Predict Customer Churn Before It Happens

Upload your customer data, train an AI model, and identify who’s about to leave — no coding required, no data team needed, and your data never leaves your browser.

How It Works

🚀

Train a Churn Model

Export a CSV from your CRM or billing tool, upload it here, select the churn column, and train an AI model with live progress — all in your browser.

Train Model →
🔮

Score Your Customers

Load your trained model and a customer list for instant churn risk scores and clear risk tiers — so you can see who needs attention now.

Score Customers →

Why Customer Success Teams Use marijoAI

🎯 No Data Team Required

Train a professional-grade churn model through a visual interface — no Python, no SQL, no data scientists

🔒 Your Data Stays on Your Device

All processing runs in your browser. Customer data never touches a server — no DPA, no vendor risk review

📊 Live Training Feedback

Watch the model learn in real time with loss and accuracy metrics updated every epoch

⚡ Works with Any CSV

Export from Stripe, HubSpot, Intercom, or any tool — if it’s a CSV with customer metrics, it works

💾 Save and Reuse Models

Download your trained model as JSON and reload it anytime to score new customer lists

🎨 Get Started in 2 Minutes

Follow the step-by-step tutorial — you’ll have a working model almost instantly

What You Can Predict

marijoAI uses AI-powered binary classification to score each customer as "will churn" or "will stay." Feed it any numeric customer metrics:

  • Subscription churn — which customers will cancel this month?
  • Revenue churn — which accounts are likely to downgrade?
  • Trial conversion — which free users will convert to paid?
  • Engagement drop-off — which users are going inactive?
  • Renewal risk — which annual contracts won’t renew?
  • Expansion likelihood — which customers are ready to upgrade?

Any yes/no customer outcome you can export as a CSV.

💻

We are open-source!

marijoAI is released under the MIT license. Read the code, fork it, self-host it, or contribute — every line that runs in your browser is public.

View on GitHub →