In the autumn of 2018, our data team at The Reference was asked to provide consultancy in the field of data analytics for the marketing department of a large retail company. The assignment consisted of two parts. Our main task involved making advanced marketing analyses that enabled the management to make founded decisions on a strategic level. Secondarily, we also delivered reports based on which the staff could manage its day-to-day operations
In addition to carrying out the analyses and reports that were asked for, we decided to go the extra mile. More specifically, we rolled out multiple predictive models that empowered the team to observe customer flows closely and react accordingly. One model for example picked up the chance a customer would churn at any given point in time. The result of this valuable insight was a marketing campaign that targeted customers who were about to drop out and eventually led to substantial decrease in churn rate. Another model calculated the probability of a prospect converting into a true customer. Potential customers that possessed a conversion likelihood of over 80 percent were personally targeted via multiple channels. Through its implementation, we also managed to improve figures drastically in the field of acquisition.
We developed predictive models with a wide range of methods such as the hugely popular random forest. This learning algorithm takes a table of predictive features as input and generates the likelihood percentage of a certain event occurring. In our case, these events were either a dropout or a conversion. A thorough business understanding and extensive deliberation were required to come up with the most fitting bag of predictive items. Eventually, we managed to achieve a predictive accuracy that was high enough to identify the right customers and tackle the predetermined targets.