Google introduces fully-automated ‘Smart’ display campaigns
The new campaign type applies machine learning to all aspects of campaign optimization.
Automation isn’t new to display campaigns on the Google Display Network, but with the new Smart display campaigns, it’s not just the creative that’s automated. Targeting, bidding and the ads all run on autopilot powered by machine learning.
The system pulls advertiser-provided headlines, descriptions, logos and images to create responsive text, display and native ads. Bids are set, based on Target CPAs, for each auction as the system determines the likelihood of conversion.
Trivago, Hulu Japan and Credit Karma were among the beta testers for this new campaign type.
From the Help Center page:
Smart display campaigns combine your assets — headlines, images, and so on — to create image and text ads. They can even transform into native ads, blending seamlessly into the fonts and feel of publishers’ sites. Over time, your ads optimize based on what wins you conversions.
For Smart display campaigns to work, advertisers need to:
- Have conversion tracking set up and receive at least 50 conversions on the Display Network, or at least 100 conversions from Search ads, in the past 30 days.
- Use target CPA bidding.
- Set daily budgets to account for at least 10–15x the target CPA bid.
- Apply any site exclusions (account-level exclusions apply to these campaigns).
Advertisers can also run dynamic retargeting through Smart display campaigns by attaching a feed to the campaign.
If you’re looking for precise creative control, this option is not for you. The system can generate thousands of ad permutations, though, and you can see example ads during the setup.
To set up a Smart display campaign, choose one of the first four objectives under “Drive Action” in the marketing objectives screen when you start to create a new display campaign. Then choose Use Smart display campaign.
To analyze performance, the table view of asset details under the Ads tab will show data for each asset (copy, images and so on), including a “best,” “good” or “low” grade based on its ability to generate conversions — note the asset grades are relative to each other. You can get rid of “low” grade assets and replace them.