Google will let Display & Video 360 advertisers bid against Adelaide's Attention Unit score from the middle of this month, adding a third-party media quality metric to the custom bidding engine that governs how much each impression is worth.

The change arrives through a Display & Video 360 announcement published in the platform's Help Center, dated for a mid-July rollout. According to Google, advertisers will soon be able to optimize for attention by using Adelaide's Attention Units as a signal within Custom Bidding. The announcement describes two activation paths: an off-the-shelf algorithm and a script-level integration.

What the integration actually does

Custom Bidding is the mechanism inside DV360 that lets buyers write their own valuation logic. Rather than accepting the platform's default optimization toward clicks or conversions, an advertiser can define a scoring function that assigns a value to each available impression, and the bidding system then prices bids according to that score. The feature has existed in the platform for years. When Google first rolled out Custom Bidding in DV360, signals were limited to what could be measured as a Floodlight activity or a Google Analytics 360 goal, with the platform advising that buyers needed a basic working knowledge of Python to write the scripts.

The Adelaide integration extends that signal pool outside Google's own measurement stack. According to the announcement, advertisers will be able to activate an off-the-shelf "Maximize Adelaide Attention Unit" algorithm, or use the Attention Unit as a multiplier within their own custom bidding scripts, in order to enhance inventory quality and campaign performance.

Those are two materially different implementations. The packaged algorithm makes attention the optimization target itself: the system bids toward impressions with higher predicted AU scores, in the same way a Maximize Conversions strategy bids toward impressions predicted to convert. The multiplier approach is subordinate. An advertiser keeps an existing scoring model, built around conversions, revenue weighting, or viewability, and applies the AU value as a coefficient that raises or lowers the computed worth of each impression. A buyer running a weighted conversion model can therefore keep the conversion logic intact while penalising placements that score poorly on attention.

Custom Bidding has historically carried inventory restrictions. Google's documentation has previously specified that the feature does not operate against YouTube, Gmail, Audio, Tag Guaranteed, or Programmatic Guaranteed inventory. The mid-July announcement does not restate which inventory types the Adelaide signal will cover, nor does it name a specific calendar date within the month.

The metric being imported

Adelaide's Attention Unit is not an eye-tracking measurement. It is a predictive score. The company's methodology analyses signals including ad size, time in view, clutter, and position to predict a placement's likelihood of capturing attention and delivering business results. AU is derived from a combination of session-level page data and outcome data, which allows the model to weigh factors beyond simple exposure duration.

That distinction matters for how the signal behaves inside a bidding engine. Attention metrics in the market divide broadly into observational methods, which use panels and eye-tracking hardware to record what people actually look at, and predictive methods, which infer likely attention from placement characteristics. Predictive scores can be computed for every impression in an auction. Panel-derived scores cannot, because panels cover a sample rather than the full inventory pool. A bidding algorithm needs a value for each impression it prices, which is why predictive models have been the ones to reach pre-bid and bid-time systems first.

The AU metric has accumulated distribution across the measurement industry over the past two years. Comscore integrated Adelaide's attention metric into its certified deal IDs through PubMatic's supply-side platform in June 2025, combining publisher rankings from its MMX and Video Metrix systems with attention scoring for private marketplace transactions. Nielsen and Adelaide integrated AU with reach data in October 2025, producing efficiency metrics that assess reach and attention together inside Nielsen ONE. That same month, Uber Advertising launched a platform-specific custom AU model built with Adelaide and Kantar, fusing brand lift results with predictive attention scoring.

Six weeks after Amazon

The competitive timing is difficult to ignore. On June 10, 2026, Adelaide made its AU pre-bid segments available inside Amazon DSP, covering Display, Online Video, and Streaming TV inventory. That launch packaged the metric into two products. AU Media Quality segments inventory into High, Average, and Low performance tiers, with the Low tier serving as an exclusion layer. AU Quality Floor operates as a single-step waste filter, excluding both made-for-advertising designated sites and the bottom 10% of AU-scored placements in one action, using Jounce Media's supply chain quality data. The Quality Floor product is exclusive to Amazon DSP.

The two integrations are not equivalent in mechanism. Amazon's implementation is pre-bid targeting: the buyer defines an eligible inventory pool before the auction, and impressions outside that pool are simply not bid on. Google's implementation is bid-time valuation: all eligible inventory remains addressable, but the price the advertiser is willing to pay varies continuously with the attention score. Targeting removes supply. Bidding reprices it.

For a buyer, the practical difference concerns scale. A pre-bid exclusion that removes the bottom decile of inventory removes that inventory absolutely, regardless of how cheaply it might have been available. A multiplier permits the same inventory to be purchased when the discount is steep enough to justify the lower predicted attention. Which approach produces better outcomes depends on whether an advertiser treats low-attention inventory as waste or as a pricing question.

Evidence from the platform

Attention-based custom bidding in DV360 has documented precedent. EssenceMediacom France ran a test campaign for a consumer packaged goods advertiser over one month in July 2024, using custom bidding scripts that scored impression combinations based on URL and device parameters. The implementation produced CPM rates 7.3% more cost effective than standard automated bidding, with improvements across several performance indicators. The agency, led by Programmatic Director Pascaline Montigny, presented the methodology during a Google DV360 podcast recording on June 18, 2025.

That result was achieved with hand-built scripts. The advertiser constructed the attention proxy itself rather than consuming a vendor score. The mid-July change removes that construction step for buyers who want it removed, while preserving script-level control for those who do not.

Wider evidence on attention optimization comes largely from vendors with a commercial interest in the category, a caveat that applies to most figures circulating in this area. Integral Ad Science demonstrated up to 130% conversion rate improvements comparing high against low attention impressions when it expanded Quality Attention into optimization. A Stellantis programme with IAS and Publicis Media measured more than 13 billion impressions across 19 countries, reporting a 165% increase in engagement on high-attention inventory, a 33% lift in ad recall among males aged 18 to 44, an 18% improvement in brand favorability, and a 29% increase in conversion rates. The full case study was published on June 4, 2026.

Standardization remains the unresolved question. In November 2025, the MRC and IAB finalised attention measurement guidelines that established minimum transparency and comparability requirements, while cautioning that attention should not be treated as a direct measure of campaign outcomes. Different vendors compute attention differently, and a score of 60 in one system does not correspond to a score of 60 in another. An advertiser optimizing toward AU inside DV360 while a sister campaign optimizes toward a different vendor's metric elsewhere is not running comparable campaigns.

Why inventory quality became a bidding problem

The framing in the Google announcement points at inventory quality, and that phrase carries weight in the current market. Made-for-advertising sites, properties built to generate advertising revenue rather than serve audiences, have proven persistent. The Association of National Advertisers has estimated that such sites drain roughly 15% of total programmatic ad spend. IAS documented a 5% increase in traffic to made-for-advertising and ad clutter environments over December 24 and 25, 2025, compared with the preceding week, and reported in its most recent Media Quality Report that mobile web display carried a 2.0% MFA rate against 0.5% on desktop, with the channel generating 71.9% of all MFA impressions measured.

Attention scores function as an indirect filter against that category. A page dense with advertising units, where each placement competes for the same viewer, tends to score poorly on the clutter and position inputs that feed a predictive attention model. An advertiser bidding on AU is bidding away from MFA characteristics without needing to maintain an exclusion list.

That logic explains why attention signals have migrated toward bid-time systems across the industry rather than staying in post-campaign reporting. Index Exchange embedded xpln.ai attention metrics into its supply-side platform on February 12, 2026, with the xpln.ai system capturing 20 to 25 exposure signals per impression and converting them into a single predictive KPI through models trained on eye-tracking datasets. Lumen Research brought attention measurement to Netflix advertising across five European markets in March 2026.

Context inside a platform in transition

The Adelaide integration lands during an extended period of change in DV360. Google published a roadmap of twelve platform changes running through August 2026, covering audience expansion deprecation, reporting migrations, and a shift to bidding-aligned creative selection estimated for August 12, 2026. Under that change, fixed bidding, maximize viewable impressions, maximize custom value, and cost-based bidding strategies will rotate creatives evenly, while conversion- and click-maximizing strategies will favour creatives predicted to perform.

The platform has also been narrowing manual controls. Google removed bid multipliers from DV360 in May 2025, directing advertisers toward auto bidding or custom bidding with rules as the alternative. That removal made Custom Bidding the primary route for buyers wanting granular impression-level valuation. Adding a third-party attention score to that route expands what the surviving mechanism can express.

Custom Bidding itself has been extended repeatedly. Google added CTV-specific multipliers on October 7, 2024, covering content duration and delivery type, allowing distinction between livestream and video-on-demand inventory. The DV360 API reached version 4 general availability on March 27, 2025, with dedicated algorithm resources for programmatic implementation of custom bidding logic.

Open questions

The announcement leaves several operational details unstated. It gives no specific launch date within mid-July, no indication of whether Adelaide charges separately for the signal inside DV360, and no detail on which inventory types the algorithm supports. Whether the off-the-shelf algorithm requires an existing Adelaide contract, or whether the score becomes available to any DV360 advertiser, is not addressed in the published text.

The question of measurement independence also sits unresolved. A demand-side platform optimizing against a third-party quality score is, in effect, allowing an outside party to influence which supply wins auctions. That has been a stated goal of the attention measurement sector for several years. It also concentrates a degree of market power in whichever attention vendor achieves the widest DSP distribution. Adelaide now holds integrations with the two largest programmatic buying platforms outside The Trade Desk, having reached Amazon DSP in June and DV360 in July.

Timeline

Summary

Who: Google, through its Display & Video 360 demand-side platform, and Adelaide, the attention measurement company behind the Attention Unit metric. The change affects programmatic buyers, agencies, and advertisers running campaigns through DV360 Custom Bidding.

What: Adelaide's Attention Units become available as a signal inside DV360 Custom Bidding, through an off-the-shelf "Maximize Adelaide Attention Unit" algorithm and as a multiplier applicable within advertiser-written custom bidding scripts. Google states the purpose as enhancing inventory quality and campaign performance.

When: The Display & Video 360 Help Center announcement gives an expected availability of mid-July 2026, without a specific calendar date.

Where: Inside Display & Video 360, Google's enterprise programmatic platform, at the Custom Bidding layer that determines how much an advertiser will pay for an individual impression.

Why: Attention measurement has been moving from post-campaign reporting into live bid decisioning across the programmatic supply chain, driven by persistent inventory quality problems including made-for-advertising sites estimated to absorb around 15% of programmatic spend. The integration follows Adelaide's June 10, 2026 arrival inside Amazon DSP as a pre-bid targeting signal, placing the metric inside two of the largest programmatic buying platforms within six weeks. It also expands what Custom Bidding can express at a point when Google has removed other manual bid controls from the platform.