The intervention was chosen before the diagnosis
Watch what an organisation does when it discovers a behaviour problem.
People are not reporting near misses. People are not using the new system. Managers are avoiding difficult conversations. Nobody is completing the compliance module properly. Whatever the problem is, the response tends to come from a small and remarkably stable set. Training. A communications campaign. A dashboard. Occasionally a workshop.
I do not think this is because diagnosis keeps arriving at the same answer. I think it is because those are the interventions that can be bought.
The menu is set by procurement
Training has vendors. It has a market, a price per head, a budget line, and a procurement pathway that somebody in the organisation already knows how to walk down. The same is true of communications campaigns and of dashboards. Each one can be specified in a document, competitively tendered, delivered on a date, and invoiced.
Now consider the interventions that a behavioural diagnosis frequently points toward. Remove a step from a process. Change who has to sign something off. Alter the order in which two things happen. Stop asking for a piece of information nobody uses. Change what gets counted in a bonus. Move a decision from an individual to a pair.
None of those has a supplier. There is no market for removing an approval step. There is no line item called “we changed the sequence.” They are often cheap or free, which sounds like an advantage and is in fact a serious commercial disadvantage, because an intervention with no cost has no budget, and an intervention with no budget has no sponsor, and an intervention with no sponsor does not happen.
So the set of interventions available in practice is not the set that the evidence base supports. It is the intersection of that set with the set of things that can be purchased. And the second set is much smaller and shaped by entirely different forces.
Running the diagnosis on the wrong person
There is a slightly uncomfortable move available here, which is to apply the field's own diagnostic tools to the people commissioning the work rather than to the people whose behaviour is nominally the problem.
If I ask what is stopping an organisation from selecting the intervention that the diagnosis actually implies, the answer decomposes reasonably cleanly.
Capability is a real constraint. Many of the people who commission behaviour change work have never been trained to distinguish a capability barrier from a motivation barrier, so training gets selected because training is the intervention that addresses the barrier most people intuitively assume, which is that the person does not know how.
Opportunity is a larger constraint than it looks. The person commissioning the work usually does not control the process they would need to change. A head of learning and development can buy a course. They cannot remove an approval step in a system owned by finance. The intervention that is within their gift is not the intervention the diagnosis recommends, and they are not being unreasonable in choosing something they can actually deliver.
Motivation is the constraint nobody says out loud. A training programme produces a completion rate, and a completion rate is a number that can be reported upward before the outcome is known. Removing a process step produces no artefact and no attributable win, and if the outcome improves six months later the credit is diffuse. When a person is judged on visible activity within a budget cycle, buying something is the rational move even when it is the less effective one.
I find that third point the most persuasive and the hardest to write about without sounding contemptuous, which I do not intend, because the incentive is genuinely there and people are responding to it sensibly.
What this does to diagnosis
The consequence is that diagnosis stops being the step that determines the intervention and becomes the step that justifies one.
You can see the fingerprint of this in how work gets scoped. A brief that arrives already naming the deliverable has done this. So has a brief that asks for a behavioural framework to be applied to a programme that has already been designed and funded. In both cases the analysis is being commissioned after the decision it was supposed to inform, which means the most valuable thing behavioural science offers has been converted into a validation exercise.
This also explains a pattern that otherwise looks strange, which is that organisations often run the same intervention repeatedly against a problem that never resolves. Annual training on a behaviour that has not changed in five years is not a failure of the training. It is what happens when the response is determined by the procurement route rather than by the barrier, so the same lever gets pulled regardless of what the last pull achieved.
What would actually change it
The honest answer is that the fix is not a better diagnosis. It is getting a different person into the room earlier.
If the diagnosis happens before a budget has been allocated to an intervention type, the space of possible answers is still open. If it happens afterward, the space has already closed and no amount of analytic quality reopens it. That timing point seems to matter more than anything about the method itself.
The second thing that would help is for practitioners to price and scope the diagnosis as a deliverable in its own right, rather than as the free preamble to a paid intervention. As long as diagnosis is bundled in front of something billable, the person doing it has a quiet interest in it concluding that the billable thing is needed. That conflict does not require anyone to behave badly. It just needs to be structurally present, and it usually is.
The limit of what I am claiming
I should be clear that this is an argument from observation and from the structure of the incentives, not from evidence I have gathered. I have not counted interventions against diagnoses across a sample of organisations, and as far as I know nobody has.
It would not be a hard study to run. Take a set of behaviour change programmes, code the stated barrier, code the selected intervention, and check whether the selection is better predicted by the diagnosis or by whether the intervention type had an existing supplier relationship. If procurement explains more variance than diagnosis, that is worth knowing, and it would be a fairly damning thing to know.
Until somebody does that, this is a hypothesis. But it is one I have not been able to stop noticing.