**From:** Emeka Okafor <emeka.okafor@tundemobile.ng> **Subject:** Customer churn prediction project Hello, I'm Emeka, I run customer retention at Tunde Mobile here in Lagos. We're an MVNO — about 200,000 subscribers, mostly prepaid, serving young professionals and small business owners across Lagos and Ogun State. Our problem is churn. We lose around 2-3% of subscribers every month, which doesn't sound like much until you multiply it out. Last quarter alone we lost about 15,000 customers. My retention team does outreach — calls, special offers, plan upgrades — but right now we're basically guessing who to call. We wait until someone's usage drops to zero and by then they've already ported to MTN or Airtel. I have twelve months of subscriber data from our billing system: monthly call minutes, data usage, top-up frequency, complaints logged, plan type, account tenure, whether they churned. It's about 7,000 subscribers in this export — a snapshot from our billing system. Clean CSV format. What I want is a model that can tell me which subscribers are likely to leave in the next 30 days, ranked by risk, so my team can reach them before they go. And I want it running as an API we can query each week. I also want to understand what's driving churn — is it network complaints? Data usage patterns? Plan type? If I know the drivers, I can feed that back to our operations team. I've attached the dataset and a data dictionary. The churn column is binary: 1 if the subscriber left that month, 0 if they stayed. Let me know how you'd approach this and what you need from me. Best regards, Emeka