Book your personalized demo
See what your data can do for
Your company can do
Your company can do
Learn how to get started and what results you can expect.
Contact us
A telecommunications provider now sees the risk of customer churn 3 months in advance.
A major telecom provider was losing 15 % subscribers annually. The customer retention team only reacted after a customer expressed a desire to cancel. Using predictive analytics, the company now identifies at-risk customers three months earlier.
Service:
Predictive analytics
Timeline:
January 2025 - June 2025
Country:
Ukraine
Industry:
telecommunications
Company:
NDA
Summary:
In the telecommunications industry, it's nearly impossible to win back a customer once they've signed with a competitor. The provider therefore shifted from a reactive to a proactive strategy. Together, we organized historical data and trained a predictive model. This model identifies at-risk customers three months before they're ready to churn and pinpoints the reasons for their dissatisfaction. This allows the team to address issues before the customer even considers switching.
Business area
The company is an established provider with over 70,000 subscribers. With an annual churn rate of 15%, it was essential to protect revenue from existing customers.
problem
The biggest challenge was timing. Contacting a customer during the termination meeting is usually too late. Trust is often already lost, and discounts are of little help. The company had to find dissatisfied customers while they could still be retained.
Solution
Step 1: Merge data from different systems
The project began with the collection of all customer information in one place. This included payment history, technical logs, and support call records. We linked data that was previously stored in different departments. This enabled a unified view of each subscriber's experience over time.
A unified view for every customer. It combines historical data with AI predictions in anonymized form.
+
Step 2: Recognizing subtle signs of frustration
We searched the data for patterns preceding a cancellation. Examples included decreasing internet speed or sudden changes in payment behavior. These signs were often too subtle for manual detection.
+
Step 3: Train the prediction model
We used thousands of examples of customers who had already left. The system learned the patterns of these departures. Now every subscriber receives a risk score. This allows for predictions even before a formal complaint is filed.
Step 4: Switch to proactive communication
The team receives a weekly list of at-risk customers. They proactively call and offer direct solutions. Technical errors are fixed or better tariffs are offered. This approach turns a cancellation into a positive experience.
Results
- 80% Accuracy in identifying at-risk customers.
- 8,400 customers are saved each year before they change.
- The exodus was reduced by solving hidden problems.
- Communication It became personal because the team knew the reasons for the dissatisfaction.
