On Halloween Eve I stepped into the Los Angeles Facebook offices for an agency event called “AI in Action” a training focused on how using Artificial Intelligence (AI) within the Facebook platform can help enhance the overall performance of campaigns. Their office was highly secure (three checkpoints with a government ID to get in) but once inside I was greeted with friendly faces, loud toe-tapping music, kitschy Halloween decorations, and a three-story industrial building decked out with modern art.
Facebook began the session by defining AI and how it uses machine learning on their platform with an analogy. They said an average pilot spends seven minutes manually piloting their plane in a typical flight. A majority of their time is spent entering coordinates into their computers, checking on mechanisms that keep the airplane safe, and filling about information about the flight itself. The rest of the flight is all powered by AI. Then the Facebook presenter asked for a show of hands, “how many people would feel safe flying in an airplane without a pilot even though most of it is AI”? A few hands raised, but the rest of the room laughed. “Exactly. Just like many clients don’t feel safe with their AI campaigns without your ad buyers.”
They then presented an eMarketer stat that by 2021 an estimated 88% ($81 billion) of all US digital display ad dollars will transact programmatically. AI and machine learning is becoming an important part of digital buys because while people can forecast data in the past and make predictions in the future, ad buying is getting better and better at making decisions in the present, in real-time, in a way a human never could. But the key to making AI work for your campaigns is “liquidity”.
Liquidity on Facebook means the ability to run a campaign across every placement possible so AI can decide where conversions are happening the best. Placement liquidity refers “to a situation in which campaign budget can flow to the most valuable opportunities, regardless of placement.” This is exactly how our Facebook and Instagram Premium products work and its value in letting impressions flow from the Newsfeed to Messenger, to the Facebook Audience Network, to Instant Articles, over to Instagram, and into Stories. Many of our advertisers are already experienced at explaining placement liquidity and why it’s so important, but a big opportunity missed is bouncing between Facebook and Instagram and all the placements both platforms offer within the same campaign. If I had to guess only 5%-10% of Vici social campaigns use both at the same time.
AI in Action also made me look at Custom Audiences and Lookalike Audiences differently. A Custom Audience is when you upload names, zips, emails or phone numbers into the Facebook platform, and you are provided a match to those profiles. When you employ a Lookalike to this audience, Facebook typically explains it as finding people who look just like your customer’s demographic profiles. But with AI it’s so much more than that and this is what Facebook is enhancing.
It’s not just finding other men the same age and income in your area to sell stuff to, it’s finding other people who also convert in the way your target customer is converting even if they aren’t in your demographic profile. So, you could have a grandmother that is converting at the same rate as your men and is it opted into the campaign with AI because it improved the overall conversion performance. And that is so much more beneficial then a Lookalike because you find people to fold in your customer mix that you would have never otherwise targeted. Facebook also teased a product called “Value Base Lookalike Audience” which is going after people that are spending higher than your average Lookalike when defining a dollar amount per conversion within your campaign.
Another topic Facebook spent a good amount of time on is understanding the importance of a human touch in the world of AI and how AI works better with better creative and placement liquidity, but placement liquidity and AI doesn’t work at all without strong creative. That’s where the human creative element comes along.
Carousel ads are still among the top performing ad placements on Facebook, but the rise of homemade videos without huge budgets is emerging as the biggest trend. When making these homemade creatives (creatives done without big budgets and studio work) Facebook says the key is to focus on displaying the brands intention within the first 3 seconds of the video, designing the ad for no sound, and focusing on shorter format duration like :15 or :30s. They also recommended many “mobile studio” apps, or apps that focus on shooting video on your phone, to take the most basic shots and make them wow, which in turn makes AI work even better. Their top recommendations for shooting on your phone are Videoshop, Photoshop Mix, Ripl, Unfold, Plotaverse, Adobe Express, and Boomerang.
Much of the event was focused on helping agencies understand that AI has been a big focus for Facebook and powers many of the tools we have access to, with the intent on taking machine learning a step farther. They talked about a newer product “advanced matching” which is coding developers can add to a client’s website to allow you to send hashed data back to Facebook along with pixel data so you can find people who may have saw your ad on Instagram, didn’t click but went to the website on their own and converted, to give the advertiser credit. They also spoke about attribution studies for real time data on a campaigns and lift testing to understand incrementally what creative is driving the campaign the best. More on that after Vici does some beta testing in 2020.
At the end of the session, they handed us all fortune cookies, because what digital marketer doesn’t love cookies?! We broke them open to reveal a message, with three at the event winning a new Facebook Portal. While I wasn’t a winner (sorry not sorry Facebook, I’m an Amazon Alexa girl through and through) the message revealed a non-inspiring non-message of the future “Thank you for coming to AI in Action, visit us for a recap of best practices you heard about here today”. Bummer! I was expecting some sort of Terminator-esque doomsday joke or optimistic vision of the future. But all and all I feel confident that digital advertising’s future is bright with the innovations and power of human and machine learning.
