DuckDB is often described as "SQLite for analytics": a full SQL engine that runs inside your process — no server to install, no connection string to configure.
Where it fits
| Workload | Good fit |
|---|---|
| Ad-hoc analysis of local files (CSV, Parquet) | ✅ Excellent |
| Powering dashboards from a data warehouse | ✅ Often |
| High-concurrency transactional apps | ❌ Use Postgres |
Trying it
DuckDB reads files directly — no import step:
SELECT *
FROM 'orders.parquet'
LIMIT 10;
That single query — no CREATE TABLE, no loading — is most of the pitch.