The Influence Model
A household decides to start recycling. Everyone agrees with it; nobody argues.
Three weeks later almost nothing is being separated, and no single person is at fault. There is one bin, and it is in the kitchen, so separating means walking outside every time. Nobody is actually sure whether the greasy pizza box counts. Nobody else in the house is visibly doing it either, so the person who tries feels slightly ridiculous. And the bin gets emptied on the same day whether or not anything was sorted, so nothing depends on it.
Four different things are missing, and any one of them is enough to stop the behaviour on its own. Agreement was never the constraint.
The four conditions#
McKinsey’s influence model names those four, and its central claim is that they are conjunctive: all four must hold, and the weakest one governs.
Role modelling. I see the people around me, especially the ones with authority, actually doing this. Not endorsing it, doing it. This is the condition with the largest gap between how leaders think they are doing and how they are perceived, because it is judged on behaviour under pressure rather than on stated support.
Understanding and conviction. I know what is being asked of me and I believe it is worth doing. Conviction is the part usually skipped, because announcing what changed is quick and explaining why it matters is not.
Skills. I am actually capable of the new behaviour. A change that requires a capability nobody has is not a change, it is a wish, and people who cannot do the new thing will quietly keep doing the old one rather than admit it.
Formal mechanisms. The systems, processes, tools and incentives make the new behaviour the easy path. This is the condition most under a team’s direct control and the one most often left alone, because it is the least satisfying to work on and the most effective.
Why “everyone agreed” is not evidence#
Agreement only tests one of the four. It is entirely consistent for a change to be universally supported and completely absent, and that combination is the normal state of an announced change rather than an unusual one.
This is also why repeating the announcement does not work. Restating it louder addresses understanding, which was rarely the problem, and leaves the other three untouched. If a change has been announced twice and is still not happening, the second announcement is evidence that the missing condition is one of the other three.
How to tell which one is missing#
Take one specific behaviour, not a theme, and ask four questions of the people who are supposed to do it.
| Condition | The question to ask |
|---|---|
| Role modelling | Can you name someone senior who does this, and when you last saw it? |
| Conviction | What do you think happens if we do not do this? |
| Skills | Walk me through how you would do it today |
| Mechanisms | What is the fastest way to get this done right now? |
The last question is the highest yield. If the fastest way to get the work done is still the old way, you have your answer, and no amount of encouragement will beat a path that is genuinely quicker.
A useful shortcut: ask someone who tried and stopped. They can usually tell you exactly which condition failed, in one sentence, and they are rarely asked because their stopping was read as a lack of commitment.
When it goes wrong, the temptation on finding all four weak is to launch four workstreams. Fix the cheapest one first, usually a mechanism, and re-check. Conditions are not independent: removing the old path often produces the role modelling for free, because now everyone is visibly doing the new thing.
Where this stops#
It is a checklist, not a sequence. It tells you which conditions are missing and offers nothing about which to fix first, how much each costs, or how they interact. That judgement is yours.
It assumes the change is worth making. Every condition is about adoption. If people are not doing it because it is a bad idea, the model will diagnose a conviction problem and recommend explaining harder, which is exactly the wrong response. Before running it, check that the resistance is not information.
It is about behaviour, not capability at the organisational level. For why a change does not fit the organisation at all rather than failing to be adopted by people, 7S is the better instrument. The two pair naturally: 7S tells you which element is misaligned, and if that element turns out to be skills, style or staff, this model tells you what to do about it.
How we use it here#
Two moments, and one of them is before rather than after.
Before introducing a new practice, walk the four conditions and decide in advance who models it, how conviction gets built, whether anyone needs to learn something, and above all what mechanism makes it the easy path. A practice introduced without a mechanism is a practice that depends on everyone remembering, which is a plan with a known expiry date.
When something we adopted is quietly not happening, run the four questions before restating the rule. It will usually be mechanisms, and the fix will usually be removing the old path rather than adding a reminder or a check.
The pattern to watch for in ourselves is the reflex to respond to a lapsed practice by writing it down again. Documentation is understanding. It is one condition out of four, it is the cheapest to produce, and that is exactly why it is the one we reach for.
Related#
- McKinsey 7S diagnoses which organisational element is fighting a change; this covers the people half.
- The Framework has the cascade checks for when a decision is not reaching anyone’s week.
- Equipping People is the skills condition in practice.
- Structured Problem Solving is the wider method.