google_analytics.error_events
Daily count of errors flagged by CanadaLogin to GA4.
Keywords: Google Analytics, GA4, errors, error codes, forms
Provenance
Pulled daily from the GA4 Data API (runReport). Only events with error or error_id fields are kept.
- Updated: Daily, at 07:00 ET
- Retention: 7 years
- Steward: EDCP Data and Research Team
- Contact: Colin Douglas
- Location:
s3://<transformed-bucket>/google_analytics/error_events/
Fields
| Field Name | Type | Description |
|---|---|---|
| date | string | Reporting day, partition key (e.g., “2026-08-16”) |
| eventname | string | GA4 event name (e.g., form_step_end, form_api_call_end) |
| form_id | string | Form that emitted the event (e.g., passwordrecovery, signup) |
| step | string | Form step number |
| error | string | error event parameter. |
| error_id | string | error_id event parameter. |
| rp_name | string | rp_name event parameter. The relying party (partner service) the user came from |
| rp_client_id | string | rp_client_id event parameter. The relying party’s client registration |
| eventcount | bigint | Number of events |
| totalusers | bigint | Number of distinct users. Do not sum. |
| error_code | string | error_id, else error. Derived in the ETL |
Notes
Use error_code, not error or error_id. Some pages report one or the other, and a few rare circumstances report both. This is an open defect. Error code is calculated as:
error_idif its not(not set)and not blank- otherwise the first word of
errorif its not(not set)and not blank - otherwise the row is dropped
Code columns are strings. This means that leading zeros should not be dropped, but an open open defect remains where they are coerced to ints at some point upstream. This means that the error code “000” is sometimes reported as 000 and sometimes as 0.
rp_name and rp_client_id are often blank, and blank rows are kept. Manage CanadaLogin sends rp_name but no rp_client_id, and one rp_name can have several client IDs. Most errors caused by a misconfigured relying party arrive with no relying party at all, so a blank value on those errors is itself a finding.
The totalusers field does not sum across rows, do not add them. This is a common pitfall and the reason is similar to the reason we can’t sum unique users in IBM Verify.