Early churn prediction is a main concern of every subscription business. We created a model for prepaid customer base churn prediction which identifies early customers’ retention decline and allows to work with such clients before their next step - switching to your rivals.

Effort and duration


Tech stack




Oracle Databases




Who is our client


the third largest mobile operator in Belarus
Basic moments
  • customer base descriptive analysis
  • clusterization of base
  • patterns and insights identifying
  • testing of hypothesis
  • modelling and validation for segmented base
  • control/target testing
  • deploy
  • effectiveness measurement and re-train of model if necessary
Project overview
End-to-end churn prediction model. It is a common true that acquiring a new customer usually costs higher than retaining the old one. The companies are interested in identifying segments of customers whose interest goes down and proactively engage them with special offers instead of losing them. Early churn prediction is a main concern of every subscription business. It is necessary to detect the first signs of customers’ behavior change before the point of unreturn.
Business value
Our team developed an early churn prediction model. It evaluates customer activities to elicit patterns of usual customer behavior. Once the customers behavioral pattern changes – it gives us a probability, churn score for a particular customer. Based on customer’s interests we allocate him in a group with similar demands and propose a relevant offer to retain him.
Contact us.

Your name

Your e-mail

Your message

max size is 15 MB; allowed extension: jpg,jpeg, png, txt, pdf, doc, docx, xls, xlsx,ppt, pptx
I agree to the processing of my personal data