By Ritvars Mētra, Founder of ReadinessCompassThe framework selection meeting is a recognisable event. Someone has a preference from a previous employer, someone else has a certification, and a decision gets made that will shape how the next two years of change work is described.
It matters less than the meeting implies. The evidence base does not support a strong claim that any of these approaches outperforms the others, and the variables that actually determine whether change lands are mostly not framework choices at all.
Start with what the evidence supports
None of the mainstream frameworks has been validated as a whole. When Steven Appelbaum and colleagues reviewed fifteen years of literature against Kotter’s eight steps in the Journal of Management Development, they found support for most individual steps but no formal studies covering the model as a whole, and concluded that its popularity owed more to its usable format than to scientific consensus.
Lewin’s three-step model — the ancestor of most of the others — turns out to be a posthumous reconstruction rather than something he proposed, according to Cummings, Bridgman and Brown’s analysis in Human Relations. ADKAR is a practitioner model developed from benchmarking rather than a research programme.
So the honest framing is not “which one is proven?” It is: these are organising devices with different shapes, and the question is which shape fits the problem in front of you.
What each is actually good at
| Operates at | Strongest for | Blind to | |
|---|---|---|---|
| ADKAR | The individual | Diagnosing which gap a specific group has — and the Knowledge/Ability distinction | Capacity, incentives, sequencing, process misfit |
| Kotter | The organisation | Coalition, momentum, and naming conditions leadership must create | Fixed dates, packaged software, anything the change leader does not control |
| Agile change | The delivery cadence | Iterating where the answer is genuinely unknown; short feedback loops | Big-bang cutovers, regulatory dates, changes that cannot be partially adopted |
Read the right-hand column rather than the middle one. Frameworks are chosen on their strengths and fail on their blind spots, and the blind spots are more predictive.
Three questions that decide it
1. Is the date fixed, and who fixed it? If a contract, licence expiry or regulator set the date, any framework that assumes the change leader controls pace will decay into ceremony. Kotter’s sequence in particular presumes you can wait for conditions to be reached. On a fixed-date programme the conditions become unrecorded risk instead — which is what makes falsifiable readiness criteria more valuable than a phase model.
2. Is use mandatory or discretionary? This is the most underrated distinction in the field. If the old system is switched off, capability is the whole problem and an individual-level model serves well. If people can carry on as before — a CRM, a collaboration tool, an AI assistant — then incentives and manager behaviour dominate, and the whole shape of the plan has to change.
3. Is the binding constraint individual or structural? Ask what would happen if every affected person were fully willing and fully skilled. If the change would then land, you have an individual problem and ADKAR will help. If it still would not — because of capacity, dependencies, incentives or a process that does not work — no individual-level framework will reach it.
That third question is the one worth asking out loud in the selection meeting. It changes the conversation from preference to diagnosis.
The case for not choosing one
Committing to a single framework has a real cost: it determines what you will notice. A programme run entirely in ADKAR language will generate individual-level diagnoses because those are the only outputs the instrument produces. A programme run entirely in Kotter language will produce coalition and vision work regardless of whether the constraint is capacity.
The practical alternative is unglamorous and works better: use ADKAR to locate individual gaps by group, use Kotter’s questions to test leadership conditions, use agile cadence where the answer is genuinely unknown, and hold all of them loosely enough to notice when the real constraint is somewhere none of them looks.
Certification and consistency have genuine value in a large organisation — shared vocabulary is not nothing. Just do not confuse having a common language with having a diagnosis.
What matters more than the choice
On most programmes, four things predict the outcome better than the framework does:
- Whether managers visibly do the new thing themselves.
- Whether people have the hours, given everything else landing that quarter.
- Whether the measurement and incentive system rewards the new behaviour or the old one.
- Whether anyone has watched a real person attempt the real task, unaided, before go-live.
Not one of those is determined by which framework is on the cover of the plan. A programme that gets all four right will succeed under any of them, and a programme that gets them wrong will fail under all of them — which is roughly what the research on models actually says.
More on frameworks and evidence in the Change Readiness Hub, or see how a Readiness Diagnostic Sprint identifies the binding constraint before a method is chosen.