Lewin's change model - unfreeze, change, refreeze

Change management requires structured processes to succeed. Here’s how we approach process improvement software for organizational change.

Here’s something that drives me a little crazy. Everyone talks about “digital transformation” like it’s a switch you flip. Buy some software, roll it out, done. But McKinsey’s own data says roughly 70% of change initiatives fail. That’s not a rounding error. That’s a pattern.

Kurt Lewin figured this out back in the 1940s. He was a social psychologist, not a management consultant, and maybe that’s why his model still holds up. He wasn’t trying to sell anything. He was trying to understand how humans behave when you ask them to do something differently.

His idea? You can’t just change things. You’ve got to melt the old shape first, pour the new one, and then let it set. Unfreeze, change, refreeze. Simple to say. Brutally hard to do.

Why Lewin’s model still matters

There’s a reason this framework from the 1940s keeps showing up in business schools and boardrooms. It works because it respects how people think.

Most change management models focus on the mechanics of change, the project plan, the timeline, the deliverables. Lewin’s model starts somewhere different. It starts with the question: why would anyone go along with this?

That’s the part leaders skip. They assume the logic of the change is enough. “We’re losing market share, so obviously we need to restructure.” But logic doesn’t move people. Proof does. Emotional buy-in does. Addressing the fear that someone’s job might disappear, that matters.

I learned this the hard way at Tallyfy this pattern play out hundreds of times. An operations leader buys workflow software expecting instant improvement. But the team hasn’t been unfrozen yet. They’re still clinging to their spreadsheets and email chains because nobody explained why the old way was broken.

here’s the thing that connects directly to where business is heading right now: If you automate a broken workflow with AI, you don’t get efficiency. You get faster chaos. Lewin’s model is more relevant than ever because it forces you to fix the foundation before you build on it.

Stage 1 - Unfreeze the status quo

Your team loves things the way they are. Can you blame them? Change is uncertain. Maybe it leads to layoffs. Maybe it makes their job harder. Maybe the whole thing’s a waste of time and everything goes back to normal in six months anyway.

Even the executive team can be skeptical. Why rock the boat when revenue’s steady?

Unless you get everyone on board, you’re dead in the water. So step one in Lewin’s model is to unfreeze. Break the grip of the status quo.

How? With evidence. Hard, uncomfortable evidence.

• Sales declining quarter over quarter
• Net promoter scores dropping
• Operational costs climbing while output stays flat
• Competitors moving faster on the same problems

Every time we onboard a new team, the same issue surfaces with workflow automation, the organizations that succeed at unfreezing don’t just present data. They make it personal. They show individual teams how the current process wastes their time. Not the company’s time. Their time.

Even with strong evidence, you’ll hit resistance. People will ask reasonable questions:

• What does this mean for my role?
• How will this change fix the actual problem?
• What if it makes things worse?
• Won’t this just blow over?

You’ve got to answer every single one. Not with corporate platitudes. With specifics.

Here’s a rough sequence that works:

The change management process we’ve seen work best treats unfreezing as its own project, not a preamble to the real work.

Stage 2 - Make the change

This is where people get energized. And also where they get reckless.

Your team’s bought in. They understand why change is needed. Now what? You experiment. But you don’t bet the farm on a single approach.

Prosci’s research across thousands of organizations shows that the most successful change efforts start small. Pilot programs. Controlled experiments. A/B comparisons across teams.

At Tallyfy, we’ve observed something interesting. Organizations running 50 or more workflows simultaneously discover what works by comparing results across teams instead of betting everything on one big rollout. The parallel approach sounds counterintuitive, running multiple experiments at once, but it’s faster and safer than sequential testing.

Don’t just hand down changes from the top, either. Sometimes the best ideas come from the people doing the work every day. A shop floor employee might see a bottleneck that’s invisible from the C-suite.

Here’s how this stage typically plays out:

One thing I’d add: document everything during this phase. Not for bureaucracy’s sake. Because when it’s time to refreeze, you’ll need proof that the new way is better. Without documentation, you’re asking people to trust your memory. That’s a bad bet.

Example Procedure

Employee Onboarding

1HR - Set up payroll and send welcome email

2IT - Order equipment and set up workstation

3Office Manager - Prepare physical workspace

4IT - Create accounts and system access

5HR - Welcome meeting and company orientation

+3 more steps

View template

Example Procedure

New Hire Orientation

1Before arrival HR: Send new employee email and company handbook

2Before arrival Manager: Send new employee email and create work-plan for month 1-3

3Before arrival IT: Set-up desk and computer

4First day HR: Meet new employee and introduce manager, set up tax forms

5First day Manager: Introduce employee to department, begin training

+10 more steps

View template

Example Procedure

Quarterly Strategic Planning & Goal Setting Workflow

1Revisit annual plan goals

2Break down goals into smaller chunks

3Review budget and benchmarks

4Create action steps and benchmarks

5Set expectations and timelines

+2 more steps

View template

Stage 3 - Refreeze the new normal

This is where most change initiatives quietly die. The excitement fades. The project team moves on. And six months later, everyone’s back to doing things the old way.

Refreezing is about turning experiments into habits. It’s the least glamorous stage, and probably the most important one.

Feedback we’ve received from fractional COOs working across multiple organizations confirms something I’ve suspected for years: the refreeze phase takes roughly twice as long as most leaders expect. The change phase feels like the hard part, but refreezing, where new habits become automatic, is where initiatives actually succeed or fail.

So how do you make the new way stick?

Build it into the system. Don’t rely on people remembering to follow the new process. Use workflow software that makes the new way the path of least resistance. When a process is digitized in Tallyfy, there’s no question about what happens next. The system tells you.

Incentivize the right behavior. Whether that’s recognition, bonuses, or simply making the old way harder than the new way, you need structural reinforcement.

Build verification checkpoints. The organizations that succeed at refreezing are the ones that build check-ins into their workflows. Not micromanagement. Visibility. There’s a difference.

Once the changes are embedded company-wide, you can call the initiative a success. Until the next one, anyway. Change isn’t a one-time event. It’s a muscle.

Force field analysis - Lewin’s other big idea

Most people know Lewin for unfreeze-change-refreeze. But his force field analysis is just as useful, and it pairs beautifully with the three-stage model.

The concept is simple. For any change, there are driving forces pushing toward it and restraining forces pushing against it. Draw them out. Literally. Put the change in the middle, drivers on one side, restraints on the other.

This does two things. First, it makes abstract resistance concrete. “People don’t like change” becomes “the finance team is worried about retraining costs” or “the VP of Sales thinks this will slow down Q3.” You can address specifics. You can’t address vibes.

Second, it shows you where to focus. Sometimes it’s easier to reduce a restraining force than to increase a driving force. If the finance team’s worried about retraining costs, maybe you show them a pilot that proves ROI before asking for budget. That’s unfreezing in action.

When Lewin’s model falls short

I’d be dishonest if I didn’t mention the criticisms. Some scholars, including this paper from Human Relations, argue that Lewin never actually formalized the three-step model the way it’s taught today. The clean “unfreeze-change-refreeze” framework may have been shaped by others after his death.

There are practical limitations too. The model assumes change is linear and that you reach a new stable state. But what about organizations in constant flux? If you’re refreezing every quarter, are you really refreezing at all?

For continuous change environments, the ADKAR model might be more practical because it focuses on individual readiness rather than organizational states. And Kotter’s 8-step model gives you a more granular playbook if Lewin’s three stages feel too broad.

My take? Lewin’s model is best as a thinking tool, not a rigid process. Use it to diagnose where your organization is stuck. Are people resisting because you haven’t unfrozen properly? Are changes not sticking because you skipped the refreeze? Those questions alone are worth the framework.

Making it work with modern tools

Here’s where I’ll tie it all together. The biggest challenge with Lewin’s model has always been the refreeze stage. Getting people to do something new for a week isn’t hard. Getting them to do it forever? That’s a different problem.

This pattern drove every design decision in Tallyfy. When you turn a process into a trackable workflow, refreezing happens by design.

The new process isn’t a suggestion people can ignore. It’s the system they work in. Every step has an owner, a deadline, and visibility. You don’t need to remind people. The workflow does it.

And with the rise of AI, this matters more than ever. If you’re going to hand processes to AI agents, those processes need to be defined, tested, and stable. You can’t hand an AI a mess and expect magic. You need to unfreeze, change, and refreeze first. Then automate.

That’s the sequence. Fix the process. Prove it works. Lock it in. Then scale it with technology.

Anything else is just making the mess bigger, faster.