How Video-Level Data And Content Identifiers Solve The Transparency Problem In CTV

Ad dollars are following viewers to Connected TV (CTV), but linear TV expectations are on a collision course with the reality of ad-supported streaming. AdWeek, in its coverage of recently published research by GumGum, highlights one of the major problems in streaming today - transparency. The research revealed that 20% of ad breaks in children’s content on CTV including free ad-supported TV (FAST) had illegal or problematic ad placements. While no one wants to see ads for alcohol in kids' content, CTV advertising is complicated.

The biggest problem is that planning, targeting, and measuring advertising in linear TV and digital video are fundamentally different.

Program-level information worked for linear TV for two reasons. The first is in linear TV, everyone sees the same broadcast and programming guide, which makes program-level information such as “show” and “episode” sufficient for transparency requirements in upfront terms and order letters. The other reason is that for more mature content, broadcasters have TV edits that remove obscenity, nudity, and excessive violence from content so there is only one version of the show.

Streaming and CTV is a personalized content and ad experience, just like digital formats including mobile, social, and YouTube. Digital advertising is personalized and manual processes like those used in TV don't apply. For example, instead of spreadsheets and manual reports, digital advertisers use technology providers for brand safety, brand suitability, and contextual targeting to understand how and where ads are placed.

In CTV, there is no equivalent of a publicly available page URL that contains all the necessary and sufficient information to analyze and segment. A buyer cannot use the currencies it relies on in digital and up until now has had to rely on self-declared metadata. Seller-defined content metadata is problematic because it is fragmented and not scalable for ad platforms or measurement partners to rely on. In an analysis by IRIS.TV and Basis, it was found that there were over 740 different ways that "comedy" was labeled in only 100,000 ad impressions. 

For example, “The Bad News Bears” and “The Sandlot” share the same content metadata - Rated PG, Comedy, Family, and Drama. One movie has a high amount of obscenity, racial and ethnic slurs and one is appropriate for all ages. 

While there have been calls for publishers to share “show” and “episode” data to solve the brand suitability problem, passing program-level information in the bid stream is a non-starter for streaming TV. While content owners are rightfully concerned about cherry-picking and data leakage, navigating regulations like VPPA and COPPA is extremely challenging. 

Even if VPPA allowed publishers to share program-level information, it would not be sufficient to determine brand safety and suitability, let alone in under 100 milliseconds. There’s too much subjectivity and inconsistency. It’s not about sharing more keywords. That just makes the problem worse. You have to watch the content to know what it’s about. 

While Chat GPT has captured our imagination, contextual intelligence solutions utilize machine learning and computer vision to analyze video frame-by-frame and create data segments for contextual and GARM brand suitability that provide superior precision and accuracy to content metadata. 

The barrier these currencies must overcome when bringing their technology from YouTube to the open web and CTV is accessing publishers’ video-level data. This is where companies like IRIS.TV and a content identifier like IRIS_ID fill this hole in the supply chain. 

IRIS.TV has solved the plumbing problem by developing interoperable technology that ingests and normalizes content data in any format. When a publisher becomes “IRIS-enabled”, each of their videos is assigned an IRIS_ID, and the normalized video-level data is passed on to data partners for enrichment. 

The segments are assigned to the video’s IRIS_ID which is then passed into the bid stream and decoded for targeting via direct, private marketplace, and open auction buying. 

To make it easier for buyers and sellers to transact, IRIS.TV has developed integrations with ad platforms and is collaborating with the IAB Tech Lab to establish standards and protocols.

The IRIS_ID is a signal and not a currency, so sellers and buyers can use it to activate whichever data partner they prefer for contextual and GARM brand-suitability. This is especially helpful during upfronts, as publishers can now provide buyers access to enriched video-level data for targeting and program-level information for post-campaign reporting in compliance with VPPA and similar regulations. 

While it is disconcerting to learn that ads for alcohol are playing on children’s content, it’s also a waste of money and a symptom of our overreliance on persons-based ad targeting. We are what we watch, which is why content-based targeting is a much better approach for streaming. As the industry transitions to ad-supported streaming TV, connecting and activating video-level data at scale will be critical to creating a transparent, brand-suitable, and ultimately valuable ad experiences.