Google Leans On Machine Learning And Scale For Smarter Display Ads

Google rolled out a machine learning-powered display product called Smart display campaigns on Thursday. The product is now generally available to all advertisers buying native, image or text ads across the Google Display Network (GDN).

Smart display campaigns are accessible via AdWords and reach more than 3 million GDN sites and apps on GDN now.

Beta testers like hotel search platform trivago have seen conversions increase an average of 20% across the board compared to standard display campaigns priced at the same CPA.

In its Smart display campaigns, which increased conversions 36%, trivago uploaded creative targeting different demos, set its price, budget and bid goals, and then Smart Display automatically generated 25,000 custom ads catering to the needs of different consumer targets.

Google claims Smart display uses machine learning to improve ad decisioning. While that capability isn’t new, Google argues it differentiates with pure scale.

“The opportunity here is to personalize ads based on what a user has previously done and deliver them in real time,” said Brad Bender, VP of product management for Google. “We update AdWords audiences in real time so when a consumer hits mute on an ad, for instance, we learn what they don’t like automatically and feed that into our machine learning.”

Within GDN, millions of signals make up each targeting and bid decision, so machine learning is vital for Google to wrangle it all. 

Although Google leveraged large-scale machine learning for certain products like Automated Insights, that was largely limited to analytics.

“We heard from advertisers that they wanted it to be easier to sort through multiple targeting options, create multiple versions of their ad or do complex calculations to figure out what right bid to set,” Bender said. “Over time, we’d developed a number of tools to make this easier, like auto-adapting a creative to fit every screen size or applied machine learning to automate bidding.”

But Google’s goal is to bring these media and creative capabilities together in a single workflow while ramping up its use of machine learning in ads.

“Machine learning and artificial intelligence are really at the heart of what we’re doing at Google, which you’re seeing play out in analytics, Google Assistant, Maps, and it’s certainly relevant for ads, too,” Bender said.

Google said that AI concepts like machine learning are part of its entire organization’s ethos.

Google AI initiatives include projects like DeepMind, Alpha Go and A.I. Experiments, its open-source library for machine-learning applications.

“If there’s a model that gets built in another part of the business, we’re able to bring it in and see how it does in our organization since we have a back end that allows us to test different models,” Bender said.

Easier said than done, but from an advertiser’s perspective, Google's goal is to strike a balance between  simplicity and performance.

“They just log in, upload their assets and set their campaign parameters,” Bender said, “and we make the determinations to help them get to the outcomes they want.”

 

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>