Customer research uncovers what people say they will do. Real product validation measures what they actually do. The two are almost never the same -and the gap is where many launches fail.
Ask a hundred people whether they would buy your product, and most of them will say yes. Not because they are lying. Because they are being polite, and because predicting their own future behaviour is one of the things human beings are demonstrably bad at.
This is, incidentally, one of the more useful findings to come out of forty years of behavioural economics: there is a vast and reliable gap between what people say they will do and what they actually do. Surveys reveal stated preferences. Real life reveals revealed preferences. The two are seldom in agreement, and where they differ, it is the revealed version that pays the bills.
Most market research, however carefully designed, sits firmly on the wrong side of that gap.
The most enthusiastic launch event ever held for a vehicle nobody wanted
On 10 January 1985, Sir Clive Sinclair – then one of Britain’s most celebrated inventors, fresh from the runaway success of the ZX Spectrum home computer – launched the Sinclair C5 at Alexandra Palace. The C5 was a small, single-seater electric tricycle, manufactured in Wales by Hoover (the washing machine people), built on a Lotus chassis, and pitched as nothing less than the future of personal transport.
The launch event was, by any reasonable measure, a triumph. Over the course of the demonstration days, more than 20,000 members of the public turned up to try the C5 on the indoor test track. The vast majority reacted positively. They smiled. They told the staff they thought it was brilliant. Sinclair himself was so encouraged that he declared, publicly, that the entire first production run of 2,700 units would be sold by the following Monday.
By the end of that weekend, around 200 had been sold.
By March 1985, when the C5 arrived in retail stores, sales were already collapsing.
Comet ordered 1,600 units and most were still sitting unsold nine months later. Production was cut by 90 per cent within three months of launch. By August, the line had stopped entirely. Of the 14,000 C5s ever manufactured, only around 5,000 were ever sold. Sinclair Vehicles went into receivership in October
1985, nine months after the Alexandra Palace launch.
The 20,000 people who said they loved it were not lying. They genuinely enjoyed the experience of trying a futuristic-looking vehicle on a smooth indoor track in a warm hall with no traffic. What they could not do, what no member of the public can ever reliably do, was predict how they would feel about it when faced with the actual purchase decision, in cold British weather, on real roads, next to actual lorries, in front of their neighbours. The stated preference was enthusiastic. The revealed preference was almost completely absent.
This is the central problem with conventional market research at any scale, whether you are a global retailer, an established manufacturer, or a founder with a working prototype. Research measures attitudes. Markets measure actions. The two are not the same data set, and they rarely produce the same answer.
Why people are so unreliable about themselves
There are three reasons the gap between stated and revealed preference is so wide, and they apply with equal force to consumer products, B2B propositions, and the sort of public demonstration that filled Alexandra Palace with C5 enthusiasts.
First, people want to be helpful. When you ask someone whether they would buy your product, you are also, implicitly, asking them how they feel about you. Most respondents reach for the answer that is socially comfortable, and that almost always tilts toward yes.
Second, the cost of saying yes in a survey is zero. The cost of saying yes with a card in hand is not. Hypothetical purchases are always cheaper than real ones, and people behave very differently when the price is theoretical.
Third – and this is the one even sophisticated researchers underestimate – people genuinely do not know what they will do. We systematically overestimate how much we will exercise, how often we will use the gym membership, how much we will spend on the subscription, how disciplined we will be with the meal-prep app. The error is not malicious. It is a basic feature of human self-prediction. We answer about the person we would like to be, rather than the person we will actually be on a wet Tuesday in November.
We see this pattern repeatedly in the validation work we run. A UK consumer brand we worked with, preparing to launch a premium product extension, had spent months in qualitative research with their existing customer base. The feedback was overwhelmingly positive: people liked the proposition, liked the brand stretch, said they would buy it. When we ran a paid-traffic test against a
live landing page with real pricing, the picture changed almost immediately. The clicks were healthy. The conversions were not. The proposition that had performed beautifully in conversation performed poorly under the simple test of asking someone to part with money. That gap, the one between “I think I would buy this” and “I will buy this”, is not a failure of the research. It is what the research can never reach.
None of this makes survey research worthless. It makes it directional. Surveys are useful for understanding how people think about a category, what language resonates, what the perceived problem is. They are dangerously misleading when used as evidence that a proposition will sell.
The other path: measuring behaviour from day one
Thirty years after Sinclair filled Alexandra Palace with smiling test-drivers who turned out not to be buyers, a small team in London took a deliberately different approach to validating a different kind of proposition. Monzo, then trading as Mondo, was a challenger bank with no banking licence and, in its first months, no product to speak of.
Rather than commissioning a survey on whether the British public might want a digital-first bank, the team did something cheaper and far more revealing. They built a prepaid Mastercard, paired it with a basic mobile app, and put a waitlist on the website. People who wanted in had to sign up. People who wanted to move up the queue had to invite others. People who got off the waitlist had to load real money onto a real card and use it for real purchases.
By the end of 2015, Monzo had around 3,000 alpha cardholders actively using the product and a waitlist of approximately 20,000 people behind them. None of that data came from asking anyone what they wanted. All of it came from watching what they did when given the option. The waitlist itself was not just a marketing mechanic; it was the validation mechanism. Each sign-up was a small
piece of evidence that the proposition had genuine pull.
This is not just a fintech curiosity, and it is not only relevant to first-time founders. The principle holds for any commercial leader weighing a launch decision. Whether you are introducing a new SKU into an existing distributor network, opening up a new vertical, or testing an entirely new brand, the
cheapest and most reliable evidence you can gather is evidence of behaviour. Stated preference is what people will tell you. Revealed preference is what they will spend.
What behavioural validation looks like in practice
The Monzo example is a consumer story, but the same logic underpins effective validation across every commercial context. The question is simply: what is the smallest, cheapest action you can ask a real prospect to take that costs them something? Time, money, social capital, anything more than
a tick in a survey box.
In practical terms, that looks like:
- A landing page with a real price and a real “buy now” or “pre-order” button, even before the product is fully ready, to measure conversion rates from paid traffic.
• A waitlist with a deposit, however small, attached. £10 separates the curious from the committed more effectively than any focus group.
• A signed letter of intent from a target B2B buyer, agreeing in principle to pilot a product on defined terms. Not a contract. Just a document with a name and a date on it.
• A pilot at cost or below cost with three credible target customers, in exchange for the right to use the results as evidence.
• A meeting accepted in someone’s diary in response to a cold outreach that names the price. If buyers will meet you knowing what it costs, that is data. If they will only meet you to discuss something free, that is also data.
Each of these costs the prospect something. That cost is the whole point. The minute someone is asked to part with time, money or commitment, their answer changes – and that change is the information you actually need.
On the B2B side, the most useful evidence is often the smallest. We worked with a manufacturer preparing to launch a new product range into an established trade channel. The internal view, supported by enthusiastic conversations with key distributors at industry events, was that the proposition was already validated. Distributors had nodded along, said the right things, and indicated interest. Rather than build a full launch plan on that basis, we asked the same distributors a single, more pointed question: would they sign a non-binding letter of intent to stock the range at agreed terms, subject to the product meeting specifications? The conversations that followed were significantly more revealing than the ones at the trade stand. Two distributors signed within a fortnight. Several others raised concerns – about margin, about pack format, about category positioning – that had not surfaced in any of the previous discussions. The proposition was, in the end, validated, but the launch that followed was a materially different one. The cost of asking that question was almost zero. The cost of finding out the same information at full launch would have been considerable.
What to ask before you ask anyone anything
Before commissioning another piece of research or running another round of customer interviews, three questions are worth holding the work against:
- Will this research tell me what people say, or what they will do?
- If it tells me what they say, am I treating that as directional – or am I about to make a commitment based on it?
- What is the cheapest experiment I could run that would tell me what they will do instead?
Most commercial leaders find that the third question, once asked properly, redirects a substantial portion of the research budget toward something that produces a much more useful answer.
BrandRefinery does not run focus groups, and we do not write market research reports. We design tests that turn propositions into evidence by measuring what your future customers actually do.
Sometimes that is a landing page test with paid traffic and a real price. Sometimes it is a pre-order campaign on a crowdfunding platform that doubles as a proof of demand. Sometimes it is a structured B2B pilot programme with three target accounts. The mechanics depend on the proposition, the audience and the budget. The principle does not. If you are about to commit serious capital to a launch or a scale-up, you should know how your real customers behave when given the choice, not just how they answer when asked.
It is, on balance, a slightly less comfortable kind of work than commissioning a research report. It gives you data that occasionally tells you things you would rather not hear. But the cost of finding out at this stage is roughly one or two per cent of the cost of finding out at full launch.
If you would like a thirty-minute conversation about how to test what your customers will actually do, rather than what they say they will do, book a call. No survey required.
Frequently asked questions
What is the difference between stated and revealed preference?
Stated preference is what someone says they will do when asked – in a survey, focus group, or interview.
Revealed preference is what they actually do when faced with the real decision, with real money, time or commitment at stake. The two rarely match, because people are reliably poor at predicting their own future behaviour. Effective product validation prioritises revealed preference
over stated preference at every stage.
Why does market research often fail to predict commercial outcomes?
Most market research measures attitudes and opinions in a low-stakes setting. It asks people what they think, not what they will do. The cost of agreeing in a focus group, or smiling on a test track, is zero; the cost of actually buying a product is not. Brands have lost very large sums of money – Sinclair Vehicles being one well-documented example – acting on enthusiastic feedback that did not translate into commercial behaviour. Research is useful for understanding language, perception and category structure. It is dangerous when used as evidence that a proposition will sell.
How do you validate a new product before launching it?
The most reliable validation methods measure behaviour rather than opinion. That includes paid-traffic landing page tests with real pricing, waitlists with deposits, signed letters of intent from B2B buyers, structured pilots with target customers, and pre-order campaigns through platforms like Crowdcube or Indiegogo. Each of these requires the prospect to do something that costs them something, and that cost is exactly what makes the data meaningful.
Is customer behaviour validation only relevant to early-stage startups?
No. The principle applies to any commercial decision where the question is whether a proposition will sell. Established businesses launching new ranges, entering new categories, or rebranding existing products face the same stated-versus-revealed problem. In some ways the risk is higher for
established businesses, because the cost of getting it wrong, in inventory, supply chain commitments, and internal credibility, is greater than for a founder with a prototype.
What does BrandRefinery actually do differently?
BrandRefinery designs and runs real-world tests that measure customer behaviour, not opinion. We do not conduct focus groups or write conventional market research reports. Depending on the proposition, the work might involve a paid-traffic landing page test, a pre-order campaign, a structured B2B pilot programme, or a combination. The output is evidence of what people will do, which is the only kind of evidence on which a launch decision is safely made.