It could be argued that the widespread coverage of subjects such as Machine Learning (ML) and Artificial Intelligence (AI) has only served to confuse many people as to the current state of play with these technologies.
The adtech industry is filled with a fear that if you are not using a certain technology then you are missing out. As a result, many business managers now use the phrases ‘ML’ and ‘AI’ as interchangeable (when in fact ML is a subset of AI) and seem to regard them as some kind of ‘plug-and-play’ solution capable of planning and running new campaigns entirely without human input.
Smart advertisers are utilising machine i.e. data-driven learning, applying both macro and micro insights, in combination with the human touch, to help reach relevant customers at every stage of the consumer journey.
In my opinion, there is still a long way to go before associating this with anything that could genuinely be called ‘AI’. AI is computer science dealing with the simulation of intelligent behaviour whereas ML teaches computers to discover insights and patterns without being told where to look.
Where we are today
Google is making some fantastic advances in AI; it is hard not to be impressed by the video of Google assistant making a phone call to book a hair appointment. However, I recently came across a very humorous tweet directed at Amazon, from a lady who had purchased a toilet seat and had been targeted with ads for toilet seats ever since. “I am not a collector!” she exclaimed. This shows exactly how quickly marketing technology can lead you down a rabbit hole, once human knowledge and experience is removed entirely from the equation.
Pure algorithms that can be left to their own devices don’t exist, algorithms need data and too often bad data in leads to bad data out. Hence why so often you see advertising retargeting for one-off purchases or targeting fraudulent inventory as it appears to perform.
Surprisingly, I still meet business leaders who say things like, “Yes, we’ve got Machine Learning” as if it’s something that you receive in a box and stick in a corner to get on with it. This approach is doomed to failure, as unless companies think about the inputs for their ML algorithms and also what goals they are looking to work towards they will always struggle. For example, how can you have the same algorithm for behavioural data looking for new customers and retargeting data looking to shore up users who have put items in their basket?
Machine learning solutions are all about understanding people’s behaviour to deliver commercial performance: we analyse online user data and optimise all our campaigns towards behaviours that have been shown to lead to a conversion.
Let’s take the example of a weight-loss company who approached us recently to roll-out a campaign targeting women only. When we started to test the campaign online, our Machine Learning technology quickly identified that the campaign was particularly effective at engaging customers who had also recently conducted online searches on the subjects of ‘weddings’ and ‘holidays’. While some may claim it is obvious that those customers displayed interest in those topics, that level of robust insight alone allowed our client to instantly optimise their budget and creative to target more of those customers with great success.
While this particular insight might seem like common-sense, Machine Learning can obviously spot far subtler correlations of online behaviour (invisible to human researchers) that could be extremely valuable in the context of a particular campaign. In essence, Machine Learning aids the people within advertisers and their agencies to identify often-hidden opportunities to apply traditional marketing skills and creativity to optimum effect.
The way forward
The Machine Learning technologies available to advertisers today are fantastic tools, but they still have a long way to go, and still remain a set of potential capabilities rather than a solution in themselves. Human input, intelligence and creativity are still very much required to deliver improved performance.
My advice to marketing directors and CMOs is not to think of ‘Machine Learning’ as simply a box to be ticked. Instead, start with the basics. Why use Machine Learning in the first place? What do you hope to achieve? What output do you want to produce? Only then can you begin to truly consider what you need the technology to do and which tech partners are best placed to deliver what you need.
Above all, remember that the capabilities of contemporary Machine Learning technology are truly amazing, but only in combination with the incredible talents, knowledge and experience of good old-fashioned human beings.
