We considered some recommendations from TM Forum for the churn model. Some recommended factors are relevant to subscriber churn:
-
A decrease of more than 30% in the monthly Average Revenue Per Unit (ARPU) or, in the case of prepaid subscribers, a decrease of more than 30% in the top-up frequency or rate (not the top-up amount).
-
A decrease in customer Net Promoter Score (NPS) score.
-
Late payments of post-paid bills.
-
A decrease in usage.
-
Poor network quality, such as dropped calls and service outages.
-
Widely publicized cheaper or unique offers from competitors.
-
Social media that identifies subscribers who solicit information about termination fees and change over offers.
Most of these factors are included with various weightings in most of the formulas that predict churn. The formula and the use of the propensity to churn metric continue to evolve as social media analytics continue to develop.
To date, the model for churn prediction includes these factors:
| Factor Type | Factor | Description | Data Source |
|---|---|---|---|
|
Customer Lifetime Value (CLTV) |
Customer Lifetime Value | The lifetime top-up value of the customer. | External Daily Usage |
| Tenure | Tenure Months | The customer tenure in months. | Subscription |
| Trend Features | Data Usage Trend | Customer data usage trend. | External Daily Usage |
| Top Up Amount Trend | Trend customer’s top-up amount. | ||
| Payment Frequency Trend | Trend the customer’s payment frequency amount. | ||
| Case Features |
|
Case | |