Skip to main content
Most journeys that misbehave are not broken. They are doing exactly what they were built to do, to a set of users the person who built them did not picture — the ones who matched two branches, the ones who matched none, the ones still waiting in a delay when the journey was paused. This page is about the decisions that settle all of that, taken before you publish rather than after a campaign has run.

What this is

Guardrails for designing a journey, the design mistakes that generate support tickets, and a checklist to run before you publish. It covers the shape of the workflow: who enters it, how the paths divide, and what happens to the people already inside it when you change something. It does not explain how to configure an individual node — each node’s own page does that, and Create a Workflow covers the build end to end. If a journey is already live and something has failed, start at Identify Journey Errors and look the failure status code up in Journey Error Codes.

Settle the entry condition first

Publishing permanently locks two things: the workflow name and the entry condition. Everything else stays editable on a live workflow. Changing either one afterwards means duplicating the workflow and publishing the copy — there is no way to unlock them in place.
A published journey starts from the moment it goes live. It evaluates qualifying events arriving from then onwards and never goes back for the users who qualified yesterday, so an entry condition that was nearly right costs you every user who passed through the window it ran in. Two things to resolve before you publish. Confirm the source actually feeds journeys. A workflow only reacts to events from a real-time streaming source. Events ingested through a batch source are stored against the user’s profile and are visible there, but they are never evaluated against an entry condition. The result is a silent nothing: the workflow publishes cleanly, raises no error, logs no trace and activates nobody. If you need a batch source to trigger a journey, contact Zeotap Support and ask for that source to be enabled for Journeys, giving your organisation ID and the source ID. Enabling it is not retrospective — events that arrived beforehand are not replayed — so re-ingest the file once Support confirms. See Add an Entry Condition. Filter at the entry rather than downstream wherever you can. Add a Condition has a single output: users who fail it are eliminated at that node and there is no second path to catch them. The canvas shows only how many records reached the node, not which of them failed it, so accounting for the drop-out means tracing users one at a time. A test you can express in the entry condition keeps those users out of the journey altogether, which is both easier to reason about and easier to count.

Keep branch logic flat and complete

A condition box holds up to ten conditions, but nesting under Have Certain Attributes / IDs multiplies the real total well past ten. An over-nested box is a common cause of a publish that fails with a generic error naming nothing in particular. Flatten it: split the logic across two condition nodes, or move part of it into the entry condition. A multi-branch split behaves unlike every other branching node on the canvas, and most of the trouble it causes comes from that.
  • Branches are not mutually exclusive. Each branch condition is evaluated on its own, so a user who matches three branches enters all three — and if more than one of those branches ends in a send, that user is sent once from each. Write branch conditions as ranges that cannot both be true, or use an Apply True/False Split where exactly one of two outcomes must apply.
  • There is no catch-all. A user who matches no branch goes no further; the journey ends for them at the split and nothing downstream runs. No error is raised. Either add a branch that covers the gap or accept the gap knowingly.
  • Every branch needs a complete condition. The workflow will not publish while any branch is missing one.
  • A branch cannot be removed on its own. Deleting the split deletes every branch and every node beneath it, and that cannot be undone. Settle how many branches you need — a split supports two to five — before you build them out.
See Apply Multi-branch Split for the configuration itself.

Name the nodes as you build

A journey of any size is unreadable without names: six condition nodes all read Performed an event on the canvas and can only be told apart by opening each one. Where you leave a node unnamed, Zeotap generates a label — condition_node1, delay_node2, destination_node4 — and the number counts nodes across the whole workflow rather than nodes of that type, so delay_node2 is the second node in the journey that happens to be a delay, not the second delay. A generated label tells you nothing about what the node checks or where it sends. Two limits are worth designing around. A node name appears on the canvas and nowhere else, neither on the metrics view nor in traces, so when you debug you are still identifying nodes by their position on the canvas. And the name strip is narrow: a long name is cut off with an ellipsis and there is no tooltip to reveal the rest. Keep names short and specific, “Abandoner check” rather than a sentence. See Name a Node.

Common anti-patterns

Treating pause as a safe no-op

Pausing stops new users qualifying, and it also removes everyone already inside the journey. They are no longer activated, and resuming does not bring them back: a resumed journey considers only records arriving after the resume, which makes it as good as starting from scratch. Per-node counts are hidden while a journey is paused and return intact when you resume. Archiving costs more. The counts are deleted outright, anyone waiting in a delay exits the journey, and un-archiving returns the workflow to Draft, so it has to be published again before it processes anyone. If the people already inside a journey must finish it, neither action will hold them — let the journey run. See Explore Journeys Listing Page Options.

Editing a live journey without deciding what happens to the people inside it

Republishing a live workflow raises a confirmation with two options. Neither is pre-selected and the confirm button stays disabled until you choose, so decide which you want before you open it. The same choice appears when you pause or archive a live workflow, so pausing is also a point at which re-entry history can be cleared.

Deleting a delay node to speed a journey up

Everyone waiting inside that delay is removed from the journey. They are not forwarded to the node that followed it and they are not picked up again when you republish. Cutting a connector with the scissors icon is more destructive still: it deletes the node below the cut and every node after it in that branch. Long waits have their own quirk. A delay of 10 days or more is released by a daily job rather than by each user’s own timer, so treat a long delay as accurate to the day rather than to the minute. See Add a Delay and Add or Delete a Node.

Expecting the re-entry field to express a calendar rule

The re-entry time says one thing only: wait this long after the user exits. Calendar-shaped rules — once per calendar day whatever time yesterday’s activation happened, or a reset at a fixed local time so that everyone becomes eligible together — cannot be set in the panel. Zeotap can configure a rule of that shape on your journey, so raise the request with Support early and build the lead time into the campaign plan. Two consequences follow for anything Support sets up. The rule is held against the journey rather than shown in the settings panel, and republishing after an edit can clear it, so ask Support to confirm it is still in place once your change is live. And because the reset time is held in UTC, the local time it fires at shifts when your clocks change — a follow-up request is needed at each daylight-saving transition. See Add Re-entry Criteria.

Treating re-entry and concurrent entry as one setting

They are different axes, and the panel makes them easy to confuse because concurrent entry only becomes available once re-entry is set to Yes. With concurrent entry off, which is the default, a trigger that fires while the user has an active run is dropped — not queued, not replayed when the run ends. Turn it on and the same user can receive several messages from the same journey at once — that is the setting working as designed, and it is the usual cause of duplicate sends. Choose on evidence: enable it when every event genuinely must be forwarded independently; leave it off for onboarding, re-engagement and suppression journeys. See Add Concurrent Entry.

Designing sends without checking the organisation’s frequency policies

A Gateway policy caps how many messages one customer can receive across every journey in the organisation, and it is enforced automatically at the moment of send. There is no Gateway node and nothing to configure on the canvas, so a journey can be perfectly built and still not deliver. A capped send is blocked for that one customer only, while everyone else still under their cap is messaged as normal. Nothing is queued and nothing is retried, so the message is lost rather than delayed, and an error is recorded against that customer on the Send to Destinations node. Gateway caps the total; it does not rank journeys. If a customer qualifies for two journeys on the same day and the cap allows one message, whichever send reaches the destination first takes the allowance.

Pre-publish checklist

The builder blocks a publish until five things are true. Save as Draft insists on only the first of them, and not its uniqueness rule — a draft saves on a name alone, and a workflow left in Draft never evaluates an event however finished the canvas looks. Nothing checks the following, and each one has cost somebody a send:
  • Does any user match two branches of a split, and should they?
  • Does any user match none of them, and is that intended?
  • Is the source named in the entry condition a real-time streaming source that is still live?
  • Is the re-entry time in the unit you meant — 2 Hours rather than 2 Days? The workflow header above the canvas reports the value and its unit.
  • Does the same header read Concurrent entry: Enabled when you did not intend it to?
  • Does a mandatory Journey Template apply to this folder? It renders as the top node of the canvas and is read-only, and because templates are referenced rather than copied, an edit to one reaches every live workflow using it.
Then verify on the day you publish rather than later.
  1. Check the listing row reads Published, and that Activated Users starts to climb. If the per-node counts climb while this stays at 0, users are entering but not reaching a Send to Destinations node.
  2. Read the counts on the canvas with their caveats in mind: a count is the records that reached that node to be evaluated rather than the records that passed it, it counts events rather than unique users, it covers the last 7 days, and the delay node reports none at all. See Understand Metrics.
  3. Follow one real user through Journey Traces and confirm they took the path you designed. Traces reach back one day at most, so a user who entered the day before yesterday is already out of range — test and check within the same day.

Next steps

Last modified on November 20, 2026