By Rob Webster, Chief Strategy Officer, Crimtan.
Last Thursday, after our conference on data science, I had the pleasure of speaking with Anatoly Zhigljavsky over a beer. Anatoly is one of the UK’s preeminent experts in this space and, for the last two years, has been working with Crimtan to help us with our machine learning. But this was the first opportunity I had to chat to him away from work and outside an academic environment about how trends in computing will affect marketing.
The conversation started off with chess, a subject I am fond of as a deeply average but enthusiastic player. I recently read a lot about the infamous match between Kasparov and Deep Blue when humanity (represented by Kasparov) lost the title of best chess player to a computer. Kasparov knew he couldn’t compete with the computer on tactics – which involved the calculation of millions of move combinations. However, he was confident he could still win on strategy, for example controlling certain key squares which would not have any advantage for dozens of moves but eventually lead to a win in the end game. So he was shocked when Deep Blue played a move that was tactically weak, but which countered Kasparov’s strategy. Kasparov resigned angrily, believing foul play and that a human had been working with Deep Blue to assist the result.
This was 1997 and computers were powerful enough on tactics but still had no way to comprehend more subtle strategies. Most of the chess playing world believe this win for Deep Blue was a hoax, a belief helped by the fact that deep blue was immediately dismantled, no rematch offered, and no inspection of the move lists allowed.
Move forward to 2015 and computers have further extended their advantage in tactics and computation. And they are now catching up on strategy – not only on the chess board but in other fields like marketing. This analogy highlights the difference between machine learning (tactics) and artificial intelligence (strategy). To truly possess AI, the machine has to be able to think more creatively and holistically than just going through a series of chess moves or mathematical functions. This point, according to Anatoly, has already come in chess and it’s beyond debate that if it hasn’t, it is close.
For digital marketing in 2017, machine learning is regularly used in programmatic advertising to optimise campaigns. Dozens of variables are analysed in real time and techniques such as ‘mutual interest’ and ‘gradient boosting’ are used to improve the ROI of programmatic campaigns. But as yet, little is happening in terms of developing the strategy. Here we would look to strategy as being typically longer-term functions which are harder to express as a simple number. For example strategy could be around the increase and measurement of brand awareness, the strategic focus on lapsed customers as opposed to regular customers or indeed new customers. Even decisions around how much to spend and on which channel.
So if we have a spectrum of events; with mostly tactical on the left hand side (micro optimisations on variables like time, frequency, browser etc.) and strategic on the right (the broad marketing objective, messaging, creative and high level planning), what can we expect to see over the coming years with regards to machine learning and AI?
Well, we will see machine learning take over the functions on the left hand side at an ever increasing pace. Channels where machine learning has little input today, like out-of-home and TV will have a heavy machine learning influence within five to ten years – by which point machine learning will have taken over almost all the tactical side of marketing.
However on the strategic side, things are going to look quite different. Machine learning will have nothing much to do with strategy as it is beyond the scope of simple algorithms, but AI will have a slow effect on strategy as areas like attribution become more advanced. For the next 20-30 years AI will have a modest (but increasing) impact on marketing strategy, mostly in attribution and scenario modelling. Only when an AI can actually answer questions from a board of directors or shareholders in a competent fashion and understand all that a marketing director (and their team) do will AI really take hold for marketing.
The implications for this are quite profound. It means that today’s marketing teams need to invest in machine learning to maximise the impact of their marketing activity and look to expand it out of programmatic optimisation. Attribution, audience creation, price setting and portfolio optimisation all need to be controlled on a tactical level by machine learning with (for now) a human taking strategic control. Skills-wise, it means marketers need to be able to bring in understanding about how to use machine learning, whilst also making sure that strategic skills are readily available to control setups. From a career perspective, there will still be plenty of jobs in marketing for the next 20-30 years, with the key skills being strategic control and the ability to set up channels to best utilise machine learning. Gradual implementation of AI in the strategic function is also an important trend for the biggest and best marketing companies and departments.
At some point over the next 30 years, however, we will have what is termed the singularity. Where the human touch no longer has an advantage, even in strategy. At this point, AI takes over and humans will just be there to sense-check and try and keep up with what is happening. It will be a world we can scarcely imagine. Fortunately, during the period leading up to this we will see AI’s develop from toddlers to petulant teenagers to fully formed adults, and this will give us adequate time to adjust.
So what are my conclusions from this fascinating chat with Anatoly? Well, if you are reading this as a marketing professional in the middle or end of your career, AI and automation is little threat. Focus on strategy and bringing in the data and tech skills. If you are at the start of your career, it is a threat to the day job of any tactical work that is still done by many young marketing professionals – so you need to hone your strategic skills as soon as possible. However, for both groups, provided the strategic talents exist, there is little threat and a huge amount to look forward too. It is our children that will grow up in a world where most of the jobs their parents did, including 99% of marketing, will be controlled by AI.