By Hailey Denenberg, VP of Strategic Initiatives at GumGum
As the CTV boom continues apace in 2023 – midyear forecasts suggest a 13% leap in streamer advertising to almost $26 billion – it’s clear marketers stand on the precipice of a golden kingdom: a rich content environment flooded with engaged consumers and exceptional ad completion rates.
However, the increasingly saturated CTV market is already giving way to growing ad fatigue, with AVOD subscribers turned off by irritating ads that erode purchase intent with every new repeat. Consumer attention – spread across multiple channels and devices – is becoming more fragmented than ever, driving the case for high-attentive ads.
The looming issue of brand safety is also ramping up demand for more innovative solutions. And one technology that is primed to play a key role here is AI-powered contextual intelligence. To date, advertisers in the CTV realm have had to rely on basic contextual tools that only stretch to property-level analysis – which involves analyzing the generic content metadata applied to a CTV video as a whole (e.g. name of the film and genre).
However, the next generation of machine-learning based contextual intelligence is able to look deeper at video content, analyzing it with frame-by-frame image recognition, as well as full audio transcription. Not only is this a much more effective way to understand the nature of a CTV video and the brand safety and suitability issues within it, it also unlocks a huge amount of untapped inventory beyond the ad break, carving out potential for enhanced monetization and a premium user experience.
The hidden brand safety threats within CTV
To understand why more basic contextual solutions aren’t fit for the CTV age, you have to consider the longform nature of most CTV content. The brand safety risks within content like this are constantly changing from scene to scene.
Take Netflix’s Stranger Things; a huge cultural hit around the world that many brands would want to show up in. But within each episode, there are moments that most marketers would also want to avoid altogether. Imagine a swimwear ad running immediately after Barb’s pool death in season 1. The current system based on property-level contextual analysis and generic metadata makes avoiding programmatic ad mismatches like this an impossible feat.
That’s where content-level contextual analysis comes in. AI-based contextual intelligence can dissect long-form CTV media on a granular basis, allowing scene-level metadata to be easily tagged to provide insights on any moment within a video and which ads will be most suitable. The automation of CTV advertising becomes a much safer and more viable prospect.
Opening up brand-suitable inventory
The very same mechanism opens up valuable inventory that might otherwise be blocked under the property-level contextual framework. For example, ordinarily, a popular PG-13 movie such as The Dark Knight risks being locked out of certain bid cycles – just because of its rating.
Intra-video, or scene-level metadata, however, can identify the nature of the few inappropriate scenes responsible for its PG-13 status. In doing so, it helps brands to capitalize on and align with other brand-safe, lucrative inventory beyond those moments.
The same logic applies to inventory within any number of box office or streaming hits – from Barbie to Doctor Who and Cruel Summer – that may contain some elements of violent or suggestive sexual content.
Scope for innovation
Alongside the potential to revolutionize brand suitability in CTV advertising, the ability to analyze content on a scene-by-scene basis can help raise the bar on innovative ad formats – elevating the overall user experience.
For example, rather than using a traditional ad break that pauses a CTV video, marketers could insert a subtle and non-disruptive overlay ad. This would sit at the bottom of the screen, and be contextually relevant to an individual scene.
Imagine the Friends episode The One With Chandler’s Work Laugh is passed into the bid stream. Intra-video metadata could be used to surface an overlay ad for a spirits brand over a bar scene; or an overlay ad for a fitness tracker over the scene of a tennis match.
Research shows that using CTV overlays in this way makes them four times more memorable than linear video ads, driving an increase in brand favorability and the perception by viewers that they’re less “bombarded” by ads (which makes sense when you consider the format’s non-intrusive premise).
Using AI-powered scene analysis, overlays can be scoped to resonate with a consumer’s active mindset by synching with their real-time content viewing. The format also sits comfortably with interactive elements, such as scannable QR codes, to ramp up conversions.
Shaping the future
Content-level analysis, then, is a powerful mechanism for brands seeking a deeper connection with CTV audiences, providing a less disruptive way to align with premium content, with greater monetizable inventory.
Advertisers therefore need to lean into this approach, rather than rely on property-level analysis – which arguably wasn’t good enough for the video world, and certainly isn’t good enough for CTV. Doing so will also help push the envelope on creativity, when it comes to formats such as overlays.
In the same breath, publishers and platforms need to up their game when it comes to tagging their videos with accurate metadata and passing this up the bidstream, ideally via appending a content_ID. Currently, even property-level data isn’t always shared, leaving advertisers and streamers exposed to unnecessary brand safety risks.
Key industry organizations like the IAB and MRC will be critical in moving the needle here; encouraging higher standards and accreditations so that content-level analysis and intra-video tagging becomes more widely embraced.
This will provide an edge for safety and innovation not just in CTV; but for the powerful new digital environments, e.g. gaming and the metaverse, that are shaping tomorrow’s ad market.
