In association with Silverbullet.
In this exclusive article written for ExchangeWire, Charlie Pinkney, business development director at Silverbullet and 4D, discusses how sophisticated computer vision and metadata analysis can be used to unlock contextual video in the post-cookie ecosystem.
The clock is ticking. The demise of the cookie is fast approaching, and many businesses are trying to determine how best to proceed. Whilst some want to sit on their laurels and wait to see what happens, others are looking to embrace a world where targeting audiences can be executed in a privacy-first manner, reaching the right person, in the right context, at the right time.
How? It’s all in the long-standing methodology of contextual targeting. In its early years, contextual targeting focused on identifying relevant keywords, which in turn helped marketers determine text-based editorial environments where consumers were investing time in. But the world has moved on since then. We are a generation of consumers who absorb content across a huge variety of channels; written, video, audio and more. The holy grail for many modern marketers is video. With the global digital video advertising market predicted to reach a whopping USD$155.18bn (£127.10bn) in 2026, you can understand why.
What it means to truly understand relevant video, including YouTube
Undertaking contextual targeting for text-based content is pretty simple. You identify the keywords and phrases, discard unnecessary content, run it through NLP, and voila. (Of course, it isn’t quite that simple, and requires a little more expertise to make it work, but you get my drift).
However, when we look to deploy the same mechanisms across video content, things can get a little more complex. In short, it isn’t that easy to extrapolate "context" from a piece of video content, as audio transcription and keyword/metadata aren’t "enough". Many have tried to adopt this as a way to understand what the video content is about, however oftentimes this just doesn’t cut the mustard, with the specific content of the video not being properly assessed. Let's dig a little deeper:
One method of analysing context in video content is to derive context from the metadata that surrounds a video. This metadata will usually include the URL in which the video is embedded, the text included on the webpage, tags, categories, and other metadata provided by the video creator. The problem is, this data isn’t always accurate or reliable, especially on user-generated content platforms, and can often be very vague. It is also just a proxy for the context within the video content itself, meaning aspects and nuance will likely be missed.
Advanced solutions such as 4D, are looking at video from programmatic sources and YouTube, from a brand-new lens. Video is more than just metadata; videos are images, stories, branding, audio, and colour, and so needs a whole new approach in order to truly understand the content, and therefore context. Methods such as content recognition are applied:
Exploring the video content itself can help to determine the relevant contextual categories. Content recognition can be divided into two separate approaches, which when combined creates an even more accurate picture.
First, machines can translate audio into text, and then analyse this text to highlight contextual signals. Audio analysis can be very useful and help add flavour to the video narrative.
Secondly, computer vision can be used to analyse images within a video to give an overall picture of what a video is about. This can create a very granular picture of people, locations, items and logos all contained within a video.
Computer vision systems such as 4D can be used to analyse keyframes within each individual video, understanding every image, sound and visual at a granular level, so it can bring all the relevant insights to best target against- across programmatic OLV, CTV, and YouTube. These advancements in contextual intelligence are really exciting for marketers adopting the solution for video.
Untapping video in the post-cookie era
Advertisers hoping to maximise the effectiveness of their ad spend both today and tomorrow, should seriously consider shifting their budgets to video - and even more granular, contextual video. Even though video-level contextual deployed seamlessly against programmatic video and YouTube may not be commonplace just yet, advertisers should start to get comfortable in designing creative for the medium, and to begin taking advantage of its high engagement rates.
This does not mean that contextual ads will replace behavioural ones. Far from it, as first-party data remains the holy grail for marketers stepping into the privacy-first future. However, context will become another data-driven proponent of advertising, helping brands reach consumers with messages they truly want to see — at the precise moments they're most likely to want to see them.
As the industry prepares for the third-party cookie depreciation, it won't be long until the vast majority of video advertising becomes contextual, so why not take advantage of the advancements now?