How machine learning powers auctions in Google Ads

3 Aug 2018

Image: hxdbzxy/Shutterstock

17m data points are core to how ads are auctioned on Google’s Ads platform.

Machine-learning algorithms are determining how ads are bought and sold on Google’s Ads platform.

At a Virgin Media Digital Evolution conference in Belfast, we caught up with Simon Balfe, agency development manager at Google Marketing Solutions in Dublin.

Balfe explained how algorithms and machine learning are core to how the auctioning process works.

“What we have done is, we apply our machine learning to our bidding platform so you can actually say to Google, I can afford to spend £5 for every lead or £7 for every sale that comes through.

“And, what Google’s machine learning can do is, look at 17m signals every time somebody types in a search query and will use all that detail to see how likely somebody is to take that action.”

He explained that those signals include time of day, the device being used and “previous behaviour to figure out how likely they are to take that action and bid down or bid up accordingly”.

The present and future of Google Ads

Google recently retired the AdWords and DoubleClick brands and is replacing them with three new primary brands with machine learning at their core.

Under the product reshuffle – nothing is changing under the hood, apparently – Google AdWords will now be known as Google Ads; DoubleClick and Google Analytics 360 Suite will be known as the Google Marketing Platform; and DoubleClick for Publishers and DoubleClick Ad Exchange will be integrated under Google Ad Manager.

At the Virgin Media event, Balfe said that every 10 or so years, the tech world experiences massive paradigm shifts. In the 1990s it was web, in the 2000s it was mobile and, in this present decade, it is AI and machine learning.

“Google’s CEO, Sundar Pichai, recently said Google is now an AI-first company. AI and machine learning is enabling us to achieve 40pc savings on energy used for cooling in all of the Google data centres. We use machine learning to understand the weather outside and the temperatures inside, and know when to open and close the blinds.”

John Kennedy is a journalist who served as editor of Silicon Republic for 17 years