From: Emeka Okafor emeka.okafor@tundemobile.ng To: You Subject: Subscriber churn — need a prediction model and API
Hello!
I'm Emeka — I run customer retention here at Tunde Mobile. We're an MVNO based in Lagos, about 200,000 subscribers on our books. Good growth, good coverage, but we have a problem: every month we lose 2-3% of our subscribers, and we don't know which ones are about to leave until they're already gone.
My retention team makes outreach calls, but right now they're guessing. They pull a list, start calling, and hope they reach the right people. We need to stop guessing.
I have a clean export from our billing system — 12 months of subscriber data ending March 2025. About 7,000 records. Each row is one subscriber: their plan type, tenure, usage, charges, complaints, contract type, and whether they churned.
What I need from you:
-
A model that predicts who will churn. I want predictions ranked by risk — not just "yes" or "no" but a score my team can sort by. Highest risk at the top.
-
Feature importance. I want to know what's driving churn. Is it the plan type? The complaints? How long they've been with us? If I know what matters, I can work on fixing it.
-
An API endpoint. My team needs to be able to send a subscriber's data and get back a churn prediction. We'd query it weekly as part of our outreach workflow.
The data file is subscribers.csv and I've included a data dictionary so you know what each column means. Everything should be straightforward — this is a clean export, no messy joins or missing tables.
I'm eager to see what you find. Kindly let me know when you have initial results — my team is ready to move on this.
Best regards, Emeka Okafor Head of Customer Retention Tunde Mobile, Lagos