Method
Time-series cross-validation
Rolling-origin cross-validation preserving time order.
6 cases from 5 companies apply Time-series cross-validation. 1 was checked in an online A/B test; 2 report results in production.
Typical tasks: Forecast demand and load, Decide who gets an offer, Predict churn and lifetime value, Detect anomalies and 2 more. Companies: Wix, DoorDash, Spotify and 2 more. Often combined with: Holdout / global control group, Time-series forecasting, Churn prediction, SHAP, Probability calibration and 14 more.
| Roonomics: How Economics Helped Us to Build a Churn Prediction Model | food-delivery | 2026-09 | |
| Real-World Forecasting with Deep Learning: How We Do It at Wix | saas-b2b | 2024 | |
| How DoorDash Built an Ensemble Learning Model for Time Series Forecasting | food-delivery | 2023-06 | |
| A Trial ML System Design Interview | — | retail | 2023-05 |
| Optimizing Markdown in Fashion E-Commerce with Machine Learning | e-commerce | 2022-08 | |
| How We Built Infrastructure to Run User Forecasts at Spotify | media-streaming | 2022-06 |