New ZealandTechnology and telecom2025 – ongoingSpark New Zealand
AI churn-prediction model for prepaid mobile customers
Spark says its prepaid churn model cut customer churn by 20-30%
Problem
Before the model, 30-35% of Spark's prepaid mobile customers left after a short period.
Approach
Spark built a churn-prediction model for prepaid mobile customers. Building it took nine months, and getting the business to adopt it took three more. Spark also gave Microsoft Copilot to 700 staff, later 2,400, and built SparkGPT, a ChatGPT-based tool, for its 5,200 staff.
Results
Takeaway
Adoption took a quarter of the total project time, after nine months of model building.
Categories
- Size
- Large enterprise
- Industry
- Technology and telecom
- Business function
- Marketing and salesCustomer service
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Sponsored content written for Spark. It says the prepaid churn model took nine months to build and three to adopt, and cut churn by 20-30% from a prior 30-35% loss of prepaid customers. It also describes Copilot engagement of 92% across 2,400 staff and the SparkGPT tool.
Full details are in the original sources.
