Ridhuan Rangga KusumaRidhuan Rangga
Case StudiesSystem Design
Home
Writing

Case Study / 2021

Product Search Platform Rebuild

Elasticsearch and multi-layer caching rebuild, cutting search latency from 480ms to 65ms and lifting conversion by 23%.

480ms to 65ms

Search latency p99

After cache and ES tuning

+23%

Search conversion

Items added-to-cart after search

78%

Cache hit rate

L1 and L2 combined

<30 sec

Index freshness

Product update to searchable

problem

Search Was Becoming a Revenue Bottleneck

The legacy search stack degraded as catalog size grew. Popular queries were slow, typo handling was inconsistent, and ranking experiments required risky deploys.

caching

Multi-layer Cache Strategy

The platform used a small in-process cache for hot normalized queries and Redis for popular result sets, with explicit invalidation when product updates changed searchable fields.

results

Measured Outcomes

P99 latency dropped from 480ms to 65ms. Search conversion increased by 23% after ranking and typo tolerance improvements shipped.