CASE 132 · BRINK · 2025
Search relevance the team can tune themselves.
A job board was paying Algolia $7,600/month for search that returned 18M monthly queries against 800k job listings. The pricing tier was tight; the relevance tuning was opaque. We migrated to OpenSearch Service with a careful relevance reconciliation.
Job board
MIGRATION
2025
RESULTS
What changed, by the numbers.
SEARCH BILL
−63%
RELEVANCE NDCG
+4%
p95 LATENCY
< 80ms
TUNING OPACITY
TRANSPARENT
HOW IT WENT
The Algolia win had been the time-to-launch — wonderful out-of-the-box relevance, near-zero engineering. The Algolia cost was the per-query and per-record-stored line items, which at 18M queries and 800k listings had reached the point where engineering felt cheap by comparison.
OpenSearch Ingestion absorbed the existing data pipeline. The hard work was relevance: we ran A/B comparisons between Algolia and tuned OpenSearch queries against a labelled judgement set the team built. After three tuning cycles, OpenSearch nDCG came out 4% ahead.
Search bill dropped 63%. p95 latency stayed at sub-100ms. The relevance tuning is now transparent — the team can read the OpenSearch query, understand what’s being scored, and adjust without filing a vendor ticket. The kNN plugin opens future semantic-search experiments.
RELATED · SAME DOMAIN
Other engagements in this space.
READY WHEN YOU ARE
Let's get your AWS bill (and architecture) in order.
The discovery call is free. You walk away with at least one concrete idea — even if we never work together.