An internal BI dashboard might tolerate a five-second query. A chart embedded in a customer's own product cannot. User-facing analytics are the queries that load when a customer opens your app, filters a table, or scrubs a date picker, and the architecture has to assume thousands of those happening at once across thousands of tenants.

This shifts the design constraints. The queries are short and parameterized. The data has to be filtered by tenant on every read. The system is judged by p99 latency and concurrent request rate, not by how fast a single ad-hoc query runs.

User-facing analytics is the most common shape Tinybird is deployed for: a column-oriented engine behind a parameterized HTTP endpoint, fed by a streaming ingestion path.