I will predict churn and late orders with machine learning

Which customers will leave? Which orders will arrive late? You probably have the data to know before it happens.

I build machine learning models on your business data and explain the results in plain language, so you can act on them.

PROOF: 100,000 REAL E-COMMERCE ORDERS

On the public Olist dataset, I built a model that flags late deliveries at the moment of purchase. Checking only the 10% riskiest orders finds 23% of all late deliveries: 2.3x better than random checks. I also found that late orders reviewed before arrival get 1-2 stars 81% of the time, vs 9% for on-time orders.

HOW I WORK

- I compare the model against simple rules, so you see what it really adds

- Time-based validation: the model is tested on future data, like real life

- No inflated accuracy promises before I see your data

YOU GET

A report in business language with charts, the source code (Python, Jupyter) and documentation. In Premium, the model runs on a schedule and sends predictions to Google Sheets or WhatsApp.

Not sure if your data is enough? Start with the Basic package: I check it and tell you honestly. English, Spanish or Portuguese.