Behavioural science explains why combining signals produces a lift greater than the sum of the parts
One signal on its own is rarely enough. Combining several beats relying on any single one.
BJ Fogg's model boils behaviour down to one line: Behaviour = Motivation × Ability × Prompt. All three have to be present at the same moment, or nothing happens.
That's why a single proxy, like a click, can't tell the whole story.
People act when three things line up, not before
Motivation – how much they want to do it right now. Wanting a new sofa "eventually" isn't the same as wanting one tonight.
Ability – how easy it is to do right now. Price, time, friction, whether they've got their card to hand.
Prompt – something has to trigger the action in that moment. An ad, a notification, a friend mentioning it
The bit that matters for a marketing audience: it's multiplication, not addition.
A perfect prompt to someone with zero motivation goes nowhere. You need all three at once, not two out of three.
So what: this is exactly why targeting on one signal, a single page visit, a single keyword, keeps missing.
That signal might tell you motivation is there, but says nothing about whether the person's in a position to act or whether anything's about to prompt them. It's a partial read on a formula that needs all three inputs to line up.
Sources: BJ Fogg, Tiny Habits
Netflix – Predictive Signals at Scale
Netflix is the commercial proof it works at scale, one of the biggest personalisation engines in the world, built and run on exactly this logic.
Netflix's Help Centre states outright: "The recommendations system does not include demographic information (such as age or gender)."
No age, no gender, no household income bracket. What it uses instead is layered:
- What you do on the service: viewing history, thumbs up/down ratings, how much of something you actually watch
- What people with similar taste patterns do: not "people like you" by age or postcode, but by behaviour
- Context: time of day, device, language
One of the most sophisticated recommendation systems in commercial use runs on combined, observed behaviour and deliberately excludes demographic identity, because identity doesn't predict what happens next as well as behaviour does.
Sources: Netflix, "How Netflix's Recommendations System Works," Netflix Help Centre · Netflix Technology Blog, "Recommending for Long-Term Member Satisfaction at Netflix"