How AI is Evolving Ad Creative
The creative process is notoriously challenging for advertisers and publishers. Beyond the difficulty of the creative talent that goes into the ad, publishers and advertisers struggle to settle on creative and match it to viewers at appropriate times. It’s a huge logistical challenge.
To clarify the challenges of creative management and how AI is addressing them, Street Fight connected with Jake Moskowitz, head of the Emodo Institute.
What are some of the pain points advertisers and publishers encounter with the creative process?
One issue from the marketer side is the need for cooperation between creative and media in order to leverage many of the best innovations in digital creative. Dynamic Creative Optimization (DCO), for example, requires many different versions of different components within the ad unit to enable the best results, but those components may be coming from a different source than the media decisions. Augmented reality is another example, as an upfront investment is necessary (although smaller than many assume) to create the immersive assets to maximize the opportunity, and media agencies may not be in a position to do that themselves. The days of handing over baked creative assets to a media team to make independent placement decisions may be changing.
From the publisher perspective, some of the most compelling and valuable ad formats require partner integrations, such as Native, and Embedded AR (an emerging version of augmented reality in which the immersive experience lives within the ad unit itself). This comes at a time when publishers are understandably trying to streamline integrations. Savvy publishers can seize this opportunity to test new integrations to enable these emerging formats.
How is AI being used to transform the development and/or management of ad creative?
There’s an important distinction there. AI can be used in the development of creative, for instance, by using natural language processing (NLP) and natural language generation (NLG) to create ad copy automatically.
But we’re more focused on AI being used in the management of ad creative. Dynamic Creative Optimization (DCO) is a perfect example, in which AI can optimize combinations of components within an ad unit to ensure the right combinations are delivered to each bid request. While DCO may not seem new, this AI approach is far more sophisticated than the more limited traditional A/B testing.
Advertisers are also already using AI to identify and segment audiences, filter out lower quality inventory, and optimize spend—all automatically, in real-time, at scale.
How will these AI-driven changes better serve advertisers, publishers, and consumers?
One of the real benefits of AI is that it is potentially a better fit for the privacy-forward path that digital advertising is on. It seems decreasingly sustainable to rely on a strategy of building the largest possible database of deterministic data for targeting and measurement. That doesn’t mean first party data isn’t valuable; quite the opposite! But the greatest value of first-party data is its application as a training data set because first party data can’t scale enough to replace opt-out cookies and ad IDs.
We as marketers need to stop focusing on quantity and start focusing on quality. That’s where AI comes in. High-quality training data sets are the key ingredient to successful AI, and that doesn’t require the level of scale needed for a deterministic targeting database, for example. That helps all parts of the ecosystem. Publishers maximize the usefulness and value of their first-party data, even at modest opt-in rates. Advertisers maintain the level of efficiency and scale to which they’ve become accustomed from digital advertising, even with addressability decreasing. Consumers have a more private journey, without being reminded at every click of their abandoned e-commerce cart from last week.
How do you expect the use of AI in creative to evolve in the next few years?
As the industry recognizes the value of data quality, I expect data consortiums to form in order to create training data sets that are thorough and diverse enough to drive great value. I expect algorithms to be more easily movable; we’ve seen the advent of BYOA (“bring your own algorithm”) in the DSP space, and that same sort of concept should exist elsewhere in the ecosystem as well.
As it becomes harder to build device history, I expect more value to be placed on real-time signal data in driving the specific creative a consumer sees. What time of day is it, what’s the weather, and what’s the context in which this consumer is seeing this ad?
I expect an increasing culture of test-and-learn that leverages AI to build many versions of creative in order to find the right combinations of components. I expect the humans to tap more into their creative brainpower to focus on ideas and themes, and less on the tactical part of getting ad units out the door.