Quality Attention is a proprietary scoring system sold by Integral Ad Science (IAS) that rates an individual advertising impression on how likely it was to have been noticed by a person and to contribute to a business result. Each impression receives one composite number on a scale to 100, plus three sub-scores. The product exists because viewability, the accredited benchmark that counts an impression as viewable once half its pixels sit in an active browser window for one continuous second, records an opportunity to see rather than the act of seeing. Quality Attention is one commercial attempt to put a price on that gap.

The distinction matters commercially rather than academically. Two impressions can both clear the viewability threshold while one appears alone above the fold on a news article and the other sits fourth in a stack of refreshing units on a made-for-advertising page. Viewability scores them identically. IAS built Quality Attention to score them differently.

How the score is built

IAS settled on three families of signal after analysing its own measurement traffic, a conclusion the company set out in its Taking Action on Attention research and has kept as the structure of the product since.

Visibility signals cover whether the impression was genuinely present: viewability, time in view and video quartile completion. Situation signals describe the environment the impression landed in: ad density, the share of screen the creative occupied, device type and ad format. Interaction signals track what the person did while the ad was on screen: scrolling, pausing, resuming, skipping, starting playback and changing volume.

Those inputs feed a machine learning model that IAS says was trained on billions of impressions and millions of conversion events, producing a composite figure where 65 or higher is classified as above-average attention and anything from 1 to 64 sits below average. The thresholds are the vendor's own, not an industry convention.

The eye-tracking layer comes from Lumen Research, a London company that runs consented biometric panels. One mechanical point is frequently misread in trade coverage: no ordinary impression is eye-tracked. Lumen panel data trains the model, and the model then infers likely attention from the visibility, situation and interaction signals available on every impression. Quality Attention is a prediction dressed in the vocabulary of observation, which is a design choice with consequences discussed further below.

Where it sits in the buying workflow

At launch the score was post-bid only, surfaced in IAS reporting after delivery, useful for deciding which publishers, formats or programmatic supply paths to keep. That limits it to diagnosis.

Activation arrived later. Quality Attention Optimization, released in beta on December 17, 2024, converts the model into pre-bid segments a buyer applies inside a demand-side platform, steering bidding toward inventory predicted to score well before the impression is bought. On the sell side, an extension announced in October 2024 gave publishers access to the same scores against their own inventory, on the theory that an inventory owner who can see which placements score badly can redesign the page.

Walled gardens required a separate route. Social platforms do not permit the measurement tags IAS runs across the open web, so Social Attention measurement, live for IAS customers from January 2025, combines platform-supplied exposure data with Lumen modelling. A bespoke Snapchat score followed on June 30, 2025, built with Snap and Lumen and reported inside the IAS Signal interface.

Origin and evolution

The first version was announced on June 15, 2023 as a post-bid measurement product, built on what IAS then described as access to more than 280 billion daily digital interactions. That release already carried the three-signal structure, and cited an average conversion rate lift of 198% for high-attention impressions against low-attention ones.

General availability came on January 4, 2024, when IAS launched Quality Attention with Lumen Research as an integrated product. The headline performance figure changed at that point to a lift of up to 130%, alongside 91% higher brand consideration and 166% higher purchase intent. IAS has not published a reconciliation of the two numbers, and both rest on its own analysis.

Coverage widened through 2024. Mobile in-app support arrived on July 31, 2024, addressing an environment forecast to take 82% of mobile advertising spend that year. The publisher extension followed in October, and the optimisation product reached beta in December 2024, announced from Singapore with Chief Product Officer Srishti Gupta framing attention as the answer to two buyer questions: whether the budget is working, and what can be done about it.

Ownership changed around the product. Novacap agreed to acquire IAS for $1.9 billion on September 24, 2025 at $10.30 per share, taking the company private.

What the evidence looks like

The largest published deployment involves Stellantis. The automotive group ran Quality Attention with Publicis Media across 13 billion impressions in 19 countries, in a case study IAS published on June 4, 2026 reporting a 165% increase in engagement on high-attention inventory and a 33% improvement in ad recall among target audiences. The programme began in 2024, ran more than six months globally and as long as 18 months in France, where survey company Dynata measured the brand lift. Stellantis also activated Total Visibility, the IAS supply path product, to set attention against cost.

The dataset is unusually large for an attention study. It is also a vendor case study about a vendor product, published by the vendor, which is the standard on which most attention performance claims currently rest.

Why it matters for media buyers

Attention has been migrating out of post-campaign decks and into bidding logic, and Quality Attention Optimization was an early instance of that move. The direction has since become general: Google made Adelaide Attention Unit scores available as a Display & Video 360 custom bidding signal from mid-July 2026, Index Exchange embedded xpln.ai signals into its supply-side platform on February 12, 2026, and Nielsen brought Adelaide's metric into its Outcomes Marketplacein October 2025.

For a buyer, the practical question is not whether attention is real but whether a vendor-specific score changes allocation enough to justify the measurement fee and the reduction in addressable supply that any quality filter imposes. Attention metrics have influenced programmatic decisions while standardisation stayed unresolved, which leaves buyers comparing numbers that are not comparable.

Limitations and disputes

Three problems sit on the product, and the first is definitional. The Media Rating Council and IAB finalised attention measurement guidelines on November 12, 2025 which state explicitly that attention should not be treated as a measure of outcomes for evaluating campaign performance, describing it instead as an exposure and engagement signal. Most vendor marketing around attention, including the conversion lift figures attached to Quality Attention, argues precisely the opposite.

The second is accreditation asymmetry. DoubleVerify secured MRC accreditation for its DV Authentic Attention metrics on January 4, 2023. IAS holds accreditations for third-party Amazon DSP impressions, viewability and invalid traffic, granted on November 13, 2025, and for YouTube viewability from March 2024, but has announced no equivalent accreditation for Quality Attention. Two competing attention products therefore carry different levels of independent audit.

The third is empirical. Kantar analysed 873 multi-media campaigns representing more than $3.2 billion in spend over five years and found that channel-level attention did not correlate with brand contribution or cost-effectiveness, with Duncan Southgate concluding that some advertisers may be overinvesting in high-attention placements. Meta ranked among the more cost-effective channels despite comparatively low attention levels.

A structural criticism applies to IAS specifically. The company supplies the media quality inputs, trains the model that converts them into a score, sells the pre-bid segments that act on the score and reports the outcome. Independent validation of any link in that chain is thin.

Not the same as

Viewability is a binary threshold test defined by the MRC and IAB; Quality Attention consumes viewability as one inputrather than replacing it.

DV Authentic Attention is DoubleVerify's competing product, built on exposure and engagement indices across more than 50 data points and benchmarked against a baseline of 100. Scores from the two vendors are not interchangeable.

Attention Unit, Adelaide's metric, scores placements rather than impressions and is licensed into third-party platforms as a currency. Quality Attention is measured and activated inside the IAS stack.

Attentive seconds, the Lumen panel measure, is an observational quantity. Lumen supplies training data to IAS, which makes it an ingredient of Quality Attention rather than a rival to it, a relationship the same company also maintains with Teads, Snap and Netflix.

Recent developments

IAS has spent 2026 attaching outcome data to its quality signals rather than expanding the attention score itself. A partnership with Mastercard announced on March 26, 2026 links media quality to anonymised purchase data for in-flight optimisation, with United States availability from the second quarter. Quality Connect opened IAS settings data to publishers on June 18, 2026, and IAS Agent, announced on December 16, 2025, put a natural language layer over campaign reporting. Connected television remains the gap: the IAS Australia Industry Pulse Report published in December 2025 found 82% of surveyed Australian media experts expecting CTV attention measurement to matter in 2026, a channel where rivals have moved faster.

Timeline

  • June 15, 2023: IAS announces Quality Attention as a post-bid measurement product in beta, citing a 198% average conversion rate lift.
  • January 4, 2024: General availability, integrating Lumen Research eye-tracking data; headline claim becomes up to 130% conversion lift.
  • March 2024: IAS earns MRC accreditation for third-party YouTube viewability measurement.
  • July 31, 2024: Quality Attention extends to mobile in-app environments.
  • July 2024: Kantar publishes analysis of 873 campaigns questioning the link between channel attention and effectiveness.
  • October 2024: Quality Attention extended to publishers.
  • December 17, 2024: Quality Attention Optimization released in beta; Lumen partnership expanded to cover social measurement.
  • January 2025: Social Attention measurement goes live for IAS customers.
  • June 30, 2025: Snapchat-specific attention measurement launches with Snap and Lumen.
  • September 24, 2025: Novacap agrees to acquire IAS for $1.9 billion at $10.30 per share.
  • November 12, 2025: MRC and IAB finalise attention measurement guidelines stating attention is not an outcome measure.
  • November 13, 2025: IAS receives MRC accreditation for third-party Amazon DSP impression, viewability and invalid traffic measurement.
  • March 26, 2026: IAS and Mastercard announce a partnership linking media quality signals to purchase data.
  • June 4, 2026: IAS publishes the Stellantis case study covering 13 billion impressions across 19 countries.
  • July 2026: Adelaide Attention Unit scores become available as a bidding signal inside Display & Video 360.

Summary

Who: Integral Ad Science, a verification and measurement company owned by Novacap since 2025, builds and sells the product. Lumen Research supplies the eye-tracking data behind the model. Advertisers, agencies and publishers are the buyers, with Stellantis, Publicis Media and Sanofi among the named users.

What: An impression-level attention score running to 100, derived from visibility, situation and interaction signals weighted by a machine learning model trained on Lumen biometric panel data, with 65 or above classified as above-average attention.

When: Announced on June 15, 2023 as a post-bid product, generally available with the Lumen integration on January 4, 2024, extended to mobile in-app in July 2024 and to publishers in October 2024, and made actionable pre-bid through Quality Attention Optimization from December 17, 2024.

Where: Open web display and video, mobile in-app, and social platforms through the separate Social Attention measurement, reported inside IAS Signal and activated as pre-bid segments within demand-side platforms.

Why: Viewability establishes that an ad could have been seen, not that anyone looked. Quality Attention scores the difference so that buyers can shift budget toward inventory the model rates highly, though the MRC and IAB guidelines caution against reading attention as a measure of campaign outcomes, and the strongest performance claims attached to the product come from IAS itself.