Question

A BI team uses Looker on top of BigQuery. A widely-used Explore joins several large tables and runs a heavy aggregation that takes 40+ seconds every time an analyst opens a dashboard. The underlying source data is refreshed once per hour by an ETL job. Analysts do not need sub-hourly freshness, but they need consistent, fast query response. What is the most appropriate way to model this in Looker to reduce query latency and cost?

Accepted Answer

Define a persistent derived table (PDT) with a datagroup-based rebuild trigger tied to the hourly ETL, so the aggregation is precomputed and cached in BigQuery.

Answer

Convert the Explore to an ephemeral (non-persistent) derived table so Looker recomputes the SQL inline on every request.

Answer

Enable Looker's result caching only and rely on the default cache policy without changing the modeling layer.

Answer

Increase the BigQuery on-demand query timeout and add more slots so the 40-second query completes faster.