Beyond the Viewability Threshold: The Evolution of Integral Ad Science’s ‘Quality Attention’

In the high-stakes arena of digital advertising, the industry has long relied on “viewability” as its North Star. Defined by the Media Rating Council (MRC) and the IAB, an impression is deemed viewable if 50% of its pixels are present in an active browser window for a single continuous second. While this standard provided a necessary baseline for cleaning up the chaotic programmatic ecosystem, it suffers from a fundamental limitation: it measures an opportunity to see, not the act of seeing.

Integral Ad Science (IAS) has sought to bridge this gap with "Quality Attention," a proprietary scoring system designed to rank advertising impressions on a scale of 1 to 100 based on their likelihood of being noticed and contributing to tangible business outcomes. By moving beyond binary viewability, IAS is attempting to commoditize the nebulous concept of human attention, transforming it from a psychological observation into a tradable programmatic currency.

The Architecture of Attention: How the Score is Built

IAS does not rely on real-time eye-tracking for every impression—a logistical impossibility at the scale of modern programmatic buying. Instead, the company utilizes a sophisticated machine learning model, the structure of which was established in its foundational Taking Action on Attention research.

The model synthesizes three distinct families of signals to generate its composite score:

1. Visibility Signals: These capture the physical presence of the ad, including standard viewability, total time-in-view, and video quartile completion rates.
2. Situation Signals: These account for the environment where the ad appears. Key factors include ad density, the percentage of the screen occupied by the creative, the device type, and the specific ad format.
3. Interaction Signals: These measure user behavior while the ad is live, such as scrolling patterns, pausing or resuming video playback, skipping, and volume adjustments.

The "human" element is injected into this machine-learning engine via a partnership with London-based Lumen Research. Lumen conducts consented, biometric eye-tracking panels. While the data from these panels is not applied to every individual ad impression, it serves as the ground-truth training set for the IAS algorithm. Consequently, Quality Attention is, in effect, a prediction dressed in the vocabulary of observation.

The resulting composite score categorizes inventory: a score of 65 or higher signifies "above-average" attention, while scores between 1 and 64 are considered below average. It is critical to note that these thresholds are vendor-defined, not industry-standardized.

Chronology of a Product Ecosystem

The journey of Quality Attention from an experimental measurement tool to a core programmatic utility has been marked by rapid iteration and corporate restructuring.

• June 15, 2023: IAS announces the inaugural version of Quality Attention, a post-bid measurement tool. At launch, the company claimed the model had access to 280 billion daily digital interactions and touted a 198% lift in conversion rates for high-attention impressions.
• January 4, 2024: General availability is achieved with the formal integration of Lumen Research’s biometric data. The performance claims are refined to a 130% conversion lift, alongside 91% higher brand consideration and 166% higher purchase intent.
• July 31, 2024: IAS expands the product to mobile in-app environments, a critical move given that mobile captures the vast majority of digital ad spend.
• October 2024: A publisher-facing extension is launched, allowing inventory owners to audit their own placements and redesign low-scoring pages.
• December 17, 2024: The beta launch of "Quality Attention Optimization" marks a pivot from passive reporting to active bidding. Buyers can now use these scores as pre-bid segments within demand-side platforms (DSPs).
• January 2025: Social Attention measurement is introduced to bypass the “walled garden” restrictions that prevent traditional IAS tagging on platforms like Meta or TikTok.
• September 24, 2025: In a massive industry shift, Novacap agrees to acquire IAS in an all-cash deal valued at $1.9 billion, signaling a new chapter for the company’s product roadmap under private ownership.
• June 30, 2025: A bespoke Snapchat attention score is unveiled, further deepening the integration with major social platforms.
• June 4, 2026: IAS publishes a landmark case study with automotive giant Stellantis, analyzing 13 billion impressions across 19 countries. The results showed a 165% increase in engagement and a 33% improvement in ad recall.

Supporting Data and Evidence: The Vendor’s Dilemma

The Stellantis case study remains the most significant deployment of Quality Attention to date. By running the program for over 18 months in some regions, the project provided a rare long-term view of attention metrics. However, skeptics point to a persistent issue: the most compelling performance data for Quality Attention is generated, analyzed, and published by IAS itself.

While independent academic studies on attention exist, they rarely mirror the proprietary logic of vendor-led tools. For instance, a major Kantar study analyzing over $3.2 billion in spend found that channel-level attention did not always correlate with brand contribution or cost-effectiveness. The study notably found that Meta, despite lower "attention" scores compared to some premium display inventory, remained highly cost-effective for advertisers.

Official Responses and Industry Standardization

The industry’s relationship with attention metrics is currently defined by a tension between utility and definition. On November 12, 2025, the Media Rating Council (MRC) and the IAB finalized attention measurement guidelines. Their stance is clear: attention should not be used as a proxy for campaign outcomes. They define it strictly as an exposure and engagement signal.

IAS, however, continues to market the product using performance-based language, such as “conversion lift.” This creates a divergence between the regulatory framework and the commercial application of the product.

Furthermore, there is a significant issue of accreditation asymmetry. While competitors like DoubleVerify (DV) secured MRC accreditation for their "Authentic Attention" product as early as 2023, IAS has yet to secure a similar stamp for Quality Attention. This leaves buyers in a position where they must weigh the potential benefits of the IAS product against the lack of third-party audit, comparing it against metrics like Adelaide’s "Attention Unit" or DV’s indices—none of which are interchangeable or standardized.

The Strategic Implications for Buyers

For the modern media buyer, the core question is whether the marginal gain in performance justifies the cost of the measurement and the resulting reduction in available inventory.

1. The Filtering Effect

By applying a Quality Attention filter, advertisers effectively shrink their addressable pool of inventory. If a brand only bids on impressions scoring 65+, they may find their scale diminished. The decision to implement these filters is therefore a strategic trade-off between "quality" and "reach."

2. The Move Toward Bidding Logic

Attention has successfully migrated from post-campaign PowerPoint decks into the actual bidding logic of the programmatic stack. As Google began allowing Adelaide scores into DV360, and Index Exchange integrated similar signals into their SSP, the market has moved toward a model where “Attention” is treated as a programmatic bid signal—much like geography, device type, or audience segment.

3. The "Black Box" Criticism

A structural concern remains: IAS occupies multiple points of the value chain. They provide the quality signals, they own the model that turns those signals into a score, they sell the pre-bid segments that act on those scores, and they provide the reporting on the success of those segments. This consolidation of power means that the "Attention" ecosystem is largely a closed loop, leaving little room for independent verification of the model’s weightings or logic.

Looking Ahead: The Future of Attention Measurement

As 2026 progresses, IAS has shifted its focus from simply refining the attention score to linking it with hard business outcomes. The partnership with Mastercard—connecting media quality to actual purchase data—is a significant attempt to provide the "proof" that the MRC demands.

Additionally, the rise of AI-driven reporting tools, such as "IAS Agent," suggests that the company is trying to reduce the "black box" stigma by offering natural language insights. However, the next frontier remains Connected Television (CTV). With 82% of industry experts expecting CTV attention to become a top priority, the race is on to see if IAS can maintain its dominance in a channel where rivals are already moving to capture the living room.

Ultimately, Quality Attention represents a sophisticated attempt to solve the "opportunity to see" problem. It is a powerful tool for optimization, but it is not a panacea. Advertisers who adopt these metrics must remain mindful that they are dealing with a proprietary prediction model, not a universally audited standard. As the industry matures, the true test for Quality Attention will be whether it can evolve from a vendor-specific benchmark into a transparent, cross-platform standard that earns the trust of the wider digital marketing ecosystem.