Lewin's Change Model
Lewin's change model describes a planned change in a group's behaviour in three phases: unfreezing the current level, moving the group to the new level, then freezing that level so that it holds. Kurt Lewin set it out in 1947. Edgar Schein detailed the work that each phase demands. The business analyst uses it to plan the adoption of a solution and to set the exit criterion of each phase.
Purpose
Lewin's change model is a three-phase framework for taking a group from one level of behaviour to another and keeping it there. It addresses a problem that Lewin states in his 1947 article: a rise in a group's performance obtained by a "shot in the arm" often falls back to the previous level. The objective therefore includes the permanence of the new level.
The model supports a decision on how to steer the change: which actions to take before, during and after go-live so that the change holds. The deliverable is a change management plan. For each phase, the plan names the actions, the force or process they target and the exit criterion: an observable fact that allows the move to the next phase. This plan is our own format. With it, the business analyst decides when moving can start: as soon as unfreezing meets its exit criterion. The analyst also decides when the adoption workstream can close: when the measurement taken at the end of the target period shows that the new level holds.
The model extends force field analysis. Force field analysis maps the driving and restraining forces that hold a level of behaviour at one moment. The change model sequences the move from one level to another over time.
Usage
When to use it
- Adoption workstream of an IT project: structure the deployment in three phases with one exit criterion per phase.
- Change that falls back after go-live: check whether unfreezing took place and whether freezing was designed.
- Training given without lasting effect: design freezing into the rules, the tools and the norm of the group that receives the trained staff.
- Habit shared by a team: have the group change the norm in a session rather than train people one by one.
- Resistance in the face of figures that nobody disputes: address learning anxiety through psychological safety before adding pressure.
- Department that already measures its current performance: its figures are where unfreezing starts.
When not to use it
- Change not yet decided: the model steers a move already chosen. Assess its feasibility first with force field analysis.
- Transformation of the whole organisation over several years: three phases sequence neither the guiding coalition nor the vision nor the short-term wins. Use Kotter's change process.
- Technical change with no effect on practice: no group behaviour changes. A release plan and a deployment strategy are enough.
Description
A level holds because forces balance
Lewin treats a group behaviour as a level: a production rate, a consumption level, a processing time. This level is a quasi-stationary equilibrium: the forces that push it equal the forces that hold it back. It moves as soon as one of them changes.
A group that lives at one level for a long time creates institutions with a stake in keeping it there. The group norm holds each member back: the more value the norm has for the members, the more each one resists departing from it. The group first mocks a member who strays too far from it, then excludes that member. Lewin notes that it is usually easier to change people formed into a group than to change any one of them separately. If the group norm changes, the force that held the individual back carries the individual toward the new level.
The three phases
Lewin writes that a successful change includes three aspects: unfreezing the present level if necessary; moving the group to the new level; freezing group life on the new level. Permanence implies making the new force field relatively secure against change.
A change that runs against neither a habit nor a group norm can therefore go to moving without unfreezing. The analyst describes the target level in the same terms as the current level: an observable behaviour and the period during which it must hold, for example until the next version of the tool. The analyst quantifies the behaviour where possible.
Schein restates the model.
| Phase | Lewin (1947) | Schein (1995) |
|---|---|---|
| Unfreezing | unfreezing (if necessary) | unfreezing |
| Moving | moving | changing |
| Freezing | freezing | refreezing |
Unfreezing: making the data acceptable
Lewin cites the catharsis that Allport considers necessary before a prejudice can be removed. He adds that breaking a group's shell of complacency sometimes requires a deliberate emotional stir-up. Schein breaks unfreezing down into processes that must each be present to some degree before the motivation to change can arise.
- Disconfirmation: data contradict the group's expectations or hopes. For Schein, disconfirmation is a primary driving force in the quasi-stationary equilibrium. The group can ignore it, dismiss it as irrelevant, blame others for it or deny it.
- Survival anxiety: the group accepts the data and connects them to what matters to it. It feels that it will fail to meet its needs or reach its goals unless it changes.
- Psychological safety: it overcomes learning anxiety. Learning anxiety is the fear of losing one's effectiveness, self-esteem or identity by admitting that something is wrong.
For Schein, learning anxiety is the fundamental restraining force. It can rise in direct proportion to the disconfirmation. Without enough psychological safety, the group denies the data, feels no survival anxiety and does not change. A project team that answers resistance with more figures and more pressure therefore strengthens the defence it set out to overcome. The change agent balances the threat in the data against the safety that lets the group accept them.
Schein lists the means that change agents use to create this safety: working in groups, setting up parallel systems that relieve the pressure of daily work, offering practice fields where errors are accepted, giving a positive vision, breaking the learning into manageable steps and coaching people on the job. On an IT project, the training environment where staff make mistakes without touching live data is a practice field. CATWOE analysis brings out each stakeholder's worldview and shows what each group holds on to.
Moving: redefining and choosing the method
Lewin distinguishes two ways to move a level: adding forces in the desired direction or diminishing the opposing forces. In the first case, the new level comes with high tension. In the second, it comes with low tension. Lewin judges the second preferable as a rule. Schein adds that driving forces usually exist already. An added driving force often produces an immediate counterforce. Restraining forces are harder to reach: according to Schein, they are often personal defences or group norms embedded in the culture.
For Schein, moving is a cognitive redefinition. The motivated person receives new information that produces one or more of these effects: a word changes meaning (semantic redefinition), a concept widens (cognitive broadening) or the yardstick for judging shifts (new standards of judgment).
The new information arrives by two routes. Identification works through a role model, for example a mentor or a colleague who has already moved to the new way of working. It is fast, but its solutions hold poorly when they do not suit the person. Scanning works through reading, visits, conversations and trial and error. It produces solutions that the person has found for themselves. Schein attributes to identification the failure of many consulting assignments and benchmarking efforts: the consultant or the "best practice" is the role model. The organisation then adopts a solution foreign to its culture for a while.
Moving proceeds by trial. The person tests the new categories through new behaviour. The trial reinforces them or sets off a new cycle of disconfirmation and search. In the plan, a missed exit criterion therefore sends the team back to the previous phase. The form of the move (direct cutover, parallel running, pilot deployment or phased deployment) is a matter for change deployment strategies.
Freezing: securing the new field
Lewin names two means. The first is an organisational set-up equivalent to a stable circular causal process: the new behaviour maintains the conditions that reproduce it. The second is group decision. Group decision has a freezing effect that comes partly from each person's tendency to stick to a decision and partly from commitment to the group. Lewin adds that decision does not explain everything. A housewife who has decided to use more milk can place a standing order with the milkman. The order keeps up consumption without a new decision.
Schein distinguishes two kinds of refreezing. Personal refreezing requires the new behaviour to fit the rest of the person's behaviour and personality. Otherwise the behaviour sets off new rounds of disconfirmation that often lead the person to unlearn what they have just learned. Schein draws a rule from this: prefer scanning to identification so that each person keeps a solution that suits them.
Relational refreezing works through a session in which the group states the implicit norm behind the old behaviour, then sets a new one. Schein recommends training the whole group that shares this norm. His textbook case is a programme that teaches supervisors to delegate, then sends them back into an autocratic culture. On an IT project, this group also includes the managers who approve the work, the neighbouring departments that receive the files and the long-serving staff who train newcomers.
The project team designs freezing before go-live. A team that deals with it afterwards finds the old workflows still open: the paper binder, the parallel spreadsheet, the printed copy that management still asks for. The plan provides for closing each of them.
Group decision
Lewin presents group decision as a change procedure that serves all three phases. Housewives in a Midwest town either heard a lecture on fresh milk or discussed it until they decided as a group to use more of it. Both groups had the same amount of time. After two and then four weeks, group decision came out ahead. In the maternity ward of a state hospital in Iowa, mothers received either individual instruction of about twenty-five minutes or the same amount of time spent in discussion and decision in groups of six. After four weeks, every mother in the decision group gave her baby the recommended dose of cod liver oil.
Lewin explains the gap as follows: a lecture or a discussion creates motivation; the decision links it to action. Citing Bavelas, he also reports an exceptionally favourable case of permanence: the production level that a factory team set through three group decisions held for nine months. He notes that the procedure is easy to apply in some cases. In others it requires skill and certain conditions. Managers who rush into a factory to raise production by group decisions are likely to fail: each case demands careful diagnosis.
Running the model on a project
The steps are our own construction from Lewin and Schein.
- Describe the current level and the target level
Both as observable and quantified behaviours. The target level also states the period during which it must hold. - Map the force field at the current level
Identify the restraining forces that come from a habit or a group norm. - Decide whether unfreezing is necessary
It is necessary whenever a habit or a group norm holds the current level. - Plan unfreezing
The data that disconfirm expectations, what each group connects them to, the psychological safety measures. - Plan moving
The restraining forces to reduce, the expected redefinitions, the form of the deployment. - Design freezing before go-live
The rules, tools and workflows that reproduce the new behaviour. The old workflows to close. Training for the whole group that shares the norm. A session in which that group states the norm. - Set one exit criterion per phase
For freezing, a measurement taken at the end of the target period.
AI considerations
A language model helps first with unfreezing. From the department's statistics, tickets or complaints, it assembles the disconfirming data and rewords them for each audience: management, staff, neighbouring departments. In procedures and specifications, it finds the rules and workflows that still reward the old behaviour. The team starts from this list to design freezing. During deployment, the model sifts the free-text answers of a survey and flags expressions of fear, loss of control or shame at making mistakes: the traces of learning anxiety.
AI does not measure a team's psychological safety. It cannot tell whether a group norm holds the current level: both judgements require knowing the people and the department's history. A solution that AI proposes as "best practice" plays the role that Schein gives the consultant: a role model whose solution does not necessarily suit the department's culture. A team that adopts it without scanning prepares a refreezing that will not hold. Survey verbatims that let a reader recognise their author (role, department, turn of phrase) are personal data under the Federal Act on Data Protection (FADP). A team that feeds them into an external AI service exposes statements made under the promise of anonymity.
Examples
A cantonal debt enforcement office with 22 staff replaces its paper files with an electronic case-file management tool. Its change management plan gives each phase an observable exit criterion. The target level is a lead time of at most 3 working days between the debt collection request and the dispatch of the payment order. This lead time must still hold three months after the migration of the existing files.
| Phase | Actions | Force or process targeted | Exit criterion |
|---|---|---|---|
| Unfreezing | Present the measured lead time between the debt collection request and the dispatch of the payment order: 6 working days at the office, 2 at the neighbouring office that already uses the tool. Link this lead time to the creditor complaints received at the front desk. Open a practice environment with mock files for three weeks without individual assessment. | Disconfirmation. Survival anxiety. Psychological safety. | In a session, staff cite the lead-time gap themselves. Each person has processed 5 mock files. |
| Moving | In a session, the team sets its own rules for naming and filing documents instead of adopting those of the neighbouring office. "Complete file" now means "every document in the tool". Phased deployment: new requests first, the existing files afterwards. The tool imports the data of requests received electronically so that nobody enters them twice. | Scanning. Semantic redefinition. Restraining force reduced: double entry. | For four weeks, 100% of new requests have been entered in the tool. No document from a new file goes into a paper binder. |
| Freezing | Remove the binders from the offices. The section head signs off in the tool. The tool sends the payment order as soon as the file is complete. Also train the section heads, the front desk staff and the apprentices. | Stable circular causal process. Relational refreezing. | Three months after the migration of the existing files: lead time of at most 3 working days, no paper file reopened. |
The team measures the freezing criterion three months after the migration of the existing files: on go-live day the new level is reached with nothing yet to show that it will hold.
Visualisations
On a project, the levels diagram is useful when each arrow at the new level is labelled with the name of a mechanism: a rule, a workflow, a norm. A new level without a named arrow has no freezing.
The plan fits in a table.
Cost
| Dimension | Level | Rationale |
|---|---|---|
| Preparation | Medium | Measure the current level, map the force field, gather the disconfirming data and identify the group that shares the norm. |
| Execution | High | The phases cover the whole deployment up to the freezing measurement. Relational refreezing requires training the whole group that shares the norm. |
| Documentation | Low | A one-page change management plan and the end-of-period measurement. |
Tools
The change management plan fits in a spreadsheet or a page in the project's collaboration space: one row per phase with the actions, the exit criterion, the owner and the date. The project team updates it at each phase transition. The team draws the levels on a physical or online whiteboard (Miro, Mural, FigJam) in the same session as the force field. The disconfirming data come from the tools already in place: business application statistics, the ticketing tool, the complaints register. A short anonymous survey in Microsoft Forms or LimeSurvey tracks learning anxiety. The team runs it before go-live and again at the end of the target period.
Sources
- 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: the article in which Lewin defines the three aspects of a successful change and the requirement of permanence. It also presents group decision as a change procedure.
- 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: the three processes of unfreezing, cognitive redefinition through identification or scanning and the two forms of refreezing.
- Schein, E. H. (1995), Kurt Lewin's Change Theory in the Field and in the Classroom: Notes Toward a Model of Managed Learning, MIT Sloan School of Management, Working Paper 3821: the working version of the 1996 article, in open access.

