Case Study: Tinder achieves real-time, global-scale matchmaking and search with Elastic

Tinder - Customer Case Study

Tinder, led in this case by VP of Engineering Maria Zhang, faced the challenge of delivering a simple, low‑latency mobile matching experience at massive global scale: ~17 billion event counts daily, 280 million+ Elasticsearch queries per day (100+ billion annually), 1B+ swipes daily across 196 countries and 52 languages. The engineering problem included real‑time bi‑directional ranking with tens of ranking dimensions, ~60 Elasticsearch nodes, ~20,000 index updates per second and up to 60,000 doc hits per query — all while keeping the UI minimalist for users.

To solve this, Tinder built a layered architecture combining front‑line load balancing, an inference layer, a core request flow (dispatch, auth, cache, query builder, federation, post‑ranking, blending, suppression, hydrating, render), dedicated Elasticsearch clusters, and an events preprocessing pipeline. Performance tuning, community-driven Elasticsearch fixes (reduced GC, improved shard concurrency, source filtering) and rigorous testing enabled the platform to sustain the scale and power features like Smart Photos, Boost and Tinder Social—supporting billions of matches and queries daily and continuing rapid product growth.

Tinder

Maria Zhang

Vice President of Engineering