Say-Do Gap

What lies behind it

Someone who says in a survey that they would take out a tariff or buy a product is giving an honest account of their attitude at that moment. But they are not making a choice. What is missing is the price compared with alternatives, the effort of signing up, the distractions of everyday life and the knowledge that the answer would have consequences. At the moment of the offer, exactly these forces are at work, and they shift the result.

Several mechanisms reinforce each other here. Intentions are often not acted upon because the moment, the opportunity or the priority is missing (intention-behaviour gap). Hypothetical statements of willingness to pay come out higher than real payments because no money changes hands (hypothetical bias). Respondents answer in the way that seems socially desirable or that matches how they would like to see themselves. And the survey itself changes how people later think about a topic (mere measurement effect).

The concept is old, even though today it is used mainly in marketing. An early and frequently cited piece of evidence comes from LaPiere in 1934: a travelling couple was turned away only once in 251 hotels and restaurants. When the same establishments were later asked in writing whether they would serve such guests, more than 90% of those that replied said no. Statement and action ran in opposite directions.

For companies, the Say-Do Gap is neither a fault of respondents nor a flaw of one particular method. It is a property of every measurement that asks about future behaviour instead of observing it. This does not mean surveys are useless. It means that what they say about demand is a different kind of evidence from observed behaviour.

How large the effect is

Research quantifies the gap from different directions. What matters is which quantity is being measured in each case: the relationship between intention and behaviour, the overestimation of willingness to pay, or how accurately purchase intent anticipates later sales.

Intention and behaviour: Sheeran and Webb (2016) summarise that intentions are translated into action in roughly half of all cases. Conner and Norman (2022) report that intentions explain 18 to 23% of the variance in behaviour. Webb and Sheeran (2006) also show causally, in a meta-analysis of 47 experiments, that a medium-to-large change in intention (d = 0.66) leads only to a small-to-medium change in behaviour (d = 0.36).

Willingness to pay: Schmidt and Bijmolt (2020) analyse 77 studies with 115 effect sizes. Hypothetically stated willingness to pay is on average 21% higher than real willingness to pay. Older meta-analyses from environmental economics arrive at higher values depending on study selection: Murphy et al. (2005) find a median ratio of 1.35 with a strongly right-skewed distribution, List and Gallet (2001) an overstatement of roughly a factor of 3.

Purchase intent and sales: Morwitz, Steckel and Gupta (2007) show in a meta-analysis that stated purchase intent is more weakly related to later sales for new products than for existing ones. Chandon, Morwitz and Reinartz (2005) find that the relationship between intention and purchase is 58% stronger among surveyed customers than among customers who were not surveyed. The survey itself therefore makes intentions appear more reliable than they would be in the market without a survey.

The range of these values is itself a finding: there is no single correction factor that reliably translates survey results into demand.

Why the gap differs from decision to decision

The studies show not only that the gap exists, but also when it grows. Schmidt and Bijmolt find stronger overestimation for higher-value products, for specialty products and in designs where the same people rate several variants. Morwitz and colleagues find the weakest link between intention and sales for new products.

In practice this means: the gap is largest where decisions are most expensive. A new offer, a premium price, a tariff that needs explaining or a promise that people like to hear but rarely pay for are typical cases. For routine purchases of existing products the gap tends to be smaller, but not zero.

Then there is the direction. Demand is usually overestimated. For variants that differ only in details, however, a survey can also change the ranking: the variant that performs best in the questionnaire is not necessarily the one most people choose in the offer environment. Because size and direction are unknown, the gap cannot be cleanly corrected without observation.

Where the gap typically occurs

Price and price increases: respondents name amounts that they often do not choose at the moment of the offer. This becomes particularly visible with premium tiers and with price increases for existing products, where the question is how much purchase intent or sign-up intent a higher price costs.

Tariffs and contracts: in insurance, energy, telecommunications and banking, signing up involves effort. Forms, contract terms and switching effort are missing from the survey but take effect at the moment of the offer. Stated willingness to sign up is therefore often higher than what a sign-up journey shows.

Sustainability and values: statements about environmentally friendly or socially responsible products are prone to social desirability. Agreeing is easy, the price premium is not. Here the Say-Do Gap overlaps with the attitude-behaviour gap.

New concepts and features: novelty sparks interest and good ratings. Whether people choose a new offer instead of their usual solution only becomes clear once price and alternative are visible.

Claims and brand: a promise can be convincing in a survey and still lead nobody to the offer. Conversely, some plain wording has a stronger effect on behaviour than its rating would suggest.

What this means for decisions

Business cases for new products, tariffs or pricing models often work with a stated willingness to buy or sign up. Every step that rests on a statement in the conditional carries the gap forward: sales, revenue, contribution margin and payback. Small deviations at the start thus become large deviations at the end of the calculation.

This does not only affect innovation. With a price increase, the deciding question is how much sign-up intent the higher price costs. With a claim, what matters is whether the statement leads people to the offer, not whether they agree with it. With brand and positioning, what matters is which variant is chosen in competition. In all these cases a survey measures agreement, while the decision depends on behaviour.

The consequence is not to abandon surveys, but a clear division of roles: surveys and qualitative research explain motives and drivers and help develop variants. The final question, which variant people choose, is answered by a test that measures behaviour before investing.

How to deal with it

Within surveys there are proven ways to narrow the gap: realistic price information, choice tasks instead of scales, designs in which each person sees only one variant, reminders of the bias before the question (cheap talk) and incentive-compatible procedures in which a choice actually has consequences. These approaches improve measurement. But they do not reliably eliminate the gap, and none of them is free of bias.

The second way is to observe behaviour directly. This happens in the market, for example with sales data or A/B tests on existing offers, or before the market with behavioural tests. A Painted Door Test shows a realistic offer with a price in a familiar online environment and measures how many people choose it. Nothing is sold: anyone who decides is then told transparently that it was a test.

Horizon puts this second way into practice. Real people reach the offer pages via Google and Meta ads, without a panel and without incentives. Up to six variants are compared under equal conditions, and it takes around four weeks from the question to the data analysis. The result is measured purchase intent or sign-up intent for each variant, which complements the survey rather than replacing it.

Example

An insurer is planning a new supplementary dental tariff in three variants: low-priced with basic cover, mid-priced with preventive care, higher-priced with dentures and crowns. In a survey, most participants say they would probably take out the middle variant, and the expensive variant receives the best ratings for its benefits.

In the behavioural test, each person sees only one of the three offer pages, each with a price and a sign-up journey. What is measured is the last binding click before the reveal. Result: the low-priced variant achieves the highest measured sign-up intent, the middle one is just behind, and the expensive one is clearly lower. The distance between low and middle is small enough for the insurer to pursue the middle variant with its higher premium, now with a robust assumption for the business case.

The survey was not wrong. It showed which benefits people like and why. Only behaviour showed which variant is chosen and how much sign-up intent the higher price costs.

Related terms

Intention-behaviour gap: the gap between intention and action from psychology, the core of the Say-Do Gap in purchase and sign-up decisions.

Hypothetical bias: the systematic overestimation of willingness to pay when no money changes hands.

Attitude-behaviour gap: the gap between attitude and purchase, particularly visible with sustainability and premium products.

Mere measurement effect: asking alone changes later behaviour and makes intentions appear more reliable.

How behavioural data and survey data complement each other is described in the entry Behavioural data vs. survey data.

Limitations

The Say-Do Gap is not an argument against surveys. It describes what statements about future behaviour are less suited for: as the sole basis for demand and price in new or expensive decisions. For motives, perception, understanding and satisfaction, surveys are often the better tool.

A behavioural test has limitations too. It measures purchase intent or sign-up intent at the moment of the offer, not repeat purchase, usage or cancellation. It shows which variant is chosen, but does not on its own explain why. And it describes the people reached through the chosen channels. For decisions dominated by advice, long contract negotiations or offline channels, careful transfer is needed.

Finally, measured purchase intent is not sales. It is a robust comparison between variants and a better assumption for the business case than a stated willingness, but responsibility for the decision remains with the decision makers in the company.

Taking the Say-Do Gap seriously therefore does not require a new method for everything, but a clear rule about which decisions need which evidence.

All terms in this topic

• Intention-Behaviour Gap
• Hypothetical Bias
• Attitude-Behaviour Gap
• Mere Measurement Effect