Run an analytics query

Time range for the query (required)

Metrics to query. Strings for flat or Nested (errors.count) metrics; objects for Map-key access (e.g. resource_sizes['script']). first_party_error_rate and third_party_error_rate split error_rate, the share of pageviews with any JavaScript error, by error_origin. AVG-only ratios 0-1. Grouped by errors.fingerprint alone, error_beacons is the weighted pageviews carrying each signature and error_rate their share of every pageview of the bucket or window. performed_dns_lookup and new_tcp_connection are dns_lookup and tcp_connection over the pageviews that did a lookup or opened a connection, the rated durations. skipped_dns_lookup_rate and reused_tcp_connection_rate are the AVG-only share 0-1 of the pageviews reporting the timing that did not. cdn_served_rate/origin_served_rate are the share answered from the layer's stored copy, the sum of its hit, stale, updating and revalidated status rates below. Cache-status rates follow the pattern (cdn|origin)_status_rate_<status> (lowercase, e.g. cdn_status_rate_miss; unreported = no status sent) - AVG-only ratio 0-1 over all pageviews. cdn_hit_rate/origin_hit_rate are legacy aliases for *_status_rate_hit. The ttfb_phase_* ids split coalesce(ttfb_final, ttfb) per pageview, and the avg of ttfb_phase_connect, _edge, _origin, _backend, _unattributed and _unmeasured sum to its average over the pageviews they cover, those whose first byte is at least the DNS lookup plus the TCP connection.

Aggregation functions to apply. good_share, needs_improvement_share and poor_share are the part of the bucket's page views that reported the metric and fall in that rating of GET /metric-thresholds, 0..1 and unweighted. They take only lcp, inp, cls, ttfb, fcp, performed_dns_lookup and new_tcp_connection. For the last two the denominator is the page views that did a lookup or opened a connection.

Grouping configuration (time and/or dimensions)

IANA time zone whose boundaries start the six_hours, day, week and month buckets, such as Europe/Berlin. A day is 23 or 25 hours long across a DST switch. Bucket timestamps stay UTC. In a zone whose offset is not whole hours a bucket starts at the UTC hour that holds the local boundary. Without it the buckets start at UTC boundaries

Optional filters

The window to compare with, in one of three forms. A mode: previous_period is the smallest whole number of weeks earlier that covers the window, at least one, previous_day 24 hours earlier, previous_week 7 days earlier, previous_month 30 days earlier. A window {from, to} of any length. A release {release_id}: from its released_at, as long as time_range, cut at the next release of its application and at now. Time buckets pair by their offset from the start of each window.

Compare before against after a release inside the window. Deprecated: send the two sides as time_range and a compare_to window.

Maximum rows to return

Offset for pagination

Sort configuration

Backing identifier for the visitors metric: 'stitch' (preferred - edge-derived, present under ANONYMOUS) or 'session' (consent-gated session_id). Null resolves to the application's default (stitch unless the application only collects sessions). Ignored by all other metrics.

The query key whose value the url_param dimension groups by, required with it and refused without it. The key has to keep its value in every application of the scope: listed in collection.query_values_for, or, in an application that does not keep query keys, a search term key a page type keeps (q, search, searchparam) or the platform's controller key (cl on OXID). Any other key answers 422