The Intention-Behaviour Gap And How To Measure Around It
By aytm
- article
- Advertising Testing
- Agile Quantitative Research
- Brand Tracking
- Conversational AI
- DIY Surveys
- Insight Communities
- Pricing Research
- Product Development Research
- Survey Research
- Survey Software
- Behavioural Insights
- Behavioural Science
Summarise with AI
This is the fifth article in aytm’s Innovation Intelligence series, which explores each stage of the innovation research funnel. Each piece draws on the Product Innovation curriculum in the aytm Lighthouse Academy.
Explore the full suite of innovation courses here.
Imagine your usage estimates holding up after launch. Your replenishment cycles landing where the model said they would, your volume forecasts surviving contact with the market and your segments behaving the way the data promised. That’s what becomes possible when behaviour frequency reflects what people actually do, and there’s a clean way to get there.
Behaviour frequency is one of the most consequential things research measures. How often someone buys, uses, replenishes or engages drives product usage estimates, purchase volume forecasts, segmentation and the volumetric inputs behind pricing and market-opportunity models. If consumers report brushing their teeth three times a day, you can estimate roughly 90 servings of toothpaste a month and a replenishment cycle that refills an 80-use tube monthly. If they actually brush once a day, that tube lasts two or three times as long, and the demand model built on the higher number is off from the start. Get frequency right, and everything downstream inherits the accuracy.
There’s a reason frequency is hard to capture cleanly, and it has a name. People intend to do things more often than they end up doing them, and when a survey asks how often they do something, that intention quietly colours the answer. It’s called the intention-behaviour gap, and once you see how it works, you can design straight through it.
Why Frequency Reporting Runs High
Understanding where the inflation comes from is the first lever you have over it, and the pattern is well documented. Citizens report voting more consistently than public voting records show. Churchgoers report attending more often than attendance records support. University students report exercising more often than sports-facility records confirm. The effect is sharpest for behaviours with strong normative standards or strong personal aspirations, the behaviours people feel they should do.
The conventional explanation is impression management. Two biases sit under that umbrella. Acquiescence bias pushes respondents to be agreeable and give the answer they think the interviewer wants. Social desirability bias pushes them toward the socially approved answer. Both are rooted in the very human work of managing how others see us.
Impression management explains part of the effect, and there’s clearly more to it. Over-reporting shows up even in the most private data-collection modes, where there’s no one to impress. That persistence points to a second mechanism operating alongside, and often beneath, the social one. People are telling us what they mean to do.
The Gap Is A Feature Of Human Behaviour
Treat the gap as a property of people rather than a flaw in your survey, and you can design for it directly. Foundational behavioural theory, the work of Ajzen and Fishbein on the theory of planned behaviour, holds that intention is the most direct, proximal cause of action. We do what we intend to do, and often we fall short. The discrepancy between intention and follow-through is so well established in social and health psychology that it has been termed the intention-behaviour gap. It opens for ordinary reasons, such as when we run short on time, have to prioritise something else or simply forget. Entire industries, from gym memberships to meal kits to habit apps, exist to address it, and it persists anyway.
The sociologist Peter J. Burke put the measurement consequence plainly: “The problem with most measurement situations is that without the normal situational constraints, it becomes very easy for a respondent to give us that idealised identity picture which may only seldom be realised in normal interactional situations”. A survey is exactly that constraint-free situation. When you ask “On average, how often do you exercise?” you’re asking a question that braids together a normative standard, a personal aspiration and an actual behaviour. You get back a single number that mixes all three.
The standard survey braids the three together because it gives respondents only one place to put the answer. With a single question, intention has no channel of its own, so it leaks into the behavioural answer. The fix is to give the intention a place to go.
A Cleaner Way To Ask
aytm built and tested that fix directly. The experiment ran with 3,200 U.S. adults, balanced to the population on age and gender, reporting frequency across 17 everyday behaviours. The behaviours were chosen deliberately to span normative and aspirational territory and a wide range of cadences, from things done multiple times a day (brushing teeth, washing your face) to things done a few times a year (seeing a movie in a theatre). Respondents were randomly assigned to one of four conditions, 800 each:
- A. Control, behaviour only. The standard approach. “How often, on average, do you do [X]?”
- B. Intention plus gap description plus behaviour. Ask intention first, then describe the gap (“It can be hard to do things as often as we intend, because we’re short on time, must prioritise other things or because we forget”.), then ask actual behaviour.
- C. Intention plus behaviour, no gap description. Ask “How often do you intend to [X]?” then “How often, on average, do you actually [X]?” Nothing else.
- D. Gap description plus behaviour, no intention. Describe the gap, then ask about actual behaviour. No separate intention question.
Two questions were on the table. Does separating intention from behaviour produce more accurate frequency reporting? And if so, what’s doing the work, expressing the intention or being told about the gap?
The first step confirmed the gap was real in this data. Comparing intended versus actual frequency, respondents intended to do nearly everything more often than they did. Across the 17 behaviours, 15 showed a significant intention-behaviour gap at the 95% confidence level and the remaining two (doing laundry and going to the movies) at 90%. The headline examples are stark: 46% of respondents said they intend to exercise most days of the week, while 27% said they actually do. 68% intend to brush their teeth at least twice a day, and 53% actually do.
What The Experiment Found
Here’s the payoff. Compared against the control, the condition that produced consistently more accurate (lower, less inflated) behaviour frequency across every domain was Cell C: ask intention first, then behaviour, with no explanation of the gap at all. Cell B, which added the gap description, produced a similar but weaker effect and left two behaviours unmoved. Cell D, which described the gap but never asked respondents to express their intention, came out statistically indistinguishable from the standard control on most behaviours.
One honest caveat keeps the finding precise. The study inferred accuracy from direction. It treated lower, less-inflated reporting as more accurate, grounded in behaviours where over-reporting is well documented in the literature (exercise among them), and it didn’t validate the responses against true behavioural records. Within that frame, the read is clean and immediately usable. Letting respondents state how often they intend to do something before asking how often they actually do it yields more accurate behaviour reporting. Describing the intention-behaviour gap turns out to be neither necessary nor sufficient. You can skip the lecture about why intentions and actions diverge. You give the intention its own question, so it stops contaminating the behaviour question, and expressing the aspiration separately frees the next answer to be honest.
That’s a free, two-question redesign any frequency study can adopt tomorrow. It costs one extra question, and it sharpens every usage estimate, volume forecast and segment cut built on top of the data. It opens a second door too. Once intention and behaviour are measured separately, the size of the gap becomes a variable in its own right. Intention may be the better predictor of initial trial and purchase, while actual frequency better predicts usage and replenishment. And “large-gap” consumers, the people who intend far more than they do, become a segment worth understanding on their own terms. They’re often exactly the people a new product is built to help close the gap.
Where Observing Beats Asking
Better questioning narrows the gap, and for the highest-stakes decisions, you can close the rest of the distance by watching behaviour directly. Some of the gap lies in the difference between any quiet survey moment and the messy moment of real choice. A respondent answering carefully on a screen is still not standing in an aisle, scrolling a feed or weighing your product against the brand they already trust. For the decisions where that distance is most expensive, the most reliable instrument captures behaviour instead of asking for it.
This is the logic behind aytm’s behavioural-validation methods, which read captured behaviour rather than stated intent. The same shift the frequency experiment makes inside a survey, the Reality Check guide makes across the whole research design. Its 4-S lens (Seen, Shoppable, Seductive, Selected) tracks a concept through the four hurdles between shelf and basket. Its premise is the gap this article is about: a concept can test at 78% purchase intent in a quiet room and still sit 30% below category average twelve weeks after launch, because purchase intent captures what people say in isolation, while real choice happens under competitive pressure.
aytm’s behavioural tools are built to capture that “do”. A Shelf Test places a product in a high-fidelity virtual store (Target, Walmart, Costco, and a growing library of retailer environments). It records what people actually do: heatmaps of where attention lands, time to first click, attention span per product and whether shoppers add to cart or walk away empty-handed. Ecommerce Research does the same on the digital shelf, capturing scroll-and-click behaviour, product-page engagement and, critically, source of volume, whether a new SKU takes share from a competitor or shifts volume from your own line. The Ad Testing Simulator captures creative performance in-feed against the ads a buyer is already seeing.
The throughline holds. Where a question is the right instrument, ask it in a way that separates what people intend from what they do. Where the stakes are high enough that even a well-designed question needs backup, observe the behaviour directly in an environment that resembles the real one. The intention-behaviour gap is research’s oldest problem because it never fully disappears, and it’s manageable. Measure intention and behaviour as two distinct things, and your data starts describing the life people actually live rather than the one they mean to.
Explore the Product Innovation curriculum and aytm’s behavioural-validation methods at the aytm Lighthouse Academy.
