First, prove the tag is really silent
"Not firing" is reported from three places, and only two of them are evidence. A report in GA4Google's event-based analytics platform. or an ad platform with no data is a consequence that could have many upstream causes, including a tag that fires perfectly and sends to the wrong destination. Tag Assistant attached to the published container is evidence about the live site. Preview attached to a workspace is evidence about a draft. Before diagnosing, decide which of the last two you are looking at.
Open preview on the exact page and perform the exact action from your test case. In the event timeline, select the event on which the tag ought to run, not the first page view, and open the Tags tab. A tag in the "not fired" list with a red condition has been evaluated and rejected; a tag that does not appear at all on that event was never in play, which points at the trigger type or at the event never arriving. Then confirm the outcome outside the debugger: in the browser's network panel, a request to the vendor's endpoint either leaves the page or it does not. The full tour of each pane lives in the preview and debugging tutorial; this page starts from what those panes show and works towards the cause.
Use the table in order. The top rows are the commonest and the quickest to rule out.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Tag in "not fired" with one condition marked red | The Variables tab on that event: the actual value against the condition's expected value | Trigger condition never matches: wrong case, a different URL form or an unexpected variable value | Rewrite the condition to match the real value, or normalise the variable feeding it |
| Tag fires on some events but the custom event it needs never triggers it | Event timeline order; when the push arrives relative to the trigger's event | Tag evaluated before the data layer push or event it depends on exists | Trigger on the custom event the site pushes once the data exists |
| All conditions green, tag still listed as not fired | The tag's exceptions section in the preview panel | A blocking trigger matches the same event | Narrow or remove the exception, or move the tag off the overlapping trigger |
| Tag held back on first page load, fires after the banner is accepted | The Consent tab and the tag's additional consent settings | Consent state or an additional consent requirement prevents it | Correct the default and update sequence, or the tag's consent requirement |
| Works in preview, silent for visitors | Live version number and contents; the environment the version was published to | Change left in a workspace, a different version live, or published to a staging environment | Publish the tested workspace to live and retest the published container |
| Only the first view of an app-style site triggers the tag | Whether a history change or route push appears in the timeline on navigation | Single-page app route changes produce no page load events | Have the router push a view event, or use the history change trigger deliberately |
Fix, verify and prevent each cause
A condition that never matches
Conditions compare strings exactly unless told otherwise, so a path written with a trailing slash, a hostname with or without the www prefix, a click text in a different case or a page URL carrying a UTMURL tags identifying traffic source. parameter all defeat an "equals" test.
Fix: copy the real value from the Variables tab on the triggering event and build the condition from it; prefer "contains" or a pattern where the value legitimately varies, and normalise messy inputs with a lookup or a small variable once rather than in every trigger.
Verify: the same event in preview now shows every condition green and the tag in the fired list.
Prevent: name triggers after what they match, reuse one trigger across tags instead of cloning near-duplicates, and keep the condition patterns in the reference the tags, triggers and variables guide recommends.
Data or event not there yet
A tag attached to a page view or DOM ready trigger fires as soon as that event is processed. If the value or the confirmation it depends on is pushed later, the tag either fires early with nothing to send or, when its trigger has a condition on that value, never matches.
Fix: ask the site to push a named event once the data exists and switch the tag to a custom event trigger on that name; for page context, move the push above the container snippet.
Verify: the timeline shows the push first and the tag firing on it, with the variable populated on that same event.
Prevent: write the trigger moment into the push contract described in the data layer guide, so "after the server confirms" is part of the specification rather than an assumption.
An exception wins
A blocking trigger written for internal traffic, a staging hostname or a debug cookie can match far more than intended, especially when its condition was loosened during testing.
Fix: read the exception's condition against the event's variables, tighten it to exactly the traffic it exists for, and remove it from tags that should never have carried it.
Verify: the tag fires for an ordinary visitor and stays silent for the case the exception was built to stop.
Prevent: keep each exception as a single reusable trigger with a name that states its purpose, and review where it is attached whenever a firing trigger changes.
Consent holds it back
Google's tags read the consent state themselves; other tags wait only if an additional consent requirement is set. Either can keep a tag silent: a default left denied with no update after acceptance, an update that arrives after the tag's trigger, or a requirement copied onto a tag that should run under the site's policy.
Fix: follow the default and update sequence in the consent mode guide, with defaults on the consent initialisation trigger, and set additional requirements per tag on purpose.
Verify: test the three states, no choice, refused and accepted, and confirm the tag behaves as the policy says in each.
Prevent: open the consent overview after every release and treat any tag without a deliberate consent setting as unfinished.
The change never reached visitors
Preview runs whichever workspace you opened. A fix made in one workspace, published from another, created as a version but never released, or published only to a staging environment leaves the live container unchanged.
Fix: confirm the live version contains the tag and trigger, then publish the tested workspace to the live environment; the version and environment mechanics are in the containers and workspaces guide.
Verify: attach Tag Assistant to the live site without preview and watch the tag fire there.
Prevent: one workspace per piece of work, version names that say what changed, and a rule that nothing is "done" until it is seen on the published container.
Route changes the container cannot see
On an app-style site, the first load produces the usual lifecycle events and every later view only changes the address. Page view triggers therefore fire once per visit and tags that rely on them fall silent.
Fix: have the router push an explicit view event with the new location and context, and trigger the tag on it; use the history change trigger only when the site cannot push, knowing it also fires on changes that are not views.
Verify: navigate between routes in preview and see one event and one tag firing per view, with the new path in its variables.
Prevent: add route changes to the push contract in the data layer guide and make sure enhanced measurement and a manual view tag are not both counting.
A prevention routine for container releases
Almost every silent tag was created on a release day: a trigger tidied, an exception added, a consent template updated, a site deploy that renamed a key. A short routine catches them while the change is still fresh. Write the test case before building: page, action, event name, tag expected. Preview every tag the workspace touches, plus the tags that share its triggers. Read the Errors tab even when the tag you came for looks fine. Publish with a version note listing what was added, changed and deliberately left alone. Then check the published container on the live site, and watch the destination's realtime or debug view for the events the release was meant to send. When a later alert fires, open the version list first, because the cause is nearly always in the newest version. The Tag Manager hub lists the governance pages that turn this routine into a team habit, and the sibling page on a data layer variable that reads as undefined picks up when the tag fires but carries nothing.
A demo-request tag that stopped after a redesign
A software company's demo form tag records 0 events in the week after a site redesign, against roughly 80 a
week before. Preview on the new form page shows the tag in the "not fired" list on the generate_lead
event, with one red condition: Page Path equals /demo/thank-you. The Variables tab reads /demo/thanks;
the redesign renamed the confirmation route. The analyst changes the trigger to fire on the custom event
alone, drops the path condition the push makes redundant, previews 3 test submissions, and publishes
version 112 with a note. Tag Assistant on the live site confirms the tag fires once per submission, and the
next day's report shows 14 leads.
Common mistakes
Debugging the page view when the tag is meant to fire on a later event. Trusting green conditions without reading the exceptions and consent rows beneath them. Treating firing priority as a dependency. Publishing a fix from the wrong workspace. Adding a second trigger to "make sure", then wondering why the tag fires twice.
