ParadeDB speeds up multi-table Postgres search joins by up to 1000x
ParadeDB says it has built native support for 'search joins'—relevance-ranked search queries across multiple normalized Postgres tables—without requiring data to be denormalized into a separate system like Elasticsearch. Using a StackOverflow dataset benchmark spanning users, posts, comments and badges tables, the company reports a three-table Top-K search query dropped from about 2 minutes on vanilla Postgres to roughly 100 milliseconds.
GoKawiil's interpretation of the reporting above, not reported fact.
If the performance claims hold up broadly, it could let teams running relational databases keep normalized schemas while still getting fast relevance-ranked search, avoiding the complexity and fragility of ETL pipelines into dedicated search engines like Elasticsearch. This suggests ParadeDB is positioning itself as an alternative to bolt-on search infrastructure for Postgres users, though the benchmark is self-reported and based on one dataset and query pattern.
- ParadeDB claims up to 1000x faster multi-table search queries versus vanilla Postgres in its own benchmark.
- The approach aims to let teams avoid denormalizing data into separate search systems like Elasticsearch.
- Benchmark used a StackOverflow-based dataset with tables up to 29.4 million rows across users, posts, comments and badges.
Source: paradedb.com, 2026-11-13
Published there as: “Do we still have to denormalize? Have your schema and search it too”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.