Increasing customer engagement and lifetime through improving the digital TV recommendations engine​


Our client, a global telecom, had a legacy recommendations engine (RENG), that was often suggesting repeat recommendations and was not learning from customer choices. Thus, our client was failing to increase customer engagement. ​


We built a robust measurement framework using large volumes of real-time data, to help evaluate the impact of the RENG on different customer interactions. We also improved the functionality of RENG: eliminated repeat recommendations, wrong recommendations, wrong system flags. We also developed and executed a three-step improvement initiative​

17% more shows watched through recommendations, 11% less churn in the groups with highest watches


We helped our client increase by 17 %
the number of shows watched through recommendations through the refurbished RENG.
The groups with the highest increase in recommendations watched also exhibited 11% lower churn rates than their counterparts, thus driving additional value for the ​telecom.


  1. Collaborative suggestions – live algorithm providing customer-level propensity scores to watch a show that someone similar to them has seen​
  2. Machine learning – live algorithm which retrains the itself based on actual customer behavior 
  3. Holistic 360 RENG –three modules (metadata, collaborative and machine learning) providing one-set of recommendations unique to each customer​