Advanced machine-learning tech company AdTheorent on Tuesday officially will launch a platform that relies on models to score and create unique advertising units in milliseconds.
The Advanced Predictive Creative (APC) platform creates unique content by combining product shots, logos, colors, backgrounds, messaging and calls to action to build ads that have the highest propensity to lead consumers to take action such as make a purchase.
AdTheorent began working on APC about a year ago. When completed, a major automotive manufacturer tested the platform, explains Matthew Groner, SVP of product management and business intelligence at AdTheorent.
The awareness campaign running through APC introduced a new hybrid vehicle on mobile devices, in apps and on the web. It was intended to drive people from the interactive ad unit to the web site to learn more, using creative copy and images to reach adults between the ages of 40 and 50 in market to purchase a car.
AdTheorent’s in-house creative division, Studio A/T, worked closely with the brand to identify and design various creative components, deploying unique combinations including vehicle images, CTAs and different copy options.
“We feed hundreds of data points into the platform and some of it depends on what the clients want to have such as their own CRM data,” Groner said. “We bring everything we know about the device, ID, location, time of day, and other information from publicly available data.”
The technology also processes data from past activities. It all aims to determine the best combination of text, image, colors and more to create and serve an ad. For example, the platform looks at whether people respond better with one image versus another. It uses the data rather than predetermined rules based on assumptions from humans.
Groner said the company is also working with a major retailer engaged in rules-based retargeting -- which involves following the consumer around the internet until they make a purchase -- helping the company to overcome a lack of consistency between the product test and the data.