Multiple evidence sources show that what people say, and who they are, poorly predicts what they will do
Targeting has always leaned on two convenient shortcuts. Ask people what they intend to do, or work out who they are, and assume the rest follows. Both feel intuitive. Both are the industry's habit rather than its evidence. Looked at closely, neither holds up, and a third, quieter factor turns out to matter more than either.
What people say doesn't predict what they do
Declared intent - a survey answer, a stated preference, a self-reported interest - is the most direct signal a marketer can ask for. It's also one of the least reliable. In The Choice Factory (2018), the behavioural scientist Richard Shotton lays out case after case where what people say about their own decisions and what they actually do diverge, often sharply.
The pattern isn't that people are lying. It's that most of what shapes a decision happens below conscious awareness, so when you ask someone why they chose something, they answer with a plausible story rather than the real cause. Shotton has made the same point about market research more broadly: traditional surveys and focus groups capture what respondents believe influenced them, not what actually did, and taken at face value they can point a brand in exactly the wrong direction.
That matters for targeting because declared intent inherits the unreliability of self-report. A signal built on what someone claims they'll do is only ever as good as their own insight into their own behaviour, and the evidence says that insight is thin.
Demographics get identity targeting wrong more often than the industry admits
63%
That's the average on-target rate Nielsen found for ads aimed at age and gender in its 2022 report, Bridging the Gap: Better Audience Targeting - and age and gender are the simplest demographic categories available, with the best data coverage behind them. Even here, more than one in three impressions bought against the target miss it.
Most targeting stacks don't stop at age and gender. They layer on income, life stage, parental status and more, on the assumption that stacking identity attributes sharpens the picture. It's a reasonable-sounding assumption, but it runs against what the base rate above already shows: identity, even measured simply, is a weak foundation to build on - and every additional attribute inherits that same weakness rather than correcting for it.
Context beats identity
The clearest demonstration of this comes from outside advertising altogether. In 1973, psychologists John Darley and Daniel Batson ran what's now known as the Good Samaritan study. Seminary students were asked to prepare a short talk - for half of them, on the parable of the Good Samaritan - and then walk to another building to deliver it. Some were told they had plenty of time. Others were told they were already running late.
On the way, each student passed a person slumped in a doorway, apparently in distress. The single biggest factor in whether they stopped to help wasn't their religious motivation, their stated values, or even whether they'd just been thinking about the parable of the Good Samaritan itself. It was whether they thought they had time.
63% vs 10%
Of the students who weren't rushed, 63% stopped to help. Of those told they were late, only 10% did - a near-total reversal, driven entirely by a situational, momentary variable rather than by who these people were or what they professed to believe.
Three failures, one pattern
Put the three together and a pattern emerges. What someone says they'll do is a weak predictor of what they do. Who someone is - even measured on the simplest, best-supported categories - predicts weakly too. What reliably moves behaviour is the situation someone is in at the moment it counts.
None of this is an argument for giving up on targeting. It's an argument for being honest about which signals the evidence actually supports, and which ones the industry has simply gotten used to relying on.
Sources
Shotton, R. (2018). The Choice Factory: 25 Behavioural Biases That Influence What We Buy. Harriman House.
"The Illusion of Choice: Behavioural Science in Marketing with Richard Shotton." AdNews, 2023.
Nielsen. "Bridging the Gap: Better Audience Targeting with Context and Behavior Data." 2022.
Darley, J. M., & Batson, C. D. (1973). "From Jerusalem to Jericho: A study of situational and dispositional variables in helping behavior." Journal of Personality and Social Psychology, 27(1), 100-108.