Paper Company


  • A German media publishing company asked us to crete a solution to predict customer churn
  • As preserved revenue from non-churning customers is highly profitable and directly contributes to the topline, our mail goal was to reliably predict customer churn cohorts and make those cohorts adressable for customer success outreach


  • Account master data
  • Demographic data
  • CRM data including products consumed
  • Customer review
  • topics, articles, authors


  • Machine learning and statistical modelling algorithms
  • Big Data technologies for creating analytical boards and models
  • Python
  • Sentiment Analysis Algorithms
  • Cloud environment: AWS

Business top line KPIs focus to identify growing or declining businesses

  • Revenues growth
  • Liablities growth
  • Asset growth
  • Web traffic, relevance, news coverage, job opening, employees growth


  • Our final model predicted churning customers with over 90% accuracy
  • Based on our model, we built a solution that was integratable into the existing CRM system and directly usable for customer success representatives do approach customers with a high churn-risk