Why Demand DNA May Matter More Than Data Exhaust In The Age Of AI

The advertising industry is making one of the largest technology bets in its history. Billions of dollars are being invested in identity graphs, clean rooms, customer data platforms, data collaboration technologies, and agentic AI systems designed to automate marketing decisions. Recent transactions, such as Publicis acquiring LiveRamp , and investments across the industry reflect a growing belief that the future belongs to organizations capable of connecting the most consumer data.

The assumption seems logical. More data should create better insights. Better insights should create better predictions. Better predictions should create better business outcomes. But what if the industry's biggest assumption is wrong? What if more data does not necessarily create better predictions? What if the real challenge is not connecting more data but understanding whether the data being connected contains the information marketers actually need? For decades, digital marketing has been built on behavioral exhaust—the digital footprints consumers leave behind through transactions, website visits, searches, loyalty programs, mobile devices, and advertising interactions.

Behavioral exhaust is valuable. It tells organizations what consumers did. What it often fails to explain is why they did it. That distinction may become increasingly important as artificial intelligence takes on a larger role in business decision-making. Most AI systems learn from historical observations. They identify patterns, correlations, and relationships within existing data. But no matter how sophisticated the model becomes, its understanding remains limited by the quality and relevance of the information it receives. A model cannot learn information that does not exist in the dataset . In 2018, before all the recent investments in AI companies, research presented at the International Conference on Research in Advertising (ICORIA) by Northwestern University's Medill School of Integrated Marketing Communications found that models incorporating consumer-reported behavioral, attitudinal, and motivational variables achieved nearly 4X greater targeting accuracy than traditional demographic and transaction-based approaches. The finding is significant because it challenges one of the foundational assumptions of modern data science: that more observations automatically produce better predictions. Instead, the research suggests that causally relevant data—data that measures the factors driving future behavior—may be far more valuable than simply increasing the volume of observed outcomes. Human behavior does not begin with a transaction. It begins with psychology. Consumers experience emotions. Those emotions influence expectations. Expectations shape intentions. Intentions drive decisions. Decisions eventually become transactions. Yet most modern marketing systems begin measurement at the final stage of that process. By the time a purchase appears in a transaction database, the most important drivers of behavior may already be invisible: confidence, financial security, economic expectations, optimism, fear, stress, motivations, and aspirations. These factors often determine consumer decisions long before those decisions appear in behavioral systems. The industry's response has largely been to collect more data. More devices. More identifiers. More transactions. More clicks. More observations. More linkages. But connecting more observations does not automatically improve understanding if the observations themselves fail to capture the causes of behavior. Imagine two datasets. One contains ten billion retail transactions. The other contains ten thousand statistically representative consumers reporting their future spending plans, economic confidence, employment expectations, purchase intentions, motivations, and financial outlook. Most organizations instinctively assume the larger dataset is superior. Yet if the objective is forecasting future demand, the smaller dataset may contain more useful information because it measures the drivers of behavior rather than the record of behavior after it has already occurred. This is where a new concept may deserve greater attention: Demand DNA . Demand DNA represents the emotional, economic, psychological, and behavioral factors that create future purchasing behavior. Unlike behavioral exhaust, which records outcomes after decisions have been made, Demand DNA seeks to measure the conditions that produce those decisions in the first place. Behavioral exhaust tells marketers where consumers have been. Demand DNA helps explain where they may be going. Research by economists Francesco D'Acunto and Michael Weber has demonstrated that consumer expectations often improve economic forecasting because beliefs and intentions frequently change before behavior changes. In many cases, expectations contain information that has not yet appeared in transactions, government reports, or market indicators. Demirhan Yenigun, Ph.D. , former Adjunct Professor of Decision Sciences at George Washington University & University of New Hampshire, explains: “Consumer intentions and expectations are not simply opinions. When measured properly, they are leading indicators of future economic and purchasing behavior. Ignoring them means ignoring information that may not yet exist in transactional data.” This challenge becomes even more important as agentic AI enters the mainstream. Agentic systems are increasingly being asked to determine audiences, allocate budgets, optimize campaigns, recommend actions, and automate decisions. The quality of those decisions depends entirely on the quality of the signals feeding the system. If those signals primarily reflect historical activity, AI risks becoming extraordinarily effective at analyzing the past while remaining limited in its ability to anticipate the future. In effect, many organizations may be suffering from what could be called ‘causal blindness’ …the tendency to measure outcomes while failing to measure the forces that create those outcomes. A retail transaction often reveals very little about the underlying consumer. Was the purchase made for personal use or as a gift? Was it bought by the account holder, a spouse, a child, or a family member? Was the decision driven by necessity, aspiration, convenience, or celebration? The transaction itself rarely provides those answers. Yet many identity systems attempt to construct highly detailed consumer profiles based upon observations that require multiple layers of inference. The resulting models can become highly sophisticated. But sophistication should not be confused with understanding. As privacy expectations continue to evolve, organizations are also asking new questions about data lineage, transparency, and explainability. This has renewed interest in zero-party data (information intentionally and proactively shared by consumers) which reduces the need for inference and provides direct visibility into motivations, intentions, and preferences. The concept is already beginning to move from theory into practice. A growing number of organizations are developing AI-ready audience models that incorporate emotional, motivational, and psychological variables alongside traditional behavioral signals. Rather than targeting consumers solely based on what they purchased, these models seek to understand underlying drivers such as optimism, financial confidence, security needs, status motivations, prevention behaviors, and emotional well-being. As these approaches become more widely available through cloud-based AI ecosystems and data marketplaces, marketers may gain direct access to causal consumer signals that were previously difficult to measure at scale. Recent industry announcements and marketplace launches suggest that emotional and motivational consumer models are beginning to move from academic research into commercial deployment. The implications extend beyond audience targeting. These same emotional and motivational signals may ultimately improve forecasting, personalization, media allocation, product development, and agentic AI decision-making by providing models with information about why behavior occurs rather than simply what occurred. The broader lesson extends well beyond advertising. Artificial intelligence has triggered a race to build larger models, more sophisticated infrastructure, and increasingly connected datasets. Yet the future winners may not be the organizations that collect the most data. They may be the organizations that collect the most causally relevant data. The Northwestern research suggesting a fourfold improvement in targeting accuracy provides an important clue. Better outcomes may not come from feeding AI more behavioral exhaust. They may come from feeding AI better information about the human motivations, expectations, emotions, and intentions that create behavior in the first place. The advertising industry has spent decades building increasingly sophisticated plumbing. The next breakthrough may come from finally understanding the people behind the pipes. Disclosure: The consumer sentiment study referenced above was conducted by my company, Prosper Insights & Analytics . This is the same dataset used by the National Retail Federation, and available from Amazon Web Services, Bloomberg, and the London Stock Exchange Group for economic benchmarking