Skip to main content

google_analytics.meta_page_traffic

Daily readership of Signal Check, the dashboards, and the Cookbook, by page and session source.

Daily traffic for the meta-analytics GA4 property, which measures who reads Signal Check, the published dashboards, and this Metrics Cookbook.

Each row represents one day of traffic to a single page, from sessions that started from one source, medium, and campaign.

Keywords: Google Analytics, GA4, meta-analytics, Signal Check, dashboards, Metrics Cookbook, readership, page views, session source, utm_source

Provenance

Pulled daily from the GA4 Data API (runReport) for the meta-analytics property, by the same Lambda and pipeline as page_traffic. The Glue ETL job adds two derived columns, pagepath_redacted and page_kind. They exist only in this table and meta_page_events.

The property has two web streams. The Signal Check stream covers the Signal Check editions and the published dashboards, which carry the same tag. The Metrics Cookbook stream covers this site, including its PR previews.

Data starts on 2026-07-23.

  • Updated: Daily, at 07:00 ET
  • Retention: 7 years
  • Steward: EDCP Data and Research Team
  • Contact: Colin Douglas
  • Location: s3://<transformed-bucket>/google_analytics/meta_page_traffic/

Fields

One row per date per streamid per pagepath per pagetitle per sessionsource, sessionmedium, and sessioncampaignname.

Field Name Type Description
date string The reporting day, as a partition key (for example, “2026-09-22”)
streamid string Numeric ID of the GA4 data stream that reported the page. Stable if a stream is renamed
streamname string Display name of the GA4 data stream, “Signal Check” or “Metrics Cookbook”
pagepath string The page’s URL path as GA4 reports it. Signal Check paths include a private token. Do not use (see Notes)
pagetitle string The page’s <title> at the time of the visit. Signal Check titles include the edition number
sessionsource string Where the session started, for example slack, teams, email, report_link, latest-link, github.com, google, or (direct)
sessionmedium string Session medium. (not set) for tagged links, since they carry a source only. Otherwise (none) for direct visits, referral, or organic
sessioncampaignname string Session campaign. (not set) for tagged links, otherwise (direct), (referral), or (organic)
sessions bigint Number of sessions from this source that included this page on this day
screenpageviews bigint Number of times this page was viewed on this day, in those sessions
totalusers bigint Number of distinct users in those sessions. Do not sum
newusers bigint Number of first-time users in those sessions. Do not sum
pagepath_redacted string pagepath with the Signal Check token removed. Derived in the ETL. Filter and group on this
page_kind string What kind of page this is, one of signal_check, dashboard, redirect, cookbook, preview, or other. Derived in the ETL

Notes

Filter and group on pagepath_redacted, never on pagepath. Signal Check pages are served at a path that includes a private token, so the links can’t be guessed. pagepath_redacted drops it.

Use page_kind to pick pages. It is derived from pagepath_redacted, and the first matching rule wins:

  • signal_check: a Signal Check edition (/r/YYYYMMDD_signal-check.html)
  • redirect: a redirect page that forwards to the latest edition. These are listed by name in the pipeline’s ETL job
  • preview: a Cookbook PR preview (/canadalogin-metrics-cookbook/pr-preview/). These are the team’s own review views, not readership
  • cookbook: any other Cookbook page
  • dashboard: any other .html page under the publishing site’s /r/. This is a catch-all, so a new dashboard needs no code change. A new one-off page there would also count as a dashboard
  • other: anything else, including (unrecognized) paths

totalusers and newusers do not sum across rows. A reader who came from two sources, or saw two pages, is counted in both rows. This table can’t give an exact count of unique readers of an edition. sessions and screenpageviews are safe to sum. The reason is the same as why we can’t sum unique users in IBM Verify.