. The group-decision arm did not just lower social restraint. It moved ownership of the decision from the nutritionist to the room. The women were not complying with a change. They were executing their own.

Any change design that leaves people zero authorship is fighting CD4 the entire way, and CD4 does not get tired.

Where social influence and meaning belong

Core Drive 5 (CD5): Social Influence & Relatedness is the motivation that comes from other people: what they do, what they approve of, what they will think.

Lewin’s whole career pointed here. Group standards, not individual attitudes, were the unit he worked on, and his finding was that it is far easier to move a group than to move the individuals in it one at a time. The reason is CD5: each person’s restraining force is partly composed of what they expect everyone else to do, so a visible group shift removes a restraint from all of them simultaneously.

This is the closest thing to a free lunch in change design, and it is why Diffusion of Innovations keeps rediscovering it.

Core Drive 1 (CD1): Epic Meaning & Calling is the motivation of serving something larger than yourself, and it is where most change programs spend their entire budget.

The vision. The mission narrative. The strategy deck. Epic Meaning is powerful, and it is a driving force, which means it lands on the side of the field that was never the bottleneck. A person who believes in the vision and still expects to be personally worse off after the change does not move. Belief and restraint sit on different sides of the line, and the restraint side is the one holding the level.

Which quadrant this sits in

Octalysis sorts motivation along two axes. White Hat drives make people feel powerful and in control; Black Hat drives create urgency and compulsion and get results fast at the cost of long-term wellbeing. Left Brain drives are extrinsic and goal-focused; Right Brain drives are intrinsic and experience-focused.

The standard change playbook is Black Hat and Left Brain: deadlines, mandates, compliance dashboards, consequences for non-adoption. It works, in the way Black Hat always works: fast, effective, and corrosive if it is the only thing running. The tell is a change that holds precisely as long as someone is watching.

Lewin’s own prescription lands in a different place. Removing restraining forces is White Hat work, and the group-decision method is White Hat and Right Brain at once: people feel more in control after it, not less. That is the mechanical reason it produced durable behavior change in a week rather than compliance for as long as the nutritionist was in the room.

Where the Three-Step Model Falls Apart

Four failure modes, in rough order of how often I see them do damage.

Failure one: refreezing assumes a pause that no longer exists

The standard criticism is the strongest one, and it survives even after you correct the river-versus-ice-cube confusion.

Kanter, Stein and Jick made it in 1992, and the emergent-change literature has hammered it since: organizations do not hold still long enough to be refrozen. A three-year transformation now overlaps with two reorganizations, a platform migration and an acquisition. There is no window in which the system sits stable while you set the new norms.

Note where this criticism actually bites. Lewin’s own framing, a river holding a level, handles continuous flux fine, because a level and a flow are compatible. What cannot survive contact with a modern org is the project management that got attached to the three steps: a stage with a start date, an end date and a completion criterion, where step three finishes and the change is declared done.

The metaphor is defensible. The Gantt chart built on top of it is not.

Failure two: it makes change an episode, so everything else becomes an interruption

Frame change as a bounded project and you create a category called “normal operations” that the project is a departure from. Everyone inside the organization can feel where that leads. People wait it out. Waiting is rational, because the frame itself promised that things would settle down afterwards.

Lewin’s actual view undercuts this. If the status quo is a quasi-stationary equilibrium, then the forces are running the entire time and the level is being re-produced every day. There is no “normal operations” to return to; there is only the current balance. That frame produces a very different behavior from leadership, because it makes the maintenance of the balance somebody’s standing job rather than a phase.

Failure three: unfreezing gets treated as an activity instead of a diagnosis

This one is the most common and the most expensive.

“Unfreeze” appears on a plan as a set of tasks: run the town hall, publish the case for change, do the roadshow, send the survey. All of that is communication, and communication is a driving force.

What unfreezing means in Lewin’s framing is reducing the forces that hold the current level. Those forces are specific, local and frequently embarrassing. A director’s bonus depends on the metric the new system stops producing. Two teams have a decade-old feud and the new process requires them to share a queue. The old tool has one clearly superior feature and nobody wants to say so.

You cannot roadshow any of that away. It has to be found, named and dismantled one item at a time, which is slower and less visible than a comms plan and is the actual work.

Edgar Schein’s 1996 elaboration in Systems Practice is the best treatment of this I know. He argued that unfreezing requires three things together: disconfirmation that the current way is failing, enough survival anxiety to make change feel necessary, and enough psychological safety to make learning feel survivable. His key condition is that survival anxiety must exceed learning anxiety, and the reliable way to achieve that is to lower learning anxiety rather than crank survival anxiety higher.

Which is Lewin’s restraining-force prescription again, arriving from a different direction fifty years later. Amy Edmondson’s work on psychological safety has since given us a large body of evidence for the same lever.

Failure four: applying a group model to individuals

Lewin studied group standards. The subsection heading says so explicitly, and his central empirical claim was that it is easier to change a group’s norm than to change its members individually.

Most modern usage inverts this. Change readiness is assessed per person, resistance is scored per person, and the intervention is one-to-one coaching. That approach fights the mechanism that made Lewin’s own method work, because it leaves the social restraint completely untouched: everyone’s expectation about what everyone else will do. Fifteen individual conversations do not produce what one room reaching a decision together produces.

What’s Really Happening Inside the Brain

Lewin was writing before any of this was measurable, and the neuroscience has been broadly kind to him.

Organizational change reliably produces two things the brain treats as threats. The first is uncertainty. Prediction is metabolically cheap and error is expensive, so a system that can no longer forecast its own near future runs a stress response: elevated cortisol, narrowed attention, degraded working memory. A person in that state has measurably less capacity available for learning a new system, which is precisely what you are asking them to do.

The second is status. Social standing is processed in overlapping circuitry with physical reward and threat, which is why a demotion that costs nothing financially can land like an injury. Reorganizations redistribute status constantly and usually without acknowledging it.

Both are restraining forces, and both get worse under high-pressure change. This is the physiological mechanism behind Lewin’s warning that high tension brings aggressiveness and lower constructiveness. Turning up the driving forces raises the threat load on people whose capacity to adapt was the resource you needed.

There is a second, quieter finding worth carrying. Habitual behavior lives in different neural machinery from deliberate behavior, and habits are cued by context rather than by intention. A person can fully agree with the new process and still open the old tool, because the cue is unchanged. Some of what looks like resistance is a lookup running from a context you did not change.

The practical consequence: changing the environment often outperforms changing the message, and it is the same insight the behavioral-insights literature packages as the EAST framework: make the new behavior easy first, and treat friction as the primary target.

Lewin vs the Models That Replaced Him

vs Kotter’s 8 steps

Kotter’s 8-step model is frequently described as Lewin’s three steps with more detail, with steps 1 through 4 mapped onto unfreezing, 5 through 7 onto moving, and step 8 onto refreezing.

That mapping is retrospective. Kotter’s model came out of his own study of why transformations fail, published in Harvard Business Review in 1995 and expanded in Leading Change in 1996, and he does not claim descent from Lewin.

The more useful comparison is directional. Kotter’s eight steps are almost entirely about generating and organizing driving force: create urgency, build a coalition, form a vision, communicate it, generate short-term wins. It is an excellent playbook for the half of the field Lewin said was the less preferable place to intervene. Run both, and Kotter tells you how to push while Lewin tells you what to stop pushing against.

vs ADKAR

ADKAR, standing for Awareness, Desire, Knowledge, Ability and Reinforcement, was developed by Jeff Hiatt at Prosci and has no documented derivation from Lewin, despite the constant side-by-side comparisons.

Its real contribution is the level of analysis. ADKAR is explicitly individual: it tracks one person’s transition and lets you locate exactly where that person is stuck. That is far more actionable than “the organization is in the moving phase.”

Its blind spot is the mirror image of its strength. An individual-level model cannot see the social restraint that Lewin’s group-decision result was entirely about, and its final stage, Reinforcement, is the same refreezing assumption in newer packaging.

vs the Transtheoretical Model

The Transtheoretical Model, running from precontemplation through contemplation, preparation, action and maintenance, is the closest structural cousin, developed by Prochaska and DiClemente for individual health behavior.

It has two features Lewin’s model lacks and badly needs. It has an explicit stage before anyone is even considering change, which is where most organizational populations actually sit when a program launches. And it treats relapse as normal rather than as a failure of refreezing, which matches what happens in real deployments far better.

The related clinical technique, motivational interviewing, is the most direct modern descendant of Lewin’s method I know of. Its entire premise is that pushing harder increases resistance, and that the change argument has to come from the person changing.

vs Diffusion of Innovations

Everett Rogers described how an innovation spreads through a population over time via innovators, early adopters, the majorities and laggards. It answers a question Lewin’s model does not ask at all: in what order do people move?

Where the two connect is the social restraint. Rogers’s S-curve accelerates once adoption becomes visible enough that not adopting starts to feel abnormal, which is the same CD5 mechanism that made the group-decision arm outperform the lecture. Lewin explains the engine. Rogers describes the shape it produces at population scale.

vs Theory X and Theory Y

Douglas McGregor was a colleague of Lewin’s at MIT, and Theory X and Theory Y is worth naming here because it predicts which half of the force field a given leader will reach for.

A Theory X manager assumes people avoid work and need to be directed, and will reliably add driving forces: mandates, monitoring, consequences. A Theory Y manager assumes people will self-direct given the conditions, and is far more likely to go looking for restraints to remove. Two leaders can run the same change plan and produce opposite systems, and the difference shows up in which side of Lewin’s line they instinctively touch.

The Elephant in the Room

Almost every change-management pitch you will ever hear opens with the same number: roughly 70 per cent of organizational change initiatives fail.

It is a superb opening. It establishes stakes, it flatters the audience for being about to do better, and it justifies the fee.

It also has no source.

The line that put it into circulation is the opening sentence of Michael Beer and Nitin Nohria’s “Cracking the Code of Change,” in Harvard Business Review in May 2000:

The brutal fact is that about 70% of all change initiatives fail.

No study is cited. No data is presented. It is stated as common knowledge, in the most influential management publication in the world, and from there it entered every deck ever built on the subject.

Mark Hughes went looking for the evidence and published the results in the Journal of Change Management in 2011, under the title “Do 70 Per Cent of All Organizational Change Initiatives Really Fail?” He traced the five most prominent sources (Hammer and Champy, Beer and Nohria, a Bain article, a McKinsey article, and Kotter) and found that each either asserted the figure without evidence or cited another source that did the same. His conclusion is that there is no valid, reliable empirical basis for it.

The figure survives because it is useful to the people repeating it, which is worth remembering the next time you see it on a slide, including a slide of your own.

Why the three-step model survives, honestly

Follow that thread and it leads somewhere uncomfortable for the whole field.

Unfreeze-move-refreeze has no empirical validation as a model. Its provenance is disputed by serious scholars. Its central metaphor was supplied by its critics. It has been declared obsolete roughly once a decade since 1992.

And it is still on the first slide, because it is teachable. Three words, one arrow, ninety seconds, and a room of executives feels equipped. Nothing that has been proposed to replace it competes on that dimension, and that dimension is the one that decides what spreads.

This is a real finding about how frameworks propagate, and it applies to mine as much as to Lewin’s. Memorability and validity are independent variables. A framework survives on the first one and earns its keep on the second, and the two get confused constantly, including, and I would say especially, by people who build frameworks for a living. The honest defense of any model is not that everyone uses it. It is what it lets you see that you could not see before.

By that test, the three steps are weak and the force field is strong. We kept the wrong half.

Designing Change With Octalysis

Here is how I actually run this when a client has a change that will not take.

Step zero: find out whether you have a motivation problem or a restraint problem

Most change work starts from the assumption that people are insufficiently motivated. Test it before you spend a budget on it.

Ask a sample of the affected population one question: if this change went perfectly, what would be worse for you personally? Ask it privately, and make it safe to answer.

If most people struggle to name anything, you have a genuine motivation problem, and the standard playbook of vision, urgency and wins is the right tool.

If answers come fast and specific, you have a restraint problem, and every dollar you put into communication is being spent on the wrong side of the field. In my experience the second case is roughly four times more common, and it is nearly always diagnosed as the first.

Build the force field with Core Drives attached

Run Lewin’s diagnostic, then add a column he could not have written.

List the restraining forces. For each one, name the Core Drive generating it. Fear of looking incompetent is Core Drive 8 (CD8): Loss & Avoidance. Attachment to a system somebody built is Core Drive 4 (CD4): Ownership & Possession. Not wanting to be the first person on your team to switch is Core Drive 5 (CD5): Social Influence & Relatedness. A workflow that is simply slower on day one is not a Core Drive at all. That is pure friction, and friction is the cheapest thing on the list to fix.

The Core Drive label determines the intervention, and that is the entire value of doing this:

• CD8 restraints dissolve when the feared loss is made concretely impossible, in advance, in writing. Not reassured. Removed.
• CD4 restraints dissolve when the owner is given authorship of the replacement rather than notice of it.
• CD5 restraints dissolve when enough visible peers move that staying put becomes the conspicuous choice.
• Pure friction dissolves when someone sits with a user for an hour and fixes the six things that make the new tool annoying.

A force field without Core Drives tells you that resistance is strong. A force field with Core Drives tells you what to do on Monday.

Give away the decision, not just the reasons

This is the food-habits finding, operationalized.

Wherever there is a real decision inside the change, hand it to the group that has to live with it. The decision to change is usually already settled, so hand over a genuine one instead: sequencing, tooling within a shortlist, what the rollout order is, which of the old capabilities gets preserved.

Two conditions make this work rather than becoming theater. The choice has to be real, meaning you will accept an answer you did not want. And the group has to reach it together, out loud, in one room, because the mechanism is social as much as it is about ownership.

A steering committee that decides everything and then runs a listening tour gets neither CD4 nor CD5, and the listening tour is often more damaging than silence would have been, because it demonstrates that participation is decorative.

Remove one restraint before you add any push

A sequencing rule I hold to fairly strictly: before the announcement, kill something.

Find the single most-cited restraint from the diagnostic and remove it, publicly, before you ask anyone for anything. Retire the duplicate report the new system makes redundant. Extend the deadline that made the migration feel impossible. Announce the ramp period where being slow is expected and will not be measured.

This does three things at once. It lowers the actual restraining force. It provides disconfirmation in Schein’s sense, since the organization visibly just changed. And it converts the sponsor’s credibility from an assertion into evidence, which is the currency you will need for every subsequent ask.

Announcing a change and then asking what is in the way reverses the order, and the answers you get will be smaller than the truth.

Design the new equilibrium, not the new state

Refreezing, taken seriously, means arranging a set of forces that holds the new level without anyone pushing.

Ask what is still pulling toward the old level after the change lands. If the old system is still switched on, that is a force. If the metric that rewarded the old behavior still appears on someone’s compensation, that is a force. If the person who knows the old process best is still the person everyone asks for help, that is a force.

Any one of those left running will return the level to where it was, on a delay long enough that nobody connects the two events. Which is exactly the decay Lewin wrote the passage to explain.

The river version of refreezing is the useful one: you are not setting concrete, you are arranging banks.

Practical Steps: A Force Field Audit You Can Run This Week

Pick one change in your organization that has stalled. Not the biggest one. One where you can name the affected group.

1. Write the current level as a behavior, not an attitude. “Forty per cent of tickets are still logged in the old system” beats “the team has not bought in.” Attitudes cannot be observed; the level can.
2. Interview five people from the affected group, privately, with one question: what gets worse for you personally if this goes perfectly? Take notes verbatim. Do not argue with any answer.
3. Split what you heard into driving and restraining forces, and score each one from 1 to 5 on strength. Most lists come back lopsided in a way that surprises the sponsor.
4. Tag each restraining force with its Core Drive: CD8 loss, CD4 ownership, CD5 social, or plain friction. Anything you cannot tag is probably not yet specific enough; go back and ask again.
5. Find the gatekeeper. Not the approver. The person who can quietly decline and make it stick. Their restraint is worth more than the other four combined.
6. Remove the top restraint before your next communication, and say that you did it and why, without attaching an ask to the same message.
7. Give one real decision to the group, decided together, and accept the outcome.

Seven steps, roughly a week, and you will know more about why the change is stuck than any readiness survey will tell you.

Lewin’s Model Was the Beginning, Not the End

Lewin’s most quoted line is that there is nothing so practical as a good theory. It is a fine line and it is worth knowing that he first used it in 1943 while attributing it to a businessman, and only later adopted it as his own.

There is something fitting in that. The most famous sentence attributed to Lewin was not originally his, and neither was the most famous model.

What was his is better than both. A social system is a field of forces holding a level. The level is being produced continuously, so nothing has to be melted. And where you intervene determines what you end up with: push harder and you get the new level plus stored tension, remove restraint and you get the new level with the system intact.

He also left the one clean experiment. Give the group the decision and the behavior changes at five times the rate of giving them the argument.

Eighty years on, most change programs are still doing the lecture.

If you want the mechanism underneath the forces, meaning which motivations generate them and what each one responds to, that is what Octalysis is for, and my books work through it in full.

Start with the audit above. The next time someone in a meeting says people are resistant to change, you will have a better question ready, and asking it in front of the right people tends to change the meeting.

Frequently Asked Questions

What is Lewin’s change management model?

It is a three-stage description of planned change: unfreeze the current state, move to a new state, then freeze the new state so it holds. It comes from a short passage in Kurt Lewin’s 1947 paper “Frontiers in Group Dynamics” in Human Relations, where he described a successful change as including three aspects: unfreezing the present level, moving to a new level, and freezing group life on the new level.

Did Kurt Lewin actually create the unfreeze-change-refreeze model?

This is disputed. Cummings, Bridgman and Brown argued in Human Relations in 2016 that the packaged three-step model is a posthumous reconstruction that only became established as Lewin’s foundational contribution in the 1980s. Bernard Burnes disagreed directly in the Journal of Applied Behavioral Science in 2020, arguing that Lewin did develop the approach deliberately and that the difference from later presentations is mostly nomenclature. Both sides agree the three steps alone are a stripped-down version of a much larger method.

Who added the word “refreezing”?

Lewin wrote “freezing.” Cummings and colleagues trace the substitution of “refreezing” to a 1950 conference paper by Leon Festinger, who had been one of Lewin’s students at the Research Center for Group Dynamics. Lewin died in February 1947, so he had no opportunity to accept or reject the change.

What is force field analysis?

It is Lewin’s diagnostic method. You map the driving forces pushing a system toward change against the restraining forces holding it in place, and score each one, on the understanding that the current state is the balance point between them. It is the analytical half of Lewin’s work and it is usually omitted when the three steps are taught.

What is a quasi-stationary equilibrium?

It is Lewin’s term for a state that stays recognizably constant while the process producing it keeps running. He compared it to a river flowing at a given velocity in a given direction, and described social change as a change in the velocity or direction of that river. The implication is that the status quo is being actively maintained rather than sitting inert.

Should I increase driving forces or reduce restraining forces?

Lewin’s own answer was that reducing restraining forces is preferable as a rule. Both routes can reach the new level, but adding driving forces gets there with high tension in the system, which he associated with aggressiveness, emotionality and less constructive behavior, while reducing restraining forces reaches the same level with low tension. Most change programs do the opposite of what he recommended.

What did Lewin’s food habits experiment find?

In his 1943 study for the National Research Council’s Committee on Food Habits, groups of housewives either heard a nutritionist lecture about serving organ meats or discussed the question and reached a decision as a group. At follow-up about a week later, 23 of 44 in the group-decision condition had served the food, against 4 of 41 in the lecture condition. The widely quoted figures of 32 per cent versus 3 per cent come from Lewin’s own later summary and do not match the counts in the 1943 report.

Is Lewin’s model still relevant?

The three-step sequence has aged poorly, mainly because organizations no longer hold still long enough for a refreezing phase to make sense as a project stage. The underlying apparatus has aged well: force field analysis, the quasi-stationary equilibrium, the preference for removing restraint over adding push, and the finding that group decision outperforms expert instruction are all directly usable today.

What is the difference between Lewin’s model and Kotter’s 8 steps?

Kotter’s model came out of his own research on failed transformations and does not claim descent from Lewin, despite the frequent side-by-side mapping. The practical difference is directional: Kotter’s eight steps are largely about generating and organizing driving force (urgency, coalition, vision, communication, short-term wins) while Lewin’s contribution is the argument for working on the restraining side instead.

Did 70% of change initiatives really fail?

There is no evidence for that figure. It entered circulation through the opening sentence of Beer and Nohria’s 2000 Harvard Business Review article “Cracking the Code of Change,” which states it without citing a study. Mark Hughes traced the five most prominent sources in the Journal of Change Management in 2011 and found each either asserted it without evidence or cited another source that did, concluding there is no valid empirical basis for it.

Which Octalysis Core Drive explains resistance to change?

Usually Core Drive 8 (CD8): Loss & Avoidance, sharpened by Core Drive 4 (CD4): Ownership & Possession. What people resist is a specific projected personal loss of competence, status, headcount, or something they built, rather than the change in the abstract. Core Drive 5 (CD5): Social Influence & Relatedness supplies a third layer, because part of each person’s restraint is their expectation of what everyone else will do.

References

• Lewin, K. (1947). Frontiers in group dynamics: Concept, method and reality in social science; social equilibria and social change. Human Relations, 1(1), 5–41. DOI
• Lewin, K. (1947). Frontiers in group dynamics II: Channels of group life; social planning and action research. Human Relations, 1(2), 143–153.
• Lewin, K. (1943). Forces behind food habits and methods of change. In The Problem of Changing Food Habits: Report of the Committee on Food Habits 1941–1943, Bulletin of the National Research Council, 108, 35–65. National Academies
• Lewin, K. (1946). Action research and minority problems. Journal of Social Issues, 2(4), 34–46. Wiley
• Lewin, K. (1936). Principles of Topological Psychology. New York: McGraw-Hill. (Source of the formula B = f(P, E): behavior is a function of the person and the environment.)
• Lewin, K. (1951). Field Theory in Social Science: Selected Theoretical Papers (D. Cartwright, Ed.). New York: Harper & Row.
• Cummings, S., Bridgman, T., & Brown, K. G. (2016). Unfreezing change as three steps: Rethinking Kurt Lewin’s legacy for change management. Human Relations, 69(1), 33–60. DOI
• Burnes, B. (2004). Kurt Lewin and the planned approach to change: A re-appraisal. Journal of Management Studies, 41(6), 977–1002. Wiley
• Burnes, B. (2020). The origins of Lewin’s three-step model of change. The Journal of Applied Behavioral Science, 56(1), 32–59. DOI
• Kanter, R. M., Stein, B. A., & Jick, T. D. (1992). The Challenge of Organizational Change: How Companies Experience It and Leaders Guide It. New York: Free Press. (The “organization as an ice cube” criticism appears on p. 10.)
• Schein, E. H. (1996). Kurt Lewin’s change theory in the field and in the classroom: Notes toward a model of managed learning. Systems Practice, 9(1), 27–47. Springer
• Beer, M., & Nohria, N. (2000). Cracking the code of change. Harvard Business Review, May–June. HBR
• Hughes, M. (2011). Do 70 per cent of all organizational change initiatives really fail? Journal of Change Management, 11(4), 451–464. Taylor & Francis
• Kotter, J. P. (1996). Leading Change. Boston: Harvard Business School Press.
• Chou, Y. (2015). Actionable Gamification: Beyond Points, Badges, and Leaderboards. Yu-kai’s books

Related Reading

• Kotter’s 8-Step Change Model: Organizational Transformation
• The Transtheoretical Model: Prochaska’s Stages of Change
• Diffusion of Innovations: Rogers’s Adoption Curve
• Psychological Safety: Edmondson on Team Learning
• Psychological Ownership: The Psychology of “Mine”
• Motivational Interviewing: Miller and Rollnick on Behavior Change

Explore the full Behavioral Framework Library for 242 pillar guides, or go deeper with the complete Octalysis Framework.

WOULD YOU LIKE YU-KAI CHOU TO WORK WITH YOUR ORGANIZATION?