Case Study: GO-JEK achieves real-time demand forecasting and dynamic pricing with Google Cloud Platform

GO-JEK: Using Machine Learning for forecasting and dynamic pricing

GO-JEK is Indonesia’s leading “super app,” offering more than 18 on-demand services (ride-hailing, food delivery, payments, etc.) and connecting over a million drivers with millions of customers across 167 cities. Faced with severe traffic congestion and hypergrowth, the company needed to match drivers and requests in real time, forecast demand, and implement dynamic pricing while ingesting massive telemetry — billions of pings and roughly 4–5 TB of data per day.

Working with Google Cloud and Professional Services, GO-JEK built a centralized data and feature platform using Google Maps Platform, Cloud Dataflow/Apache Beam, Kafka/Pub‑Sub, BigQuery, Cloud Bigtable, Cloud Memorystore, and Cloud Machine Learning Engine. The solution enabled real‑time feature serving, scalable model training and streaming inference for demand forecasting and dynamic pricing, meeting 30 ms SLAs at 10,000+ rps. Outcomes include optimized routes for 1 million drivers, improved customer experience, faster time to market, lower infrastructure overhead, and readiness for international expansion.

GO-JEK

Ajey Gore

Chief Technology Officer