People Rarely Tell You Why They Buy. They Show You.

Ask a hundred customers why they chose your product. You'll get a hundred clean, logical explanations. Price. Features. Timing. A recommendation from a colleague. Read those answers carefully, build a strategy around them, and you might still watch your best cohort quietly churn six months later while a product with half your feature list doubles its user base. Something is off. The problem is not that customers are lying. The problem is that most of the purchase decision happened somewhere they don't have access to, and the explanation they give you is constructed after the fact. This post is about what that means, why conventional research keeps missing it, and what you can actually do with the signal that's always been sitting right in front of you.

Key Takeaways

• According to Harvard Business School professor Gerald Zaltman, roughly 95% of purchase decisions are made in the subconscious mind, meaning customers often cannot accurately report their own motivations.
• The "say/do gap" is not a data quality problem. It is a fundamental feature of how human decision-making works.
• Traditional research methods like focus groups create artificial conditions that produce artificial responses, as demonstrated by the New Coke failure.
• Behavioral data (what customers repeat, when they churn, which paths they take before buying) is a more reliable signal than stated preference.
• Closing the gap between what customers say and what they do requires watching patterns over time, not just asking better questions.

The Classic Idea: Customers Know What They Want, So Just Ask Them

The standard move in consumer research has always been to go straight to the source. Run a survey. Convene a focus group. Build a voice-of-customer program. The logic is intuitive: if you want to understand why someone bought something, ask them. If you want to predict what they'll buy next, ask them about that too.

This assumption runs so deep in market research that it became institutional. Entire methodologies were built on it. Hundreds of millions of dollars flowed into focus group facilities, panel providers, and survey platforms, all on the premise that consumers, given a comfortable room and a leading question, could surface their own motivations on demand.

The economist Paul Samuelson actually pushed back on this idea as far back as 1938, when he proposed what became known as revealed preference theory: the idea that behavior is a more reliable indicator of consumer preference than stated intent. Samuelson's point was not that people lie. His point was that choices, when made with real money under real constraints, reveal preferences that self-reports simply cannot.

That insight sat in economics literature for decades, occasionally visited and then quietly set aside, while market research kept asking the same questions in the same focus group rooms.

What Everyone Gets Wrong About the Say/Do Gap

Here is the part that makes this genuinely interesting rather than just frustrating.

Customers are not withholding the real reason they bought. They are not running a game. When someone tells you they switched shampoo brands because of the price, they are not being strategic. They genuinely believe that is why. The construction happens fast and invisibly, a post-hoc narrative that feels like memory but functions more like explanation. The purchase decision has already been made by the time the conscious mind gets involved, and what the conscious mind contributes is largely the story told afterward.

Gerald Zaltman, a professor at Harvard Business School, put a number on this. His research found that approximately 95% of purchase decision-making takes place in the subconscious mind. His fieldwork revealed something even sharper: many consumers report comparing multiple competing brands and price points at the point of purchase, but direct observation of those same consumers shows they often do not look at alternatives at all. The stated behavior and the observed behavior are two different things. Neither is fake. They simply belong to different cognitive processes.

The New Coke debacle sits here as the clearest case study that money can buy. Coca-Cola ran extensive taste tests before its now-infamous formula change. By most accounts, the new formula performed well in blind tests. Consumers said they liked it. Some said they preferred it. Around $300 million worth of product went on to gather dust on supermarket shelves once the product launched, while customers flooded call centers expressing something closer to grief than product dissatisfaction.

What the research missed was not the flavor. It was the accumulated emotional weight of the original formula, the nostalgia, the identity, the cultural ownership that consumers felt but could not name when handed a paper cup in a testing room. The Irrational Agency, writing on the incident, described it well: "Consumers rarely mean what they say, say what they mean, and know what they want. At least not when you are conducting simple focus groups and surveys."

That is not a cynical conclusion. It is a structural one.

The gap between those two columns is where most marketing strategy goes wrong. And it is worth saying plainly: surveys are not useless. They reveal things. But they reveal a particular kind of thing, a reconstructed, rationalized, socially acceptable account of behavior that may or may not correspond to what actually drove the decision.

What Shifts When You Watch Instead of Ask

Behavioral data does not ask anything. It just records.

The digital environment created something that ethnographers working in the field have known for decades: the most useful data comes from observation in context, not responses in isolation. Ethnographic research methodology holds that studying behavior in natural environments consistently surfaces motivations that interviews and surveys miss. A shopper observed in a grocery aisle reveals more about their decision-making than the same shopper answering a questionnaire about that aisle later.

The behavioral signals that accumulate in digital products and commerce environments operate on the same logic. A customer who returns to a pricing page four times before purchasing is telling you something. A user who opens your mobile app at 10pm three nights in a row is telling you something different than what they would report in a monthly satisfaction survey. The question is not whether the signal is there. It almost always is. The question is whether anyone is reading it.

What changed is volume and accessibility. Behavioral data at scale is now readable in ways it was not for most of the history of consumer research. That does not make interpretation automatic. Pattern recognition still requires judgment, and not every behavioral signal means what it appears to mean. A spike in page visits before churn could indicate high interest or the frantic last effort of a frustrated user trying to solve a problem your product failed to solve. Context matters. But the raw material is there.

What This Means for How You Actually Operate

Operationally, the implication is not to abandon surveys or qualitative research. It is to stop letting stated preference carry more weight than revealed behavior, especially when the two conflict.

Pay attention to what gets repeated, not just what gets praised. The most reliable signal of satisfaction is a return purchase made without a discount, a prompt, or a campaign. If customers who say they love your product still require a promotional push to come back, that gap deserves investigation. Praise is easy to collect and difficult to act on. Repurchase behavior under ordinary conditions tells you whether the value was real.

Churn timing is a behavioral confession. When a customer leaves is not random. Customers who churn in the first week after onboarding are usually telling you the product failed to deliver on the expectation set during acquisition. Customers who churn at month six are often telling you something hit a ceiling: a usage ceiling, a value ceiling, or a relationship ceiling. Exit surveys will give you reasons. Churn timing gives you the chapter.

The path before the purchase matters more than the purchase itself. How many times did the customer visit before converting? What did they read, watch, or click in that window? Which objection pages did they linger on? The pre-purchase path is a map of what they were actually weighing, and it is often more revealing than anything they will say in a post-purchase survey.

Feature usage patterns reveal what actually creates value. Product teams frequently build roadmaps based on feature requests gathered from customers or sales teams. But the features customers ask for and the features that correlate with retention are often different sets. Looking at which features get used consistently, by the customers who stay and spend more, will tell you where the real value lives. That is worth knowing before the next roadmap conversation.

None of this requires exotic infrastructure. It requires treating behavioral data as a first-class input rather than a supporting exhibit.

The Signal Was Always There

Here is the honest version of this: reading behavioral signals is not some new invention. Ethnographers have been doing it for over a century. Economists built preference theory around it decades ago. Smart retailers have been reading foot traffic, return rates, and shelf placement data for as long as shelves have existed.

What makes the current moment interesting is not that the signal is new. It is that the gap between the signal and how most organizations make decisions has gotten harder to justify. The data exists. The tools to read it exist. The decision to weight a focus group transcript over a cohort behavior report is, at this point, a choice rather than a constraint.

Customers do not usually know why they buy. That is not their job. Understanding the gap between what they tell you and what they actually do is yours.

Frequently Asked Questions

What is the say/do gap in consumer behavior?

The say/do gap refers to the consistent difference between what consumers report about their purchasing motivations and what their observable behavior actually shows. Customers who claim to compare multiple brands before buying often do not. Customers who report price sensitivity still repurchase at full price. The gap exists not because consumers are dishonest, but because most purchase decisions are processed subconsciously, and the explanations offered afterward are constructed narratives rather than accurate accounts.

Why do focus groups and surveys often fail to predict purchase behavior?

Focus groups and surveys create artificial environments that strip away the contextual distractions, emotional associations, and habitual patterns that normally shape real buying decisions. When respondents are placed in a structured setting and asked to evaluate a product in isolation, they respond to that setting, not to the conditions they would face in an actual store or purchasing moment. The New Coke product launch is one of the most cited examples: research pointed to consumer approval; actual market behavior told a completely different story.

How did Gerald Zaltman's research change how marketers understand consumer decisions?

Harvard Business School professor Gerald Zaltman's research found that approximately 95% of purchase decision-making takes place in the subconscious mind. His fieldwork also showed that consumers frequently report behaviors (such as comparing competing brands) that direct observation shows they do not actually perform. Zaltman's work pushed marketers toward methods that probe non-conscious motivations rather than relying solely on self-reported attitudes.

What behavioral signals should businesses track to understand buyer intent?

The behavioral signals that tend to be most informative include: repeat purchases made without promotional prompting, the number of pre-purchase site visits and the content consumed during those visits, feature usage patterns in digital products, the specific timing of churn, and whether customers who say they are satisfied actually behave in ways consistent with satisfaction (such as referring others or upgrading). These signals do not require customers to explain themselves, which is precisely why they are reliable.

Is there any value in asking customers why they bought?

Yes, but with specific caveats. Customer interviews, particularly in-depth one-on-one conversations that use open-ended prompts and narrative techniques, can surface the metaphors, emotions, and associations customers hold around a product or brand. These insights, when they complement rather than replace behavioral observation, add genuine texture. The problem is not asking. The problem is treating the answer as a primary source of truth without checking it against what customers actually do.

How does revealed preference theory apply to modern marketing?

Revealed preference theory, developed by economist Paul Samuelson, holds that consumer behavior observed under real conditions is a more accurate indicator of preference than stated intent. Applied to marketing, this means that what customers repeatedly choose, especially under conditions where alternatives exist and no artificial incentives are in play, is the most honest signal available about what they actually value. Modern behavioral analytics, repurchase data, engagement patterns, and product usage metrics are all practical expressions of the same underlying idea.