Skip to content
All field notes
Measurement
Published
Nov 4, 2026
Reading time
11 min read

GA4 Custom Channel Groups: Audit Rules, Order and Primary

Diagnose GA4 channel misclassification with scoped source evidence, ordered rule tests and separate checks for custom reports and the Primary channel group.

AP
AdProof Research
Independent measurement desk
Source slips pass through ordered rule gates into separate folders, alongside a comparison of two reporting folders.
Rule order and the selected reporting dimension determine classification. Conceptual illustration without campaign results.
In this guide (9 sections)

First establish whether the source information is missing or merely classified differently. If the required values exist, test the custom channel rules in their actual order and inspect the exact report dimension. If the information was never collected, assigning a familiar channel name does not reconstruct it.

A custom group selected in one report and the property's Primary channel group are different reporting choices. Preserve the original report and rule configuration before changing either. The useful result is an explainable classification for a defined population, not simply a smaller Unassigned row.

Real questions expose the rule and report boundaries#

An older question about agency-managed advertising describes Unassigned and (not set) results. After creating custom groupings, the author asks how to avoid selecting them separately in every report. That follow-up narrows the task: creating a group does not establish which dimension a report uses.

A separate social-ad reporting question receives advice about UTMs and custom-rule order. The thread does not prove the cause of its reported discrepancy or establish a universal rule that Direct must always be last.

Exact creation and latest-activity dates were not verified from those accessible originals. They are historical questions with current diagnostic value, not evidence of a new platform incident.

Related guide

Trace channel rules in their actual order

Record source values, rule precedence and the report dimension you are checking.

Use the tracking QA checklist

Opens a practical guide. No email required.

1. Separate missing information from unmatched classification#

Google's (not set) explanation describes a placeholder used when information for a dimension was not received. Its causes depend on the dimension. The default-channel reference separately defines Unassigned as the result when no channel rule matches.

These can occur together, but they do not mean the same thing. A populated source/medium pair with an unconventional medium can pose a classification problem. An absent source needs an earlier collection or attribution investigation. A custom rule based on a known campaign value can be useful, but it should not be presented as recovering the missing source.

For one affected row, preserve:

EvidenceWhat to save
Exact dimensionDefault, Primary or named custom group, including scope
Underlying traffic valuesSource, medium, campaign and source platform where available
PopulationProperty, dates, timezone, metric, filters and comparisons
Group identityGroup name, identifier and ordered rule definitions
ObservationRetrieval time and whether the row changed after an edit

Use the missing-attribution guide when the underlying value is absent. Do not rename every unresolved row “Paid Social” merely because the organization was running social ads during the period.

2. Read the full dimension name before testing the rule#

Google's traffic-source scope reference distinguishes First user, Session and event-scoped acquisition information. A person first acquired through an advertisement can return through another route. Grouping that initial acquisition and grouping the returning session answer different questions.

Choose one task explicitly, such as: “For this completed week, classify Sessions by the session-scoped custom group using the recorded campaign context.” Keep that choice fixed during testing. Do not compare it with a user-acquisition or key-event-credit report and attribute every difference to rule order.

Google's Data API schema exposes separate custom-group dimensions by scope and group identifier:

Requested contextAPI dimension pattern
SessionsessionCustomChannelGroup:GROUP_ID
First acquisitionfirstUserCustomChannelGroup:GROUP_ID
Key-event attributioncustomChannelGroup:GROUP_ID

The same reference lists sessionPrimaryChannelGroup separately. A friendly heading reading “Channel” does not prove these fields are interchangeable. For API work, inspect the property's available metadata rather than copying another property's identifier.

A channel-label audit also does not validate a performance denominator. Keep Sessions, key-event occurrences, users and advertising clicks distinct. The key-event rate guide covers that measurement boundary.

3. Inventory conditions and precedence#

Google's custom-group guide says traffic enters the first matching channel in the group's order. Therefore, testing each condition separately is insufficient: two correct individual conditions can overlap, and the earlier one wins.

Write a compact rule specification before editing:

Rule itemDiagnostic question
InputWhich recorded dimension does this condition read?
MatchExact, list, contains or regular expression?
LogicWhich conditions must all be true, and which are alternatives?
OrderWhich earlier rule can also match this input?
Intended resultWhich channel should this known test row receive?
Unknown valuesDoes the rule have evidence, or is it guessing?

The Admin API resource reference describes grouping rules, logical filter expressions, exact/list matches and full versus partial regular-expression matching. Its string filters are case insensitive. Preserve the match type as well as the visible pattern; the same letters can mean different conditions.

For example, a partial match for paid can include the invented value unpaid. An exact match for paid would not. Likewise, “source is in an approved social list AND medium is an approved paid value” differs from allowing either condition independently. Those are examples of logic to test, not a replacement for Google's complete default-channel definitions.

Where an authorized integration already exists, Google's channel-group lookup method provides read-only retrieval of one group. An ordinary configuration export or recorded UI inspection can serve the same inventory purpose. This guide does not require creating credentials or changing a property merely to inspect its current rules.

4. A hypothetical first-match audit#

Assume a deliberately simplified, disjoint population of 100 invented sessions. Each row below represents an aggregate slice, and each session belongs to exactly one slice. The values are supplied for the example; none describes an AdProof customer or Google's complete classification engine.

SliceRecorded sourceMediumCampaignSessions
Afacebookpaid_socialspring_launch30
Bbuying_deskcpmspring_launch20
Ccommunity_sitereferralcommunity40
DMissingMissingMissing10

The proposed custom group has these invented rules:

  1. Launch: campaign contains launch.
  2. Paid Social: source equals facebook AND medium equals paid_social.
  3. Display: medium equals cpm.
  4. Referral: medium equals referral.

With that order, A and B both enter Launch before their specific channels are reached. Launch receives 50 sessions, Referral 40, and 10 have no matching rule. Paid Social and Display receive none of the fixture's sessions.

Now suppose the reporting owner confirms that specific paid channels should take precedence over the broad campaign label. Reorder Paid Social and Display before Launch, keeping the same conditions and input population:

Output channelBroad rule firstSpecific rules first
Launch500
Paid Social030
Display020
Referral4040
No matching rule1010
Total100100

Exactly 50 ÷ 100 = 50% of the fixture's sessions move between labels. No sessions are created or recovered. The unresolved ten remain unresolved because the example supplies no evidence for another classification.

The alternative order is correct only for the stated business specification. If “Launch” is intentionally the highest-priority category, its first position may be appropriate. Do not substitute an arbitrary order recommended in a forum for an agreed reporting requirement.

This first-match exercise was executed locally with the stated conditions. It is not a GA4 account test. Validate the actual saved group, including copied channels and all remaining rules, on an authorized representative report before accepting a change.

5. Distinguish historical custom reports from future Primary values#

The custom-group documentation distinguishes two behaviors: a selected custom group can be applied retroactively in reports, while making it Primary populates Primary-dimension values using the chosen criteria from that point forward.

Treat the questions separately:

Reader questionEvidence to compare
How would old traffic classify under this custom definition?Selected custom group, fixed historical period and saved rule version
What did our Primary channel reporting record across a change?Primary dimension, effective configuration time and comparable report extracts
Why does an overview still show the old categories?Actual dimension referenced by that card or report

Do not promise that changing Primary rewrites every historical default-channel report. Google's Default channel group remains a separate predefined classification. Inspect reports that explicitly reference it instead of assuming their dimension changed with the property's Primary selection.

For the agency question, first demonstrate the custom group's result in a supported acquisition report. Then identify which overview or snapshot the owner needs to maintain. Save the exact card configuration and field name. A report's primary dimension means its leading grouping field; that phrase alone does not establish that it uses the property's Primary channel group.

Prepare the proposed setting change with its effective time and acceptance checks. Re-run the chosen reporting comparison after an approved change, rather than judging success only by the disappearance of an unfamiliar label.

6. Keep import and connector boundaries visible#

The retained BigQuery custom-channel question from October 27, 2023 asks where an already configured GA4 group appears in exported data. Current custom-group documentation states that custom channel groups are unavailable in the BigQuery export schema. Report/API availability does not establish an exported field.

If a warehouse creates its own classification, document its input scope and versioned rules separately. Do not copy a first-user source field into a session classifier simply because its name looks suitable. This is a separate transformation whose equivalence needs evidence.

A historical connector failure also does not justify deleting a group to make a dashboard load. Preserve the failing request or metadata, connector version and affected identifier; separate connector compatibility from the group's intended reporting definition.

For advertising cost comparisons, keep the import join distinct from the channel label. The Meta campaign-cost import guide explains why a clean category does not prove source, medium, date and campaign keys actually joined.

Common mistakes and acceptance checklist#

Avoid broad rules that unintentionally capture later categories, mixing scopes, treating a renamed label as recovered traffic, or interpreting historical reclassification as a performance improvement. Do not assume an API field name or custom-group identifier belongs to every property.

Before accepting the configuration:

  • Retain the original ordered rules and report export.
  • Document the intended population, scope and business category definitions.
  • Test positive, negative, overlapping and missing-value rows.
  • Verify the first winning rule for each test.
  • Reconcile total eligible population separately from movement between labels.
  • Record the selected report dimension and any Primary effective time.
  • Recheck downstream reports, imports and connectors under their own contracts.

The fashion-retail display case illustrates establishing a delivery baseline before interpreting performance. Its measurement does not establish a channel-classification repair or a sales outcome.

Contact AdProof for a scoped campaign-delivery review using supported exports with selected campaign, dates, timezone, dimensions and metrics. Keep that delivery evidence beside the explicitly defined GA4 classification report. This workflow does not imply an AdProof channel-group integration or automatic agreement between impressions, clicks and sessions.

Frequently asked questions#

Will a custom group recover missing source or medium values?

It can classify using information that exists under its rules. It does not recreate information that collection or attribution did not supply. Preserve unresolved values and investigate the missing dimension separately.

Can custom-group edits and Primary changes affect history differently?

Yes. Selected custom-group reporting can apply definitions to historical data, while the documented Primary criteria populate values going forward from the change. Record the exact dimension and configuration interval before interpreting a trend.

Does the first matching rule always represent the best category?

It represents precedence, not business correctness. Test overlapping conditions against the intended category specification. A broad campaign rule may intentionally win, or may unintentionally hide a more specific paid channel.

Will a GA4 custom group automatically appear in BigQuery or every connector?

No. Current documentation excludes custom channel groups from the BigQuery export schema. Other reporting paths need their own supported dimension, property identifier and compatibility checks. A configured group alone does not prove end-to-end connector support.

Your next step

Trace channel rules in their actual order

Record source values, rule precedence and the report dimension you are checking.

Use the tracking QA checklist

Opens a practical guide. No email required.

Back to all field notes