• by Laurie Sullivan  @lauriesullivan, April 21, 2017

Google is using machine learning to make its display ad network smarter. It now can automatically target and bid on specific advertising units on publisher sites, pulling from pre-existing copy and content that advertisers pre-set to use.

The Smart display tool gives marketers a way to personalize ads served to consumers, based on machine-learning technology and the data Google collects across its network of sites, such as searches and clicks, to improve the relevancy of the ad served. 

Although Google does not specifically point to using data from entries in calendars or keywords in Google Home and Gmail, the platform learns from online and offline behaviors. They could also include travel, by pulling data from Google Maps for each user signed into the network on their mobile device.

For Smart display campaigns, advertisers add pre-set headlines, descriptions, logos and images to create responsive text, display and native ads, so the platform can pull the ads and display them in real time. The system determines the likelihood of the conversion based on the predetermined price for the bid and target CPAs, for each auction.

Trivago, Hulu Japan and Credit Karma tested the new campaign type.

For example, the hotel travel site Trivago used Smart display to help travelers find rooms. Marketers created headlines such as "Find Great Hotel Deals," and added images, hotel listing, and its logo, along with a target CPA and daily budget.

AdWords created more than 25,000 custom ads and served them to travelers shopping for hotel deals. Travelers searching a travel blog might see a message with Trivago’s "Find Great Hotel Deals" headline and an image of the Coliseum in Italy.

During the test phase, Trivago drove 36% more conversions at the same CPA, compared with its other similar display campaigns, David Margines, product manager at Google AdWords, wrote in a post.