Open BigQuery in Google Cloud Console and navigate to the public dataset chrome-ux-report under the all dataset
Write a query joining tables by their YYYYMM name pattern, selecting your origin and extracting the p75 density for LCP, INP, and CLS from the metric histogram structs
Use date partitioning or explicit table name enumeration to build a time series across multiple months
Join against the experimental dataset variants for faster query performance and lower bytes billed if trend resolution at the monthly level is sufficient
Export results to a spreadsheet or visualization tool and annotate notable deployment dates to correlate code changes with field data shifts
Known gotchas
CrUX BigQuery tables are updated once per month and represent a 28-day rolling snapshot ending near the table's month boundary; the data is not real-time and cannot reflect changes made within the current month
Querying large date ranges across multiple tables scans significant data; use column-level selection and filter on origin early in the WHERE clause to minimize billable bytes
The FID metric was retired and replaced by INP in the dataset; queries referencing first_input do not reflect current performance standards and should migrate to interaction_to_next_paint
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp