SnappFood · Data Science & ML · 2019–20
SnappFood ML
Delivery-time, churn and fraud models for a food-delivery platform with 10M+ users.
- Machine learning
- Production
- Analytics
Overview
On the customer-experience team of Iran’s largest food-delivery platform, I worked on the models that decide when your food arrives, who is about to leave, and which vendors are gaming the system.
Contributions and results
- Adapted Uber’s DeepETA to motorbike delivery: ~27% better ETA accuracy and 24% fewer delivery delays.
- Churn-prediction pipeline over 3M+ users feeding reactivation campaigns that cut monthly churn by 13%.
- Vendor-fraud detection that lifted CSAT by 10% and NPS from 5 to 7.
- Built the company-wide Octopus BI layer of department-specific KPI dashboards.