As part of our A-Z of Adtech series, our Associate Audiences Manager, Juan Vásquez, speaks about the importance of keywords in programmatic advertising, and the work his team are doing to take contextual targeting to the next level!
What do we mean by keywords?
For as long as I can remember, keywords have been the basis of contextual targeting. At Unruly we work with Grapeshot which allows us to scrape keywords across the different sites that make up our network. We can then categorise them according to either our clients’ needs or our own proprietary taxonomy. This includes emotive, personality-based, cultural and motivational categories.
After classifying the active pages in our network, we can make them directly available in DealIDs so our clients can buy ad space in pages that are more likely to have content that matches their ads. This approach is called contextual targeting, as ad serving is based not on information about the user themselves, but on information about the context, they are in (e.g the webpage they are currently on).
Where are we now?
This approach to keywords has been instrumental in helping us scale across our network, in addition to the audiences we offer. But there is still a huge question we need to ask ourselves; can we do anything to bring together contextual and audience-led targeting? The short answer is yes!
We’re currently looking at the consumption patterns of individuals and how we’d target them based on that information. This means we would look at the last pages that the consumer had viewed, especially if they’re pages they’ve spent a considerable amount of time on (indicating that they’d been interested in the content). As before, we would analyse the page’s keywords to understand what it’s about. But instead of serving an ad based on that information, we would create a semantic profile of what the consumer had viewed in terms of content. Finally, by aggregating and anonymising them with those of other people, it would give us the ability to target based on this information.
When you think about it, it’s amazing! We don’t need to know anything about the consumer to be able to serve an ad that they will find much more relevant because it is, quite literally, what they are interested in! By using both the traditional keywords approach, as well as this complementary targeting capability, we can be more flexible and precise in our offering.
What does the future look like?
Keywords are great because they are easy to process and scale. But as technology evolves, Natural Language Processing (NLP) will become more commonplace, and this is exactly what we’re looking at. NLP refers to an algorithm that is able to read and understand a sentence in the same way a human would. Although they already exist and tend to have relatively good accuracy, they’re neither fast nor cheap and tend to be available mainly in English, which is a big issue when you have Unruly’s global footprint.
Going forward, I expect vendors will keep pushing for more precise algorithms that understand the nuances of language, making them fast enough to be reliable in a real-time programmatic landscape, and cheap enough for them not to eat at the margins of the ad tech ecosystem.