Source:- thedrum.com
In recent months you’d be forgiven for thinking that media buying as a discipline is going to hell in a handcart, with Google itself being the latest to suffer the slings and arrows of programmatic media buying’s outrageous fortune.
But instead of viewing recent events as some kind of advertising apocalypse, we should perhaps be viewing this period as a painful adolescent relationship between adland’s humans and machines. A period where, having rushed headlong into the data and automation, we take a step back, have a bit of a reboot and make some serious choices about how we work alongside machines, for the long-term benefit of our industry.
P&G’s Marc Pritchard was of course right when there is far too much that is murky about today’s media buying landscape. But it’s a misconception that technology is not to blame. Technology is simply a tool that has been handed to a sector that initially had no idea how to use it. A degree of chaos where two worlds collide should perhaps have been expected, and is precisely why seeing Google making a commitment to “building systems with people in mind at the start of the process” is exactly the direction we should all be going in.
Not all technology is created equal
What this confusion does tell us is that it’s time for the growing demand for artificial intelligence (AI) and machine learning to change the process of planning and buying media for the better, redefining client experience and the role of industry creatives.
First, though, we need to understand why we can’t just lump all technology in a bucket and label it ‘bad’. AI and machine learning are two terms that are increasingly used and it’s essential that clients understand exactly how they will alter the industry and what the differences are between the two.
AI is the broader concept of machines acting intelligently in carrying out specific tasks. It carries everything that is the act of learning and its goal is to recreate human intelligence.
Machine learning is a subset of AI. It is pattern recognition and uses algorithms based on data to learn how to solve specific narrow tasks. You need AI researchers to build the smart machines and you need machine learning experts to feed the AI with tools that will enable it to act intelligently. AI is driving the car, machine learning recognises the car in front.
Share and share alike when it comes to data
One of the reasons the use of machines in media buying has fallen is that so many actors have so many different data sets being used independently of each other as each jealously guards their information.
Instead, vendors need to be transparent about the amount of data used in their AI and machine learning processes. Instead of looking at different factors independently, the data must be used to collate different insights together to make sure the problem is solved.
Expanding your data landscape is critical to success. We use 82 different data sources from YouGov and Nielsen to the client’s sales information and the weather to predict the most probable outcome of a media plan.
Model behaviour
Using data democratically is just one way in which the media buying model needs to change. With so much inheritance from the creative and human side of media and marketing, it can be hard to conceive that planning and buying should be taken over by an algorithm, that decisions can be made by machine.
We would argue that this is exactly how it should happen. The scale of decision-making in a fully automated sizeable media campaign is more than a human brain can cope with. And with publisher content changing every second of every minute of every day, the variables are too huge to contemplate.
With machine learning, the client simply goes to a platform, chooses their end goal, decides how much budget to allocate and for how long to run the campaign. AI generates a proposal based on a mathematically optimised media plan including every variable the client has chosen. Because all the data is online and up to date from both client and publisher, there’s the opportunity to dial elements up or down and optimise on the fly.
It would seem like this is making the human redundant. Certainly, it cuts out the politics that has been part and parcel of media buying relationships since the bad old days and which continues to muddy the waters in Pritchard’s view.
The machines will still need people. There are nuances and bespoke arrangements where you just can’t excise the human touch. But the one thing AI just doesn’t need is its slice off the top. It shouldn’t just be the times that are a changin’, it should be the whole media model.