When we first started applying Machine Learning technology to solving marketing problems, it was a discipline that few outside of the operations research community had heard of. Cloud computing was in its infancy and big data was yet to become the ubiquitous term it is today. I remember going to some early meetings with prospective clients who wanted to know how we might help them analyze their marketing mix and finding that if I mentioned Machine Learning or Artificial Intelligence, eyes would quickly glaze over and people would lose interest (of course that could have just been me). It quickly became apparent that talking about technology and techniques wasn’t the best way to win friends and influence people! Instead we focused on the problem at hand and how we would help our clients solve the challenge. And that made sense at the time.
Since 2007, the world has of course moved on. There are now thousands of articles talking about ML and AI, countless movies looking at the hellish consequences of a dystopian future we all face when the machines take over (!) and a million shiny suited sales people walking into offices every day talking about their AI solutions as being the missing link in performance for their prospects (despite them rarely understanding the key concepts).
So why might now be the time for us to talk about this topic? Well as somewhat “old hands” in this area, we know a thing or two about some of the practical benefits and how to make them work for you. For example, in our strataQED solution, we can use our technologies to crunch through literally thousands of complex mix models per minute – testing combinations and permutations that a human simply can’t test – and checking that no stone is left unturned in the quest to build great insights. We also understand that machines on their own can be dumb – they need human insight to make smart decisions and results often require a human to interpret them before these insights are deployed. We know Machine Learning is valuable but it also has limitations. Our job is to maximize the usefulness whilst minimizing the constraints. I think we’ve done a good job on this in the mix modelling and optimization space.