Machine learning enables dynamic selection of ad formats, creative, and text with each impression.
Originally published by, Taylor Peterson
Facebook is bringing more machine learning-driven capabilities to its ads platform to improve personalization, the company announced earlier this month.
Dynamic formats and ad creative for Dynamic Ads. Advertisers will now be able to automatically serve different ad formats to audiences based on the machine learning model’s prediction of a user’s format preference. Campaign managers can access this capability from the Facebook Dynamic Ad unit when creating ads for the catalog sales, traffic and conversions objectives in Ads Manager or the API.
“The dynamic formats and ad creative solution aims to meet people where they are in the customer journey by delivering a personalized version of the ad to everyone who sees it,” Facebook said.
Multiple text optimization in single-media ads. As we reported last month, Facebook has also rolled out a responsive ad feature that allows advertisers to input multiple text options for the primary text, headline and description fields when building single-media ads for traffic, app installs, and conversions objectives. Facebook said it uses this data to optimize for delivery and performance using variations of the text options provided, based on individual preferences identified by its machine learning models.
Auto-translated languages for single-media ads. In Ads Manager, advertisers can now add different languages to be auto-translated for international audiences. By automatically producing translations for key languages, this feature can help speed up the campaign setup process while still giving advertisers control over the review process.
Why we should care. For smaller organizations, creating highly personalized content can be challenging if time and resources are limited. By leveraging machine learning to dynamically select ad formats, creative, and text on the ad impression-level, advertisers can more efficiently deliver campaigns that reach the right audience, at the right time, with less effort.
More on the news. Facebook machine learning blends data from its platform with target audience insights in order to predict which people might be most receptive to a brand’s message. As people take different actions on and off Facebook, the company explained, intent signals are created to help achieve a more one-to-one experience for consumers served with both organic and paid content. Facebook said it’s committed to its investment in prediction models that can help build stronger business outcomes and better digital experiences for customers.
In a test of 12 e-commerce and retail advertisers, Facebook reported that dynamic formats and dynamic ad creative outperformed carousel-only ads in driving lift across content views, add-to-cart, purchases, and sales. Facebook said the showed an average of 34% improvement in incremental ROAS, 10% improvement in lift and 6% lower cost per incremental purchase, compared to carousel-only ads.