Custom AI has become an indispensable tool for agencies seeking a competitive edge in the rapidly evolving digital marketing landscape.
Digital advertising has shifted from predefined audience targeting to adopting more sophisticated, custom AI-driven methods. Initially, brands relied on predefined audiences for user targeting, a necessary compromise given the technological limitations of the time. However, this approach often sacrificed accuracy for simplicity.
Lookalike modeling represented a significant leap forward, enabling brands to expand their target audiences by identifying users with characteristics similar to their specific brand audience. This technique became a staple in the toolkits of major platforms like Facebook and Google.
The latest advancement in this evolution is fully customized targeting designed for the cookieless web.
This approach employs custom AI to build campaign-specific machine-learning models using first-party data and contextual signals. These models analyze URLs, scoring them based on their semantic relevance to a brand’s campaign brief. The result is a refined selection of content that aligns closely with the campaign’s objectives, surpassing the accuracy of standard segments.
A critical aspect of audience targeting with custom AI is the quality of the underlying audience data and the integrity of the matching process. A study by Truthset highlighted the reliability issues in data used for ad targeting and audience measurement. The study found that matches between hashed email addresses and postal addresses across various data providers were accurate only about 51% of the time, casting doubt on the accuracy of such audience data matches.
Several innovations underpin custom AI’s data integrity and advanced targeting capability. For example, network-level analysis (NLA) is crucial, examining the entire universe of URLs to discern content clusters, trends and semantic relationships. Content retrieval techniques scan this network, identifying URLs that align with the advertiser’s brief. A custom AI model, built and trained with this filtered content set, classifies new articles and ensures that only the most relevant ones are selected for the campaign.
The efficacy of custom contextual AI is evident in its results. For instance, Seedtag’s Affinity Index, which measures context relevancy for the intended audience and message, is typically 92% higher than scores derived from predefined taxonomies. Moreover, ads placed using this technology enhance ad/content fit by 9%, leading to significant uplifts in ad recall (22%) and message association (19%) compared to standard IAB categories.
In a post-cookie landscape, audience targeting will increasingly rely on first-party data. However, translating this limited data into scalable marketing campaigns poses a significant challenge.
Contextual targeting, focusing on the environment of the ad placement rather than gathering information from potentially unreliable audience data, ensures relevance to the content being consumed at the moment. This approach bypasses the uncertainties of personal data matching, offering a powerful and sustainable alternative to traditional methods.
Custom contextual advertising, therefore, emerges as a future-proof solution in a privacy-first world. It adapts to the evolving digital landscape and outperforms standardized segments, offering a more accurate and reliable method for placing ads in relevant contexts.
As the digital advertising industry grapples with signal loss and heightened privacy standards, custom contextual AI stands as a beacon of innovation, guiding the way to more effective, responsible and sustainable advertising practices.
By Chad Schulte, Senior Vice President of Agency Partnerships and Strategy at Seedtag.
Interested to know more about contextual advertising?
This article was originally released on Digiday