Investigate a journey enrollment

Use this when someone asks "why did Kari get this email", "why is Ola stuck", or "why did nobody from the webinar enter the journey".

Find the run

journeys_list to find the journey, then journeys_list_enrollments with contactId (find the contact with crm_search_contacts). Several rows mean several runs: the journey allows re-entry after a run ends.
journeys_get_trace on the enrollment.

Read the trace

Every node visit is one step, in order. Each step has the node, the edge taken and a reason:

branch steps say which rule matched (rule 2 matched) or no rule matched. Rules are checked in order, so an early broad rule hides later ones.
split steps record the bucket (0–99) the person hashed into; the same person always lands in the same path.
wait and wait_for steps with status waiting are where the person is now. openWaits shows when a timer fires or which event the run is waiting for and when it times out.
action steps show the result, or failed: … with the error. A step with status retrying will try again; permanent errors (unknown contact, missing webhook) do not retry and follow the node's onFail.
A step that throws an unexpected error (not an action's own failure) rolls back on its own, keeps the steps before it, and runs again after 1, 5, 15 and 60 minutes. The enrollment's stepErrors counts the attempts and lastError holds the message; after five in a row the run ends with step_failed and the step shows status failed. Steps a run was in when it ended show status cancelled.
The enrollment's exitReason says how the run ended: reached_end, goal_reached, matched_exit, idle_timeout, loop_detected, wait_timeout, action_failed, step_failed, journey_archived, holdout, ejected, or node_missing. idle_timeout means the run made no progress for exit.idleDays — the deadline moves forward every time it advances.

context.trigger holds the event that started the run; context.lastEvent holds the last event a wait_for received.

Someone never entered

Check, in this order:

The journey is live (journeys_get). Draft and paused journeys enroll nobody.
The trigger: for an event trigger, the event name must match exactly (analytics_list_product_event_names), and events older than entry.lateEventWindowHours are ignored.
The entry segment: the contact must match it and have a recorded lawful basis, must not be unsubscribed, and must not be marked do-not-contact.
Re-entry: with once a person enters at most once; with once_at_a_time not while a run is open.
The holdout: an enrollment with holdout: true entered on purpose and was held back to measure lift.

Someone is stuck on a step

If the trace stops at a step that never started and the run is active with no next run time, the step is paused — by a person or by the monitor on a guardrail breach. journeys_get lists pausedNodes with the reason; journeys_get_guardrails shows the rates behind a monitor pause. Resuming the step releases everyone held there at once. Messages from that step that were already approved are held too, not failed: they stay approved and are checked again every 15 minutes, as they are while the whole journey is paused.

The trace is empty

Step traces of runs that ended are deleted after settings.traceRetentionDays (365 days unless the journey sets 30–3650). The run itself stays, with its exit reason and goal, so reports still count it; only the step-by-step log and the run's context are gone.

What to do

A journey that behaves as designed but not as intended needs a new draft (skill://journeys/build-journey); people already in it finish on their version.
To take one person out, journeys_eject — but only when the operator asks you to.