Connected TV ad targeting is getting a contextual boost with an AI assist.
Streamers may be looking to generate ad revenue with marketers turning attention to CTV, but amid fragmentation, a lack of transparency and consistent video metadata, and identity and privacy challenges looming, reaching the right person at the right time on streaming isn’t always easy for advertisers.
Contextual advertising could be one avenue to address some of these issues as it helps brands align their message with content consumers are watching, potentially enabling more effective targeting at the appropriate moment while avoiding privacy concerns.
While not a new concept, some publishers and vendors have taken recent steps to advance the potential of contextual advertising on CTV for ad sellers and buyers.
Vendors IRIS.TV and AppLovin-owned Wurl each recently introduced respective tools to help advertisers zero in on elements like emotion within content to better align ad creatives or more easily buy segments that match advertisers with the mood or other contextual aspects of programming.
The new products come as media companies are also trying to marry the right content and the right audience for advertisers. That includes Disney’s debut of “Magic Words,” a contextual capability in the beta phase that analyzes scenes and visuals across the Disney library so brands can tap into a specific moment, mood or emotion. And NBCUniversal last week unveiled a combination of content analysis with first-party datasets to create 300 emotion-based, GenAI-powered audience segments for advertisers. Roku last year ahead of NewFronts introduced AI-powered contextual capabilities for its free ad-supported streaming TV (FAST) service The Roku Channel.
As for the recent vendors announcements, Wurl on Tuesday introduced BrandDiscovery, a GenAI-based ad product that lets brands match CTV ads with content in real-time. The ad targeting tool provides scene-level information, so advertisers can align an ad’s emotion with the content that’s closest to the ad break. The scene-level contextual data matches ads based on emotion, genre and brand safety – all of which Wurl says helps improve brand awareness and purchase intent.
Wurl also promises scale, with BrandDiscovery ads shown across more than 300 premium CTV publishers, including FAST channels and leading streamer. Wurl offers more than 60 billion monthly available ad impressions and boasts its single source of access as providing advertisers with cost and time savings. The data segments from Wurl are free of charge and available to advertisers through their SSP (supply-side platform) and DSPs (demand-side platforms).
In a statement, Peter Crofut, VP of Business Development for Agencies and Brands at Wurl, categorized the CTV ad ecosystem as “ripe for disruption.”
“We know emotional resonance drives positive attention for brand campaigns, leading to higher brand awareness, recall and reduced cost per engagement for advertisers,” Crofut stated. “Now, with BrandDiscovery, advertisers will have greater control over the context of their CTV ads, more easily earning attention of viewers and, as a result, improving measured campaign outcomes.”
Wurl is using GenAI to analyze video content, as IRIS.TV, meanwhile, has pulled together contextual data companies, along with publisher content and ad serving platform partners, to offer a centralized contextual CTV PMP deal library, which launched last month.
The IRIS-enabled CTV deal library is meant to make contextual ad targeting and buying easier by offering nearly 100 curated programmatic PMP (Private Marketplace) deal segments that are available in a self-serve environment.
In an interview with StreamTV Insider, IRIS.TV co-founder and COO Richie Hyden explained how, essentially, IRIS.TV has built a content spine (or graph) - likening it to an identity spine used for audiences, except containing data for content instead of people.
And in offering contextual CTV deal segments, the company isn’t making guesses based on broad or more general data inputs sometimes seen in the bid stream – such as the name of an app, or genre, that’s often based on publisher or SSP data. That type of data, while not purposefully incomplete, is not standardized and is error prone, according to Hyden, and has issues with scalability when it comes to applying across a vast and fragmented CTV and advertising ecosystem.
IRIS.TV, through its content graph, creates a standardized approach to programming data using computer vision, where advertisers can target against content using its platform of pre-built contextual deal segments in a marketplace that allows buyers to more simply transact deals based on video-level data that’s been analyzed frame-by-frame.
“If we want CTV to be as healthy a format as we have seen with search, social and display, the content that someone’s consuming needs to have an as equal seat at the table as the person,” he said, adding, if not more so in the case of CTV when there’s often co-viewing and it’s not always clear who’s watching.
The IRIS.TV CTV deal library allows advertisers to contextually target millions of viewers watching shows and movies across thousands of premium streaming apps and FAST channels. It’s currently available through a handful of SSP partners, with plans to expand the SSP footprint a focus for 2024.
A component for IRIS.TV is that it’s compiling video-level data with files directly from content providers.
Comparing IRIS.TV to others in the market, Hyden categorized existing products as providing “a good guess” when it comes to program data, where a company may see an app like ESPN, for example, and categorize the content for advertisers as “sports.” Whereas under the hood, IRIS sees a lot of variety among content within a single app like ESPN, which requires actual video files from content providers to provide a more fine-tuned approach to define the programming “on a content level, not an app level or a channel level.”
Specifically, the CTV deal library includes segments for over 40 IAB contextual categories, GARM Brand Suitability segments for brand safe content, emotional resonance (with over 25 segments matching emotions like surprise, anticipation, joy, etc), audience lookalikes, and holiday and tentpole event programming with seasonal segments. It also offers custom deal segments, where advertisers can target with more nuance – say if a brand wanted to target an ad to run against content that is more appropriate for commercials featuring alcohol products.
Creating a deal library for contextual-based CTV targeting
To be clear, IRIS.TV isn’t a contextual data company itself, but rather has built a platform that makes utilizing the approach possible and easier.
IRIS.TV serves as the central point among three legs of partners: publishers and content owners in the CTV space, contextual data partners, and ad server platforms. Content supply or publisher partners send IRIS their content files including shows or livestreams, with video files then sent to data partners. On the data side, IRIS.TV counts Gumgum, Oracle, KERV, Pixability – and previously mentioned BrandDiscovery from Wurl – alongside several others, among its 17 data partners. Once data companies get video files from IRIS.TV, they use their respective proprietary AI and computer vision technologies to analyze video-level data from content frame-by-frame, including imagery, audio and text data. Based on that analysis they create the categories or contextual segments for use in the CTV deal library. All videos within the library have an IRIS_ID content ID attached, with different segments or options from the video-level data partners listed below it.
Different data partners bring different elements to the table and IRIS.TV doesn’t believe everything needs to be unified under one taxonomy, as that takes away “all the special sauce of each data company,” Hyden emphasized, but rather thinks the ecosystem needs a graph to hold them all together. With lots of data crossing the advertising supply chain he cited a desire to help partners avoid a “where’s Waldo” scenario of where to find it. The third leg is IRIS.TV’s ad platform partners. The aim is to make contextual program datasets available through ad server SSPs and DSPs, so they can be used for targeting and reporting in a similar way an audience data marketplace would be made available through The Trade Desk or another activation platform.
In creating the deal library, Hyden described a push towards curated or self-service ad buying capabilities and making it simple to create deals using supply-side data signals, which he said can help contextual and similar targeting signals get adopted across the market at scale.
Ultimately, IRIS.TV wants to make it as easy as a click of a button for contextual CTV ad buying and targeting. For example, a buyer could log in to the platform and say they want to target people watching sports content.
“Just click, grab the deal, it will push the deal to their seat and that’ll basically give [the advertiser] inventory across the CTV landscape which is only sports content, based on the process that we run through to create that data,” he said.
Contextual for broad or nuanced CTV targeting
When it comes to targeting contextually on CTV, Hyden said advertiser aims run the gamut, but the method can be used by essentially any brand category, albeit in different ways.
Some advertisers are looking for broad awareness. Such as CPG brands selling toothpaste, for example, where everyone is addressable, but brands might want to target consumers watching certain content like travel, where they’ve seen specific results.
It “can be as simple as just big broad awareness of a category, or it can be as nuanced as let’s take one creative and put it in the right moment in time to drive app downloads, or CPA like foot traffic for a QSR restaurant,” he said.
Brand safety is a separate area, where the operating chief said CTV buyers are less concerned with other contextual segments but could choose to focus on GARM segments because they “just want to know that everything in there is low risk.”
News, meanwhile, is a popular viewing category on CTV but some advertisers will avoid buying against news content altogether on so as not to be aligned with negative emotions or coverage. However, as Hyden noted, a lot of news channels and apps run feel-good programming and other non-news content that could be a fit for advertisers, and where video-level data and contextual segments could let them target appropriate content within news apps or channels more precisely.
“We do see very commonly that people want a brand-safe news deal,” he said.
IRIS..TV has also seen traction in the CTV deal library with performance buys, particularly around the emotional segments it can create.
“What we’re seeing is that if you take a ‘happy’ ad, you put it in a ‘happy’ moment in time, it actually works really well, resonates,” Hyden commented, adding it has seen higher rates of aided and unaided awareness, high brand recall, and even bottom funnel sales lifts.
Separately, Wurl also claimed success of its BrandDiscovery product with customer Media.Monks, for a campaign that was activated through the Comcast-owned FreeWheel technology platform. In Tuesday’s announcement, Matthew Kramer, head of Brand Investment at Media.Monks said the campaign for a financial services client using BrandDiscovery’s emotion targeting led to a 33% lift in aided brand awareness and a 15% lift in purchase intent, outpacing performance of all other CTV campaign strategies.
“What’s even more promising, the campaign saw a 200% improvement in cost per engagement – as measured by EDO – when the ad creative’s emotions matched the content right before the ad break,” Kramer stated.