Economizing attention: Markets and barter in video advertising platforms

Video format as an economization device

A video format is the material device through which a platform organizes when, where, and how an advertisement appears on a screen, including its placement, triggers, timing, and interaction rules. Across digital video advertising, three formats have become dominant. An in-stream video ad is fully embedded within the data stream that generates the content the viewer sees, such as a pre-roll on YouTube, a mid-roll inside a TikTok video, or the ad breaks inside Netflix's ad-supported tier. An out-stream video ad—like the video that suddenly starts playing in the middle of a news article— is digitally added to that stream at the last minute. Connected TV (CTV) ads are those that appear on smart TVs via apps like Netflix, Hulu, or YouTube. Together, these three formats organize nearly all video advertising today, scripting how viewer attention is economized. But how? To explain this, we need to analyze the materiality of a video ad.

A digital video ad is often described “simply” as content with an ad attached (Lobato, 2019). Yet in reality, a video ad is not a mere media file paired with another. What looks like a straightforward attachment is actually the outcome of an extended, multi-agent, energy-intensive, latency-sensitive computational production process. In practice, a digital video ad is not {video + ad} but a composite object produced through encoding pipelines, delivery protocols, auction engines, verification systems, and prediction models. The seamless moment on the viewer's screen is the final output of a rapidly executed industrial supply chain.

Furthermore, video ad manufacturing is governed by protocols such as VAST, which specifies how video ads are requested and delivered, SIMID, which enables secure interactive features, OMID, which standardizes viewability and measurement, and OpenRTB, which structures real-time bidding auctions, all developed and maintained by industry bodies such as the IAB Tech Lab (2023). It is delivered across layered infrastructures that include browsers and apps, player engines, content delivery networks, ad servers, device operating systems, and data centers, each contributing to the rendering, verification, and monetization of impressions. These standards and infrastructures do not merely support advertising delivery; they actively structure how impressions are defined, measured, and made exchangeable within programmatic systems (Alaimo et al., 2018; Paparo, 2025).

The video ad does not simply emerge after simple processes of digital attachment. It is produced in a complex supply chain as the materials used to marketize ads are produced and exchanged with customers, or viewers. At the center of this production lies a deceptively simple economization formula:

Digital Video Ad = {V, S, t_0−x}

Unlike analogue advertising, digital video ads do not exist as stable, pre-given units. They are fleeting, reactive, and valuable only within tightly bounded, personalized intervals of time. Their economization depends on the synchronized presence of three elements: a viewer or viewers (V); a screen capable of rendering the ad (S); and a micro-temporal interval between t_₀ and t_x. Only when {V, S, t_₀−x} align does an impression come into being; if any element drifts—if attention falters, the device changes state, the temporal window ends—the economic object dissolves. In digital video, the commodity is not the ad file but the momentary conjunction of attention, device, and time. Without this triad, there is nothing to sell. Every impression is radically contextual and vanishes the instant attention shifts, the screen updates, or the micro-temporal interval expires.

In its simplest form: viewer × screen × time. A person encounters content on a device and remains present long enough to be measured. This triad is the minimal unit from which all other processes—valuation, segmentation, auctioning, optimization, and reporting—are constructed. Video formats are the devices with which platforms capture, stabilize, and convert this fragile triad into tradable inventory, each doing so according to its own material constraints and delivery architectures.

To empirically describe this in In-stream video, consider what happens when a viewer taps a YouTube video. A skippable pre-roll becomes active because the player timeline has reached a cue point. The player calls a VAST tag, which fetches a markup code (like XML) describing the ad, including skip-offset, companion creatives, tracking events, and interaction options. The platform computes whether a user is eligible to receive an ad—based on device type, network conditions, watch history, predicted completion probability, and brand-safety constraints. Several exchanges respond with bids, each calculated in microseconds. An auction selects the winning ad. Only then do the video file streams begin, and only then do the viewability and completion timers begin. In this moment, the viewer × screen × time triad is sliced into discrete economic intervals: the first 5 s (t₀₋₅ high-value), the next 15 s (t₅₋₂₀ medium-value), and the remainder (t_20−x low-value). The ad's position in the timeline becomes a valuation device with agency—a computational industrial clock for producing the ad as an economized object.

Out-stream video reorganizes the formula yet again. Here, the viewer is not watching one video; they are scrolling many shorter videos in, for example, Instagram feeds. The system waits until a video unit comes into the viewport—often just 50% of its pixels for at least 2 s—to activate a viewable impression. The advertisement is not tied to content; it is tied to the surface. Any page—news article, feed, blog, weather app—can be converted into latent video inventory because the DOM itself (Document Object Model, or the page's structure in live memory) is programmable territory. Autoplay initiates silently, obeying browser-level constraints; viewability thresholds kick in when economization begins; and a brief pause in the user's scrolling produces sufficient “time” to count as attention (note the barter exchange took place between the platform and customer). In out-stream, the device is not a player, but the page, and the economic unit is not watch-time but dwell-time. The page becomes an impression workshop, turning viewer + screen + time into salable surface-attention.

The third format of video, CTV, renders the same triad, but at the scale of a household or group of viewers (sometimes including even unrelated people in, for example, a waiting room: that is, “out-of-home,” in industry parlance). Here, the “viewer” can be plural—the source of endless measurement headaches for practitioner interviewees used to the individualization of web or advertising, although individuals can sometimes be identified if a “device graph” includes both the screen they are watching and their phone. The “screen” can be a 55-inch smart TV, governed not by browser policies but by the OEM's (i.e. the TV manufacturer's) operating system, which determines ad-pod placement, home-screen tile auctions, and pause-ad eligibility. Location is often a living room, the viewer is comfortably seated, “time” becomes long-form: a 42-min episode, two mid-roll clusters, a “pause ad” triggered if someone temporarily leaves the room.

Across all these settings, programmatic video integrates these three formats into a shared economic topology. Auctions unify the disparate materialities of In-stream, Out-stream, and CTV, making their differences commensurable. A mobile pre-roll with a skip button, a scroll-triggered out-stream unit, and a CTV mid-roll stitched server-side are radically different objects from a technical perspective. Yet in auctions, they can appear as comparable bids: a CPM (the cost of showing an ad a thousand times), an attention score, a predicted completion rate. The auction mechanism marketizes heterogeneous devices of attention capture into standardized, exchangeable units. Programmatic infrastructure performs a role crucial to marketization: it makes unlike things alike, long enough for them to be priced.

What looks to the viewer like a brief ad interruption, a clip sliding into view, is in fact the tip of an iceberg, a vast computational production process, an assembly line of economized objects. Yet before any ad can be monetized—before it can enter the world of markets—an entirely invisible layer of barter must take place. Users exchange data for access, attention for entertainment, and viewing labor for platform services. This barter stack is always at work before the market stack in digital video platforms, quietly producing the conditions under which impressions can be priced and sold.

Video advertising can be situated alongside other dominant formats such as search and display, which we analyze in detail elsewhere (MacKenzie and Caliskan, 2026). Across these formats, ads exist as contingent events defined by the alignment of a consumer, a screen, and a moment in time. They differ, however, in how inputs are generated and structured. Search advertising is organized around keywords that capture explicit user intent and are produced through highly dematerialized computational supply chains, while display advertising relies on contextual placement across publisher sites with relatively lighter creative and infrastructural requirements (Coromina et al. 2023, Mellet 2025). Video advertising, by contrast, depends on materially heavier processes, including large file transfers, streaming infrastructures, and extended creative supply chains (MacKenzie and Caliskan, 2026). The barter dyad and stacked economization are therefore not unique to video. They operate across formats, including in search, where data are exchanged for visibility, and in display, where attention is traded for access to content (Caliskan et al., 2025). What distinguishes video is the intensity and visibility of these processes: it amplifies the material, infrastructural, and organizational demands of platform economies, making them especially observable.

Barter as the first economization stack

Digital video economies rest on an often-invisible exchange that must occur before any ad can be monetized. The socio-technical life of impressions begins not with the auction (i.e. the market) but with the user's encounter with the platform. This is the domain of barter—the first layer in the stack of video ad economization—where production and exchange collapse into one another. Platforms depend on user behavior to generate the very ad opportunities they simultaneously sell. At this foundational layer, the basic unit of exchange is a barter dyad: two different valuables paired in an asymmetrical economic exchange process where viewers provide attention, data traces, watch-time, and/or engagement signals in return for access to content. These dyads (data vs. entertainment, attention vs. content, etc.) are embedded in the platform's design. Their exchange is made possible, formatted and scaled by programs whose decisive input is user data couples of platforms’ algorithms. Before any video ad becomes auctionable, platforms must format these behavioral materials so that an “impression” can be constructed. Barter thus precedes marketization: dyads must be paired and their barter is stabilized successfully before economic things can be commodified.

Formalist anthropological accounts challenged the assumption that barter was primitive or economically marginal, showing that it emerged in highly structured social, political, and institutional contexts. Chapman (1980) argued that barter appeared when actors lacked either the institutional means or the desire to use money, and that such transactions depended on pre-existing social relations, shared norms of equivalence, and culturally specific rules of obligation, not price-based calculation. Dalton (1982) similarly emphasized that barter was rarely a simple spot exchange; it was embedded in broader systems of hierarchy, reciprocity, and constraint, often occurring between strangers, across group boundaries, or in settings where credit and trust were limited.

Critiquing Dalton and Chapman's “formalist” accounts a decade later, Humphrey and Hugh-Jones (1992) emphasized that barter depends on negotiated equivalence without a general measure of value, making it fragile and context-specific, not universal and formal. Hart (1988) noted that barter persists among, for example, households and small firms in today's economies, although he argues that it remains peripheral because it requires face-to-face negotiation, shared expectations, and situation-specific trust. Zelizer (1994) offers an implicit critique of the assumption that exchange without money is simple: even intimate or informal barter involves boundary work and moral evaluation. Despite these contributions, however, barter is rarely a core analytic object of contemporary social research. Barter is understood either as limited and exceptional, occurring, for example, where money is absent, or as “thinly relational”: a one-off negotiation between parties without obligations.

These limitations leave a gap for understanding platform economies, where barter is continuous, infrastructurally enforced, and central to value creation. Platforms must actively construct equivalence between heterogeneous valuables through interfaces, thresholds, and predictive systems. To address this gap, we introduce the concept of the barter dyad. The dyad refers to the paired set of two heterogeneous, non-priced valuables that are exchanged in a barter relation. The dyad is not the relation itself, but its object-form: it is to barter what the commodity is to market exchange or the gift is to gift exchange. In platforms, these valuables are not held by whole actors but are co-produced and paired within infrastructures. They consist of fragments such as attention, behavioral traces, and watch time on one side, and content, access, and relevance on the other. The dyad thus names the unit through which barter is enacted, making visible how platforms extract, format, and couple dividualized signals into exchangeable forms that can later be stabilized and marketized.

Seen in practice, dyads organize economic participation across platforms. On TikTok, a labor–entertainment dyad emerges when a user lingers on a short video, as micro-temporal behaviors such as pausing or rewatching generate signals that feed algorithmic circulation while the platform returns content and relevance (Bhandari and Bimo, 2022; Maris et al., 2024). On YouTube, a data–entertainment dyad takes shape as watch duration, skips, and hover signals are recorded and translated into personalized recommendations (Banerjee and Pal, 2021). On Facebook Reels, a device–computation dyad is activated as the phone renders auto-play videos, exchanging energy and memory for platform use. On CTV, a household attention–visibility dyad forms as aggregated viewing behavior is exchanged for access to long-form content. Across these cases, dyads make visible how platforms organize barter through the continuous pairing of behavioral fragments with infrastructural outputs.

Dyads are not commodities: they are not necessarily fungible or priced, yet economically valued and exchanged. They are co-produced within architectures that platforms design and maintain, and their terms cannot normally be negotiated by users. A viewer encountering a dyad cannot change the terms of barter, but they can refuse the barter itself by not continuing to watch, thus dismantling the economization formula. When the V of {V, S, t_0−x} steps back, the economization fails.

These dyads structure what we call “prodexchange,” the defining logic of digital video: every user gesture is simultaneously consumption, production, and exchange, processed in the supply chain we are describing. TikTok's recommendation engine produces the next watchable sequence by predicting monetizable engagement from the viewer's last swipe. YouTube's encoding profiles are tuned to reduce buffering to protect the ad-stitching pipeline. Facebook's video player is both a rendering device and a pricing instrument, firing skip-event APIs, viewability timers, and attention signals that determine whether the impression will enter the auction. Even Netflix's emerging ad-supported tier now embeds tracking intervals into long-form content to prepare mid-roll “pods” (groups of ad impressions) for household-level targeting. In digital video, the viewer cannot simply watch; the act of watching generates future prediction, future ranking, and ultimately future auctionability. Prodexchange is this fusion, in which each viewing gesture produces the very object that will later be exchanged in the ad market.

Barter is complex and depends on an acceptable level of commensuration. If watching a cat video required sitting through a 5-min cat-food ad, the viewer would most likely reject the exchange; if it is only 5 s, they may accept it. There is, therefore, an implicit economic calculation of the non-monetary cost of barter to the viewer. Yet entertainment and behavioral data are not easily comparable, so platforms must actively construct their equivalence. TikTok attracts attention by making videos extremely short, transforming micro-attention into a payable unit that can be bartered instantly. YouTube's 5-s skip window defines the minimum attention cost for pre-roll access, making the exchange predictable and tolerable. Meta's out-stream ads trigger only when a video enters the viewport, ensuring users are charged attention only when necessary. Netflix uses long-form formats to extract extended durations of household viewing, converting 30- or 60-min segments into stable barter intervals for CTV ad pods. Across these platforms, commensuration is achieved through interface design, temporal rules, autoplay constraints, and thresholds that determine when the dyad is complete.

These processes of commensuration also point to a further feature of barter that is often overlooked: it does not require explicit recognition by those who participate in it. Barter can operate across a spectrum of awareness. Classic ethnographic accounts show that barter does not necessarily depend on mutual understanding, transparency, or even shared language. Practices such as silent trade and intergroup exchange demonstrate that actors can engage in barter without fully recognizing the terms of exchange or the identity of their counterpart (Humphrey, 1985). This insight is crucial for platform economies. In digital video environments, viewers may not explicitly recognize their participation in barter, as the exchange is embedded in interfaces, timing rules, and background processes. The dyad is enacted without being named. This opacity does not weaken the exchange; rather, it allows barter to be infrastructurally enforced while remaining experientially subtle. At the same time, its stability depends on the limits of tolerance. When the perceived terms of exchange shift, for example, when the duration or intrusiveness of compulsory ad exposure exceeds what viewers accept, the barter relation can weaken or collapse. Barter in platform settings is therefore not only structured but also continuously calibrated, operating within shifting thresholds that define its viability.

Barter is also mediated by qualculative agencies that judge whether the exchange has occurred sufficiently to proceed to marketization: “qualculative” action is simultaneously qualitative classification and quantitative measurement (Cochoy, 2002). TikTok's skip-prediction models determine whether a user is likely to tolerate a future ad, thereby influencing where and when the platform inserts breaks. YouTube's viewability engine assesses whether half the pixels remained in view for at least 2 s, translating this into an economic “enoughness.” Netflix's “household graphs” identify how multiple viewers in a living room barter their joint attention to long-form content. Meta's brand-safety classifiers evaluate whether particular content is “safe” enough to trade attention for ads. These agencies never merely measure activity; they decide whether a dyad has achieved sufficient value to produce an impression.

Barter is enacted in specific encounters that platforms design. On YouTube, the pre-roll moment is high-stakes barter: the viewer must provide at least apparent attention before accessing desired content. TikTok's infinite scroll transforms each swipe into a micro-barter in which attention is exchanged for relevance. Facebook's out-stream videos erupt between paragraphs, turning reading space into bartered visibility. Netflix's pause ads convert the act of pausing—a moment of user departure—into an opportunity for barter. CTV mid-roll breaks are placed at narrative tension points, encouraging viewers to remain seated and complete the barter. These encounters are architected with precision, not mere interruptions; they are designed settings through which dyads become actionable—the barter mode of economization.

Barter valuation determines when attention or behavior becomes “enough” to move the impression into the ad market. TikTok may treat a 2-s linger as a complete barter; YouTube treats full views, partial views, and skips differently; Facebook counts a view only when 1 s of video has been watched; Netflix evaluates commercial breaks by household-level watch continuity. Because these bartered goods are completely demarketized, their valuation must be constructed by thresholds, sensors, logs, and predictive models that convert micro-attention into a barterable input. Barter valuation, much like market valuation, is always contingent, always measured through devices, and always shaped by platform incentives.

The barter exchange system also requires continual maintenance. YouTube must update VAST and SIMID specifications, maintain skip-button logic, and tune video-player engines to support SSAI (“supply-side ad insertion”). TikTok must ensure scroll-trigger scripts fire reliably and that its recommender systems continue to produce watchable sequences. Facebook must keep out-stream rendering scripts compatible with browser autoplay policies. CTV platforms must sustain consistent ad-stitching so that mid-rolls appear seamlessly. Without constant technical, policy, and interface maintenance, the barter relationship collapses, and video ad impressions cannot be produced.

Finally, barter is enacted through a series of economic devices: skip buttons, autoplay engines, scroll-trigger APIs, mute-on-load constraints, rewarded-video interfaces, attention timers, household identity graphs, VAST tags, SIMID sandboxes, CSAI/SSAI architectures. These are not decorative user interface elements; they are devices with agency that structure the dyad, regulate the exchange, and prepare the resulting attention-data object for marketization. On Netflix, even the act of selecting a show initiates a sequence of device-level signals that stabilize future ad opportunities. On TikTok, the pause before a swipe becomes a device-level event used to construct monetizable predictions. On YouTube, the skip button is an extraction device, converting an unavoidable 5 s into a calculable, auctionable unit.

Across TikTok, YouTube, Meta, Netflix, and CTV ecosystems, barter emerges not as a marginal feature but as the primary mechanism through which video ad economies begin. Before the market stack can operate—before auctions, prices, or budgets enter—the barter stack must convert interactions into dyadic exchanges. Attention becomes data; data becomes prediction; prediction becomes valuation; valuation becomes auction; auction becomes revenue; revenue becomes platform power. Digital video economies rest on dyads that constitute the precondition for commodification and, ultimately, marketization. They are the foundation on which ad economization is built.

Marketization as second economization stack

Digital video advertising is often presented as if markets simply “exist” around it: buyers on one side, sellers on the other, and auctions clearing impressions in real time. Yet, as our broader framework insists, markets do not stand ready-made. They must be built, stabilized, and continually repaired. Before prices can form, before bids can be submitted, and before an impression can be sold, something must be made marketable. In digital video, that “something” is not the video file, not the creative ad, nor the viewer's attention in an abstract sense. Materially, it is the impression—a micro-temporal, device-specific, behaviorally contingent value that must first be commodified, then made comparable, and only afterward fed into the process of marketization.

Marketization begins with passiva(c)tion: the transformation of heterogeneous, unstable phenomena into objects—still active, but domesticated, so to speak—that can circulate in markets (Callon, Caliskan and MacKenzie, forthcoming). Video impressions are among the most unstable economic goods ever produced. They may last milliseconds; depend on motion, sound, and scroll position; vary by device, bandwidth, and user gesture; and may vanish before they are economized. They require extensive work before they become commodities. They must be passiv(a)ctivated into measurable, commensurable units; qualified, indexed, and calculated by human and non-human agencies; staged within market encounters that platforms design; priced through multi-layered bidding architectures; maintained through standards, repair work, and governance; and enabled by a dense ecology of devices.

Platforms accomplish this through a suite of formatting rules. The 6-s bumper, 15-s non-skippable, and 20-s standard pre-roll are not mere cultural norms: they are market units. Viewability rules translate a fleeting phenomenon into a stable commodity spec. For example, “skip windows” on YouTube create a standardized minimum unit of compulsory attention, passivactivating visibility as a marketable form of attention. Out-stream video adds additional layers: impressions count only once a video enters the viewport, often accompanied by scroll-trigger thresholds that define the minimal amount of human movement required to generate a saleable unit. In CTV, standardization emerges through ad pod lengths, break structures, and co-viewing assumptions that convert household time into uniform blocks of inventory. Through these formatting rules and thresholds, platforms perform the core act of marketization: turning unstable micro-temporal experiences into countable, comparable, purchasable goods.

Once video impressions have been formatted and thus passivactivated, they must be judged, evaluated, and classified. Marketization also depends on qualculation: the intertwined work of characterization and calculation (e.g. of value) through heterogeneous agencies. In digital video, these agencies include humans (advertisers, publishers, media buyers), organizations (DSPs, SSPs, verification firms), and an array of computational systems (viewability models, brand-safety classifiers, attention predictors).

DSPs qualify impressions by deciding whether they meet the advertiser's goals. SSPs qualify by filtering for deal-specific rules, brand safety, and “invalid” (e.g. fraudulent) traffic, although this last form of filtering is not always as rigorous as it might be. Viewability vendors (such as DoubleVerify) apply their own measurement models, sometimes contradicting platform-native viewpoints. Netflix's ad-supported tier qualifies household-level segments using co-viewing models and account-behavior clustering, while TikTok's ranking engine qualifies user micro-actions (rewatches, pauses, completions) as indicators of future monetizable attention.

Here, viewers are treated as Deleuzian dividuals (Cluley and Brown, 2015)—bundles of behavioral fragments distributed across devices and sessions. A “viewer” on YouTube is partitioned into watch-time segments, click sequences, device types, and skip probabilities. On Meta, the same user is decomposed into scroll velocity, dwell time, and interaction residue. CTV platforms qualify households through device graphs and inferred demographic clusters. Marketization is impossible without these qualculative agencies. They do not merely observe value; they help produce it.

Markets require encounters—moments in which buyers can meet goods under conditions engineered to make exchange possible. In digital video, encounters are tightly designed architectures. On YouTube, encounters occur at precise cue points: pre-roll, mid-roll, and end-roll insertions, each chosen by algorithms optimizing predicted user tolerance and ad-yield. TikTok organizes encounters through continuous swiping: every new clip is an encounter opportunity triggered by user gesture.

In CTV environments, encounters are shaped with even more precision. Ads are structured into pods; break positions are negotiated among multiple actors; and streaming apps enforce rigid pod structures regulated by device-level governance (e.g. Roku's operating system rules). Pause ads introduce encounters triggered not by narrative structure but by turning domestic gestures (getting a glass of water) into economic events. Market encounters are therefore engineered points of contact: timestamps, gesture triggers, viewport crossings, pod structures, and inactivity signals. They are the sites where passivated impressions become available for bidding.

Once impressions have been formatted, qualified, and staged within encounters, prices must be realized. In digital video, price realization is a multilayered process involving real-time bidding, pacing algorithms, soft floors, bid shading, and post-auction attribution. Most premium video now trades via first-price auctions, though soft floors discreetly enforce minimum prices. DSPs employ bid-shading models to avoid overpaying, while SSPs use dynamic flooring to maximize revenue. Out-stream video ads introduce additional complexity: milliseconds of exposure can trigger multiple auctions as the scroll position fluctuates.

CTV platforms layer even more pricing mechanisms. Programmatic guaranteed deals fix inventory at pre-negotiated CPMs, while private marketplaces (PMPs) allow preferred buyers to access premium pods. Netflix and Amazon Freevee integrate dynamic pricing based on household-type segmentation and time-of-day patterns. In many cases, the final price a buyer pays is the output of several nested calculations: auction clearing price + pacing adjustments + frequency caps + attribution rules. Price realization is therefore not a single moment but a distributed process, enacted across multiple systems whose outputs must be aligned for a transaction to count as revenue.

Markets do not run themselves. They require ongoing design, maintenance, and repair. Digital video ads depend on a heavy infrastructure that must be continually updated: CDNs, encoding farms, ad servers, identity graphs, measurement scripts, and fraud filters. Standard-setting bodies update VAST, SIMID, and OMID specifications, which platforms must implement and publishers must test. Player engines require constant maintenance to ensure skip events fire correctly, quartile beacons are sent, and mute states are captured. SSAI systems must keep ad stitching seamless to avoid user churn. Fraud detection must adapt to new forms of invalid traffic—CTV spoofing, out-stream misrepresentation, and viewability laundering.

The maintenance is endless: operating system updates break measurement scripts; app redesigns require re-certification; new creative formats require new validations; privacy rules require rewriting entire components of identity resolution. Markets function only because hundreds of maintenance teams across platforms and intermediaries keep repairing the conditions that allow impressions to be produced and monetized.

Finally, digital video ad markets are constituted and operated by devices—technical, organizational, and cognitive instruments that enable, shape, and sometimes distort market relations. VAST tags specify how ads should load, track, and report. SIMID provides the sandbox that enables interactive ads while restricting what they can access. OMID verifies that an impression was actually viewable. OpenRTB defines the grammar of bidding. Header bidding wrappers enable parallel auctions, redistributing power from platforms to publishers—or, in CTV, provoking new tensions as device manufacturers enforce their own rules.

On Netflix or Hulu, ad servers orchestrate the compatibility between creative requirements, pod structures, and user segments. TikTok's machine-learning recommendation engine acts as a market device that shapes both the supply of attention and the likelihood of future monetization. YouTube's player user interface—skip buttons, mute icons, double-tap gestures—functions as a hybrid device that both delivers content and produces data for future pricing.

In digital video advertising, market devices are agentic: they determine what is counted, measured, and valued, and in doing so constitute the market itself. Video impressions do not precede exchange; they are produced as marketable objects through formatting rules, qualculative operations, encounter architectures, and pricing systems. Through this socio-technical work, impressions become commodities that can be priced, traded, and optimized within the digital advertising infrastructure.