It seems like the world is getting smarter around us. Technologies seem to know what we are going to do before we do it. Our phones know what restaurants are nearby. Our computers know just what we have been thinking about buying. Even our TVs and Kindles know what we want next to entertain us.
In this digital era of revolution, where technology is effectively learning our interests, behaviours, and preferences, an important question is raised: how will humans in the advertising profession complement AI if the advertising machines are like marketing fortune tellers?
Before we answer the question, we have to think about what’s been happening over the years in the advertising world.
It’s been a dream of marketers – certainly since my father’s generation – to extend what was direct marketing to be a two-way, one-to-one conversation with consumers. We have always known that it would be computers that would do this. What we didn’t know was the enormous computing power needed to make it a reality.
This new data-driven marketing world is a world of milliseconds. For an advert to be changed for each consumer and customized to their product interests, there needs to be a transaction to buy that ad, just that one ad, just for that consumer. To do that, we needed to completely change the way we bought and sold ads — not by the thousand to blind masses but to individuals, one-by-one.
We live in a world where, in less than 100 milliseconds, a single ad is put up for sale, the suitability of the potential customer and the value of advertising is assessed by the marketer, a bid is made, and an ad creative is created or chosen, then delivered. You can now understand why it has taken us generations to get the processing power and internet speeds necessary to make this even possible.
Needing machines isn’t new to advertisers
Some of the well-known American credit card companies have been pioneers in direct marketing for decades. They first adapted media such as direct mailers sent to houses by snail-mail, and even then, used computers to help them with sorting and targeting. Then came email marketing with data. The problem with both was not only the cost, but that the marketer was probably adding more unwanted media for the consumer to read.
Looking back, needing machines or data to do their jobs is not new to marketers.
So, what’s really changed now? The thing that has changed are the data inputs. In the past, demographics, geography, and purchase behaviour were the central inputs. This remains true today, however, the media is also now feeding back data. Not by some old-fashioned mail-in reply cards or months-later consumer surveys. The media feeds back data in real-time, driven by millions of bits of data from social and other digital media.
So marketers can concentrate on one-way mass media, like radio, print, and terrestrial TV. Or they can adapt to two-way digital media including digital audio, online display, and connected TV. Consumer behavior is moving away from the older mass media sources. Life is too short to wade through a newspaper or wait a week for the next episode of a TV show. The audience is moving and so should marketers.
Human skills can’t be replaced
Regardless of the changing landscape, the marketing strategies and dollar spending decisions continue to rest with humans.
I was lucky enough to work with machine learning early on. I observed that a machine needs time and lots of examples to learn. The machine lacks common sense. What would be obvious to you and me needs to be proven multiple times to the machine. To us the sky is blue, grass is green, and men and women buy different things. A machine knows none of these things. The marketing machines of today need to be driven by people.
So where do humans sit?
Maybe we should re-frame the question to “what do humans do well and what do machines do well?” This may be a bit more enlightening. Firstly, humans have deep understanding of their market, their product, and their competition. Humans can make the ideas, the structures, and the broad strategies. Humans see the bigger picture.
Machines, on the other hand, handle the minutia. In the world of programmatic advertising today, more than 10 million ads are auctioned each second. Knowing what ads to buy and buying them within 100 milliseconds is clearly out of the human skillset.
Optimization of campaigns can be seen as a joint enterprise between humans and machines. Computers can look at the patterns, see probabilistic trends and extrapolate audiences way beyond human skill. However, human creative thinking can reimagine the problem and look at different approaches. Where machines can try to do the same repetitive task better, humans can find new paths.
It’s a bit of a stretch but Henry Ford’s quote “If I gave the people what they asked for, I would have given them faster horses” comes to mind. The computer is looking for “faster horses”. But only a human imagination could have sold them cars.
I think we humans are safe. Our place is in the driver’s seat.
By Matt Harty, SVP of Asia-Pacific, The Trade Desk