What’s more important: The money you have or the money you earn?

By Michael Froment - May 20, 2019 | 255 0

Generally speaking, teams are rather excited when they start trying out machine learning, AI, predictive analytics and algorithms that have the potential to boost their work performance. However, the feedback in the press frequently overestimates the results achieved and gives the impression that soon everything will be done by a machine.

Unfortunately, marketers who have tried AI-powered marketing software have found themselves disappointed. Our team was very surprised to hear this, and so opened discussions with the first users of AI-powered solutions, mainly based in the US.

The first lesson to come out of this workshop was the feeling of a “Blackbox”. This term covers many different things, but the primary issue was that they wanted to know and understand how the machine worked. So, while the tool may have been doing its job well, it ignored the fact that marketing teams aren’t there to simply copy and paste the decision it has made; at least not at our current level of maturity. And that’s not without forgetting that the results weren’t even meeting the expectations of teams or management.

Nevertheless, this situation provided the perfect opportunity for us to rethink our R&D strategy. First of all, our team decided to help marketers understand the criteria that influence a conversion.By conversion, we mean any type of value that leads to events such as making a purchase, filling in a form, signing up to a newsletter, engaging on social media or viewing pages.

Machine learning is a vast topic, but we decided to apply quite a classic and popular approach called the Prediction Tree. It’s used in decision analysis to help decision makers identify a strategy that will most likely achieve a goal.

A decision tree is a flowchart-like structure in which each internal node represents a “test” on an attribute (e.g. whether a coin flip comes up heads or tails), each branch represents the outcome of the test, and each leaf node represents a class label (decision taken after computing all attributes). The paths from root to leaf represent classification rules.

Read more