The Thread Through the Project

Teams tend to buy behavioural science as a single appointment. A research method here, a nudge review there, an adoption rescue when things go wrong. Each is treated as a discrete service, dropped into one phase and lifted out again. But behaviour does not sit in one phase. It runs through the whole project like a thread, changing shape as the work changes, and if you only consider it once, you quietly inherit the behavioural risks of every stage you skipped.

The thread has seven stages: discovery, research, strategy, design, testing, rollout, scale. Each asks a different behavioural question and produces a different output. The skill is not knowing a method for each. It is knowing which question belongs to which moment, because the most common and most expensive failure in the whole practice is bringing the right contribution at the wrong time.

The first job of the early stages is to replace the story the organisation tells itself with what people actually do. In discovery, you find the gap between the official journey and lived practice, and you pay special attention to the workarounds, because the real service is often the workaround rather than the designed process.

Research then explains why that behaviour happens in context: what makes the current behaviour rational, what the person believes they risk by changing, which moments carry the most weight. And strategy makes the hard choice. Which few behaviour shifts actually matter?

The later stages turn diagnosis into conditions, evidence and habit, and their character is different because this is where the old behaviour fights back. In design, the clever part is never the intervention itself. It is the fit between behaviour, context and intervention. A nudge that ignores the context is just a decoration. Design the conditions around the action, not only the prompt at the moment of it.

Testing then checks whether action actually changes, under real friction, not whether people liked the idea in a calm room. Preference is not behaviour, and a good test is not there to prove the idea works. It is there to find out where reality disagrees, while disagreement is still cheap.

The final stage is the one most often misunderstood. Scale is not copying a working solution into more places unchanged. It is adaptation with enough fidelity to preserve the behavioural mechanism that made it work. The mechanism travels. The surface details often have to change to fit each new context, and a team that copies the surface while losing the mechanism has scaled the wrong thing.

And the real deliverable of scale is capability: what remains when the expert leaves the room. A toolkit sitting unused is not capability. A toolkit that shows up in real decisions is. The test is never whether the tool exists. It is whether it changes what people actually do when you are no longer there to remind them.