Machine learning and advertising, what does the (not so distant) future hold?
2018 is just starting, and already we’re being reminded of the importance and soon to be the prevalence of Machine Learning (ML) & Artificial Intelligence (AI).
Given the advances of last year, should we be surprised? 2017 was a breakthrough year for AI even though it showed just a glimpse of what it can accomplish. A new 8-planet system was discovered using Google AI technology, processing an immense amount of data in order to just point the telescope in the right direction and guide it on its search. The data was processed, and the planets were found, so simple yet nearly impossible for humans to do on their own.
Remember in May when Alpha Go defeated the world’s number 1 Go player? For those who don’t know, Go is an incredibly complex Chinese strategy game, where a game can possibly last up to 2.08 x 10^270 moves (for reference, 10^82 is the estimated number of atoms in the universe).
As incredible as all of this may sound, remember that this is just the beginning, it’s raw data, and as marketers, we deal with one incredibly unpredictable variable: Humans. Can we predict human thought patterns to know what product to showcase and at what time?
Well.. yes!
Let’s remember the Psychological axiom that has gained traction in the past few decades, now coined by Dr. Phil that says “the best predictor of future behaviour, is relevant past behaviour.”
Just a few months ago, AI was used by Carnegie Mellon to defeat some of the top professional poker players in 120,000 hand game over a 20-day competition. Why is this important? The machine had to learn to compete in a game where bluffing is a core component. It had to use data to be able to win and bet through the seemingly unpredictable strategic component of lying. If that’s the case, then what’s not to say it can’t pick out a time when we randomly go down the advertising rabbit hole clicking on random links and products and be able to accurately tell that it’s not relevant information for our core consumer needs?
New AdWords features such as Custom Intent Audiences begin to tap into ML while being targeted at an audience or user. It utilises data from your existing campaigns to automatically define an audience segment based on the users recent and relevant browsing history. Think of it as a remarketing list, except instead of defining the audience list yourself, you had a computer automatically do it for you based on all of the same data and signals that you were going to scroll through manually. Not only is it saving you time ($), but it may also be providing you with new ideas to trial.
Universal App Campaigns, are another great example of ML in advertising (click here to see how Canny can help with advertising your app). These campaigns leverage ML and the huge amount of search data available for Google to perform in order to reach your intended audience. UACs are not only effective for achieving your cost per install goals but also begin to exceed them as well, allowing you to effortlessly scale for growth.
Eventually, we’ll get to a point where all advertising is digitalized and targeted to the specific user (Black Mirror anyone?). Not just online either, but also on TV, radio, billboards, store fronts, etc. When that time comes, advertisers won’t be looking at bids to manage, or new keywords to target, or what media to buy, we’ll be looking at the main audience drivers of results. Who’s converting? What are they looking at? What content can we create to better engage with a specific audience?
A results guided inductive analysis into strategy and creatives will be crucial in a ML dominated world. Be at the forefront of advertising, begin experimenting with some of these machine learning features that you may have available, and use ML for the complementary powerful tool it is. 2018 is shaping to be a big year for ML & AI, and there’s no better time to get started than the present.
Curious about how AI could impact other business areas? Check out this article on Tenfold, there's an endless amount of possibility in this field, and it's best to stay ahead of the curve!