The client was struggling to simultaneously identify the customers, who are the most critical churners while also sustaining the highest return on investment from marketing campaigns.
We first applied a logit model to understand what were the most impactful drivers of churn. Additionally, the effect of the drivers was quantified. Next, we used extreme gradient boosted trees as the approach for predicting the churn on an individual level.
We helped our client to efficiently spend the marketing budget and target efforts only at customers that were classified as part of the top percentile churners, reducing churn by 20% among the likely churners.