After the Attention Economy: Media Is Shifting to the Prediction Economy

From Competing for Eyeballs to Predicting Behavior

For the past two decades, the core narrative of the media and advertising industry has been the "attention economy": impressions, reach, and click-through rates formed the yardstick by which everything was measured. Whoever could capture more of users' time could win advertisers' budgets. But this logic is loosening its grip.

The source material puts forward a judgment: the media industry is turning into a "prediction economy." The implication of this shift is that the industry's value anchor is no longer just "how many people saw it," but "can we accurately predict who will do what and when." Attention is a stock; prediction is an increment. The former relies on scale, the latter on precision.

This transformation is no conceptual game. It means that media's business models, technology stacks, and competitive moats are all being redefined, and the way advertisers evaluate campaign performance is changing along with them.

The concept of the "attention economy" was first articulated by economist Herbert Simon in 1971: an abundance of information necessarily leads to a poverty of attention. Once the internet era arrived, the advertising and media industry concretized it into a business model centered on impressions, reach, and click-through rates, whose implicit premise is that users' time is a finite and scarce resource, and that the essence of platform competition is the battle for that time. Understanding this background helps clarify why, as users' time approaches saturation and ad tolerance declines, the growth logic aimed at "competing for screen time" hits a ceiling.

Why the Attention Economy Is Unsustainable

The predicament of the attention economy lies in diminishing marginal returns. When every platform is competing for the same chunk of time from the same pool of users, the room for impression growth tends toward saturation, while users' tolerance for advertising keeps declining. Click-through rate, as a performance metric, has long proven easy to misread and even to manipulate.

More critically, attention itself is a coarse-grained signal. Knowing that a piece of content was seen by a million people does not tell you which of those million will make a purchase, subscribe, or stay as a result. Advertisers are increasingly unwilling to pay for the vague "being seen," and instead demand a verifiable "prediction."

This constitutes the direct driving force behind the rise of the prediction economy: the demand side wants to move from "broadcasting to the masses" to "predicting with precision."

"Coarse-grained signal" means that an impression record typically only indicates that a certain piece of content was seen by a certain user, but cannot distinguish whether the user actively paid attention, scrolled past unintentionally, or was forcibly shown the content—let alone characterize their subsequent intent. By contrast, the prediction economy pursues fine-grained signals—behavioral sequences, dwell time, interaction paths, revisit intervals, and so on—data that can support probabilistic estimates of a single user's next action. The finer the grain, the closer to "actionable," and the greater the need for stronger data collection and modeling capabilities.

The Core Logic of the Prediction Economy

The underlying assumption of the prediction economy is that user behavior is more or less modelable and can be estimated in advance. The behavioral data, content consumption sequences, and interaction patterns accumulated by media platforms form the raw material for prediction.

For advertisers, this means that campaign decisions shift from "choosing channels, buying impressions" to "buying audiences, buying probabilities." The actual product platforms sell changes from ad space and time to the ability to predict whether a specific user will convert, churn, or upgrade.

For media companies themselves, predictive capability also feeds back into content operations: which topics work for which audiences at which times can be judged in advance based on historical data, rather than relying entirely on editorial intuition or after-the-fact reviews. Once this layer of capability is established, it becomes a moat that is hard to replicate in the short term.

The "feedback loop" is the key mechanism for whether the prediction economy can keep running: the platform makes a prediction based on the model (for example, whether a user is likely to convert), obtains the real outcome after the campaign runs, and then feeds the result back to calibrate the model's parameters, repeating the cycle. The value of the loop is that prediction accuracy improves with use, rather than remaining at a one-off modeling stage. A prediction system without a feedback loop will drift and fail over time—this is also the essential difference between it and a mere data analysis report.

Impact on the Industry Landscape

The barrier to entry for the prediction economy is higher than for the attention economy. The attention economy is a contest of traffic scale; new players can quickly build volume with content and subsidies. The prediction economy is a contest of data density, modeling capability, and feedback loops, requiring long-term accumulation and continuous calibration. This may widen the capability gap between leading platforms and small and mid-sized media.

At the same time, it pushes media companies into a position more akin to technology companies. Content production remains core, but the prediction systems, data infrastructure, and performance validation mechanisms surrounding content are becoming equally important assets.

For advertisers, the metrics for measuring campaigns will also migrate accordingly: from impressions and clicks to the accuracy of conversion probability estimates and the actual hit rate. This will force participants across the entire chain to realign their goals.

Questions Still Unanswered

The prediction economy does not automatically equal a better media ecosystem. Prediction accuracy depends on data, and the boundaries of data collection, users' aversion to being predicted, and the risk of prediction being used to manipulate rather than to serve are all unavoidable issues in this shift.

The source material offers a directional judgment on this trend but has not yet elaborated on specific implementation paths and industry disagreements. What is certain is that when media's core product changes from "attention" to "prediction," the industry's evaluation criteria, modes of competition, and ethical frameworks all need to be rebuilt in tandem.

One-line Summary

The media and advertising industry is shifting from the impression logic of competing for attention to the prediction logic of modeling user behavior.

Paragraph Summary

The article points out that the core value anchor of the media and advertising industry is shifting from the "attention economy" to the "prediction economy." The attention economy measures everything by impressions, reach, and click-through rates, but it suffers from diminishing marginal returns, declining user tolerance for ads, and the fact that impressions are only coarse-grained signals that cannot answer the key question of "who will convert." The prediction economy instead uses behavioral data, content consumption sequences, and interaction patterns as raw material, selling probabilistic predictions of whether a specific user will convert, churn, or upgrade. This shift raises the industry's barriers, relying on data density, modeling capability, and feedback loops, and may widen the gap between leading platforms and small and mid-sized media while pushing media companies closer to technology companies. The article also cautions that the accuracy of prediction brings unanswered ethical issues such as the boundaries of data collection, user aversion, and the risk of manipulation.