Google Display Network goes AI

* This article originally appeared on smk.co – as one of our resident guest blog posts.

In September, I was fortunate to have been invited to the Google APAC Display Network Summit held in Google’s new Singapore head office.

The Summit was the first time Google has ever held a Summit focussing entirely on the Display Network. The announcement that really stood out was Google’s new offering of Machine Learning applied to Display Network: the Smart Display Campaign.

Smart Display has been in Beta for quite some time, but what struck me here was the serious weight that Google have put behind this. Essentially Google have gone ‘all in’ for this to be the AI driven “one campaign to rule them all”.

Let’s explore what Smart Display means for advertisers and agencies.

 

Think Assets not ads

The first impact is that Smart Display will no longer show ads per se. Instead we supply assets – headlines, descriptions, images – and the campaign will then combine all of these in multiple combinations turning 15 assets into more than 500 ad variations, then use Machine Learning to serve the combinations which have the highest propensity to convert.

 

Consolidate instead of Segment

In quite the reversal, Smart Display requires us not to create and target multitudes of granular segments, but instead to consolidate and reach as *large* an audience as possible then let Machine Learning determine who, within that audience, are the best candidates to reach. In doing so, Smart Display automatically factors in Remarketing, DSK, In Market, Custom In Market, Custom Affinity and Affinity audiences (as well as thousands more signals) for us.

 

Multiply Conversion Signals

Without doubt the biggest obstacle for most advertisers will be supplying Smart Display with enough conversions. If advertisers can achieve around 100 conversions per month on Display normally, they can transition to Smart Display and expect little variance in conversions & CPA and also do so within a short 2 week learning window. Fewer conversions than this will mean more unpredictability in results and a longer learning window. Smart advertisers will go one step further and feedback Sales/CRM data via Google Analytics Measurement Protocol to provide the highest value conversion signals back to the algorithm.
As Google continues building momentum under CEO Sundar Pichai’s ‘AI First’ philosophy, we will continue to see Machine Learning creep into more and more of our tools. Good advertisers will use these, but great advertisers will know the signals and conditions that these algorithms require in order to get the best out of them.

No Comments

Leave a Reply