First, see which source and medium pairs are unassigned
Open Traffic acquisition with Session default channel group as the primary dimension, then add Session source / medium as a secondary dimension and filter to the Unassigned row. That list is the whole diagnosis. Read it against the default channel rules, which the help centre documents and the attribution settings guide summarises in a table: each channel is a rule over source, medium and sometimes campaign, so a pair either fits a rule or it does not.
Check the scope as well. The session channel and the first-user channel answer different questions, and a problem that appears only in one of them narrows the search: an Unassigned first-user channel points at how new people arrive, an Unassigned session channel at how all visits are tagged. Finally, sort the pairs by sessions. The top two or three nearly always cover most of the row, and each usually traces back to one team or one sending system. If the row also shows a large "(direct) / (none)" share elsewhere, the sibling page on direct traffic that is too high covers links that lost their tags entirely.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Unassigned pairs with a sensible source and an invented medium, such as a platform or newsletter name | The medium against the controlled vocabulary and the default rules | utm_medium values the default channel group does not recognise | Retag with a medium from the documented vocabulary |
| Pairs showing a real source with "(not set)" medium, or a medium with no source | The landing URLs that produced them | utm_source without utm_medium, or the reverse | Always send source, medium and campaign together |
| Pairs of "(not set)" on events that never had a landing page | Whether the events come from a server or the Measurement Protocol | Events sent without session information | Send the session and client identifiers with server-side events |
| Unassigned appears only in a custom channel group, not in the default | The custom group's rule order and its fallback | Custom channel group rules leaving gaps | Add rules for the missing pairs, ordered from specific to general |
| Ad clicks landing with incomplete source or medium | Auto-tagging status, the account link, and the landing URLs | Ad platform traffic arriving without auto-tagging or tags | Enable auto-tagging where supported; tag every other platform's links |
Fix, verify and prevent each unassigned pattern
A medium the rules do not know
The default rules read the medium above all. A medium written as a platform name, a campaign type the team invented or a word such as "newsletter" fits no rule and lands in Unassigned, even when the source is perfectly clear.
Fix: retag the links with a value from the controlled medium list in the UTM and naming conventions guide, and map each value to the channel it should produce.
Verify: click a corrected link and watch Realtime report the expected channel for that session.
Prevent: generate links from a shared builder whose medium field is a fixed drop-down, so invented values cannot be typed.
Half a tag
A link carrying utm_source but no utm_medium gives a source and a "(not set)" medium; a link with a
medium and no source does the reverse. Neither can satisfy a rule that needs both, so the session drops into
Unassigned. This is common with partner links, QR codes and hand-edited email footers.
Fix: send source, medium and campaign together on every tagged link, as the five-parameter table in the UTM and naming conventions guide sets out.
Verify: the pair disappears from the Unassigned breakdown for sessions after the change.
Prevent: have the builder refuse to produce a link with any of the three fields empty, and include a link test in the launch checklist.
Events without a session
Events sent from a server or through the Measurement Protocol arrive without the browser context that carries source information unless the sender includes the identifiers that tie them to an existing session. They then report with session-scoped traffic dimensions unset, which the channel group cannot classify.
Fix: pass the client identifier and session identifier from the browser through to the server so its events join the right session; the server-side tagging basics guide explains how clients and event data carry those fields.
Verify: server-sent events in DebugView sit inside the visitor's session, and the "(not set)" pair shrinks after release.
Prevent: add the identifier fields to the server event specification and test them whenever the backend integration changes.
Gaps in a custom channel group
A custom group replaces the default rules with your own, evaluated in order, first match wins. A rule set built for the channels the team cared about at the time leaves later sources unclassified, and anything that matches nothing falls into Unassigned within that group only.
Fix: add rules for the missing pairs, ordering them from most specific to most general, and end with a deliberate catch-all if the business wants one; the attribution settings guide covers how custom groups are built from a copy of the default.
Verify: compare the custom and default groups side by side for the same period and confirm the gap closed.
Prevent: review the custom group whenever a new channel or partner launches, and keep the number of groups small enough that someone owns each one.
Ad clicks without usable tagging
Some platforms pass campaign detail only when the account is linked and auto-tagging is on; others pass nothing unless links carry manual tags. Ad clicks that arrive with neither can end up with incomplete source or medium information and fall outside every rule, or into the wrong channel.
Fix: link the Google Ads account and enable auto-tagging as described in the Google Ads integration tutorial, and use the URL parameter templates on other platforms with the convention's vocabulary.
Verify: click a live ad from a clean browser and confirm the session's source, medium and campaign in Realtime.
Prevent: keep a per-platform tagging rule in the convention and check one ad link per platform at every launch.
Prevention: tagging rules everyone uses
Unassigned grows when more people build links than know the rules, so prevention is mostly about removing the chance to improvise. Keep one convention document with a short medium list, each value mapped to the channel it should produce. Build every link from a shared builder with fixed fields, and log what it generates. Review the Unassigned breakdown monthly, alongside the campaign report, and treat each new pair as a conversation with the team that created it rather than a silent clean-up. Give server-side integrations a written field list that includes the session identifiers. Where a custom channel group exists, give it an owner and a review date. Channel reporting, and every AttributionAssigning conversion credit to touchpoints. report built on it, is only as good as this discipline: a model can credit only the channels it can recognise. The UTM link request SOP turns the builder into a request process for teams that do not build links themselves.
A charity's Unassigned row, traced pair by pair
A charity sees Unassigned at 11% of sessions in a quarter. Breaking it down by source and medium shows three
pairs covering almost all of it: mailer / newsletter with 2,900 sessions, partner-site / (not set) with
1,100 and (not set) / (not set) with 760. The first is the email team's own medium, which no default rule
recognises; it is retagged as email. The second is a partner's footer link carrying only a source; the
partner is sent a complete link from the builder. The third comes from donation receipts sent server-side
without the session identifier, which the developers add. The next month Unassigned falls to 2%, and a
custom channel group places the quarter's historical newsletter sessions under Email for the trustees'
report.
Common mistakes
Fixing the report with a custom group and leaving the links broken. Diagnosing with first-user channel when the question is about sessions. Adding a new medium without mapping it to a channel. Tagging internal links, which restarts attribution mid-visit. Assuming a UTMURL tags identifying traffic source. value is case-insensitive.
