First, find where the flow loses people
Put four counts side by side for the same weeks: the trigger events the site or shop recorded, the people who entered the flow, the messages the flow sent, and the messages accepted by receiving servers. Each gap points at a different part of the system. If the site recorded sign-ups or orders but the flow shows almost no entrants, the event is not reaching the flow or the entry rules reject it. If people enter but few messages leave, something inside the flow removes or holds them: an exit, a filter on a consent value, a suppression, or a rule giving another flow precedence. If messages leave but are not accepted, the problem has moved from automation to the receiving side.
Then compare against a time when the flow worked. A sudden collapse on a particular day usually lines up with a change: a new site template, a shop platform update, a moved sign-up form, an edited flow, or a switch of email platform. A slow fade points more often at rules that tighten as the list changes. The automation hub describes every flow as three decisions, a trigger, timing and exits, and that framing is the map for the table below.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Sign-ups or orders recorded on the site, almost no entrants in the flow | The event name and properties arriving in the platform against the trigger the flow listens for | The trigger event renamed or no longer sent after a site or platform change | Point the trigger at the event as it now arrives, or restore the old event |
| Entrants fell steadily while the event volume held | Each entry condition and exit rule, tested against a handful of real recent contacts | Entry or exit rules that exclude almost everyone | Rewrite the rule so it matches the people the flow exists for |
| Nothing at all since a particular date | The flow's status, its version history and whether it exists in the current platform | A flow left off, in draft or never rebuilt after a migration | Switch on or rebuild the flow, with its exits, and test it |
| People enter and leave without a message, often only some sources | The consent field and suppression status on contacts who entered and got nothing | Contacts missing the consent field the flow filters on, or suppressed | Map the consent value from every source; confirm suppressions are correct |
| Entrants wait in the flow, then exit untouched | Which other flows the same contacts were in, and the caps applied that week | A priority rule or frequency cap holding people in another flow | Set priorities and caps deliberately so the flow is not always the one that yields |
| The flow reports sends, but deliveries fell and refusals cite authentication | Bounce reasons, and whether signing or the sending domain changed | Messages sent but refused after an authentication change | Restore signing on the business's own aligned domain |
Fix, verify and prevent each cause
A trigger that no longer fires
Most flows listen for an event with an exact name: a form submission, an account created, an order placed. When a site redesign, a new shop plugin or a change of checkout sends the same moment under a different name, or stops sending it at all, the flow waits for something that never comes. The welcome series tutorial defines the trigger as the sign-up itself, received at the moment it happens, and warns that a batch import is a delay rather than a trigger.
Fix: compare the events the platform now receives with the one the flow listens for, then either reconnect the trigger to the new event or restore the old one at the source.
Verify: submit a real sign-up or test order from each source and watch the contact enter within minutes.
Prevent: add "which flows listen for this event?" to the release checklist for any change to forms, accounts or checkout.
Rules that shut nearly everyone out
An entry condition written for last year's data can quietly exclude most of this year's people: a filter on a property the new forms no longer set, a "not a customer" test that now catches everyone because the shop marks subscribers as customers, a minimum basket value nobody revisited. Exits do the same from the other side, removing people before the first message. The abandoned cart flows tutorial sets out entry rules (identified, contactable on a recorded basis, not a recent buyer of the same item) and exits matched on the person, and both can be too strict as easily as too loose.
Fix: test each rule against a small set of real contacts who should qualify, and loosen or rewrite the condition that rejects them.
Verify: those contacts now enter and receive the first message, and an exit still fires for a test purchase.
Prevent: write every rule down in plain words beside the flow, so a change in the data model is checked against it.
A flow left off or never moved
Flows get paused for a sale, a stock problem or an edit, and stay paused. After a change of email platform they may simply not exist: the tutorial on migrating to a new email platform notes that a form still posting to the old account means the welcome series never fires, and asks for each flow to be rebuilt one at a time from an inventory of its trigger, timing, branches and exits.
Fix: list every flow the programme is meant to run, confirm each is live in the platform that now sends, and rebuild any that are missing from the written register.
Verify: each live flow shows entrants and sends for the same days as its trigger events.
Prevent: keep that register current, and when a flow is paused, record who paused it and the date it should restart.
Contacts without the consent value the flow requires
A flow that filters on marketing consent sends only to contacts carrying that value. If a new form, an integration or an import creates contacts without it, they enter and drop out silently. Where Double Opt-InA sign-up process in which the subscriber confirms through a link sent to the address given. is used, entry waits for the confirmation click, so confirmations that never arrive stall the flow as well. The list building and consent guide asks for the source, timestamp, wording and choices to be stored for every address, which is exactly what the filter reads. Contacts on the Suppression ListA list of addresses or contacts excluded from every send, such as unsubscribes, hard bounces and spam complaints. are excluded by design.
Fix: map the consent value from every source that creates contacts, and confirm that suppressed contacts are suppressed for a real reason.
Verify: new contacts from each source carry the field, and the share of entrants who receive the first message returns to normal.
Prevent: make consent capture part of any new form or integration's acceptance test, never an afterthought.
Another flow taking precedence
Priority rules and frequency caps exist so a person in two flows is not mailed twice in an hour, and the automation hub asks for the winner to be decided in advance. When a new, busy flow is given top priority or a cap is tightened, a quieter flow can find that its entrants are always held or dropped.
Fix: list the flows a held contact was in, and decide which should win for that combination; consider delaying rather than dropping the losing flow.
Verify: the share of entrants who wait and exit untouched falls without a rise in people mailed twice the same day.
Prevent: review priorities whenever a flow is added, not only when one stops.
Messages refused after an authentication change
Sometimes the flow is working and the receiving servers say no. A change of sending domain, a platform that signs with its own domain, or an edited DNS record can leave mail unauthenticated or out of alignment. The deliverability guide explains that Google's email sender guidelines document authentication for all senders, and for bulk senders both SPF and DKIM, a DMARC policy and alignment with the From domain, as reasons mail can be rejected.
Fix: restore signing on the business's own domain and check alignment.
Verify: bounce reasons citing authentication stop and accepted messages match sends again.
Prevent: treat any change to sending domains as a deliverability change with its own test sends. Refusals for other reasons belong to diagnosing a high email bounce rate, and accepted mail that lands in junk to emails going to spam.
A prevention routine for unattended flows
Campaigns fail on the day someone sends them; flows fail on a day nobody is watching. The routine that catches them is mostly a habit of looking. Make entrants and sends per flow a named KPIA metric tied directly to an objective. for the automation programme, read weekly beside the trigger events that feed them, so that a gap opening between the event count and the entrant count shows up within a week rather than a quarter. Give every flow an owner who is told about changes to the site, the shop platform or the sending setup before they ship.
Before any release that touches forms, accounts or checkout, run one real sign-up and one test order through every source and confirm the flows they should start actually start. After any change to a flow, run it again through the email automation flow launch checklist instead of trusting that a small edit changed nothing; the checklist exists for launches, and an edited flow is a relaunch. Keep a written register of every flow, with its trigger event name, entry and exit rules, priority, status and owner, so a migration or a new hire starts from a list rather than from memory.
Finally, decide in advance how a broken flow is announced and who restarts it. A shared note recording when a flow was paused, why, and when it should resume prevents the commonest cause of all, a flow that someone switched off for a good reason and nobody switched back on. The email troubleshooting section lists the neighbouring symptoms, such as falling opens or rising unsubscribes, that a flow can show once it sends again.
Grumbletuck Coffee Roasters finds its flows waiting for old events
Grumbletuck Coffee Roasters, an invented roaster selling beans by subscription and one-off order, moves its shop to a new platform in March. Three weeks later the team notices that welcome emails have stopped. Lining up the counts for those weeks shows 1,140 newsletter sign-ups recorded by the shop, 96 entrants into the welcome series, and 41 welcome emails sent. The new shop sends the sign-up event as "Subscribed to newsletter"; the flow still listens for "Newsletter signup". Reconnecting the trigger restores entrants, but sends stay at about 40% of entrants. Following one contact shows why: the new integration does not pass the marketing consent field, so the welcome's consent filter drops everyone created since the move. The replenishment flow has the same fault, listening for an order event that was renamed. After both are fixed and the consent field is mapped, the team sends a single dated welcome to the 860 people who signed up in the gap and are still subscribed, and skips the replenishment reminders that were already out of date. Entrants and sends now track sign-ups week by week on the dashboard.
Common mistakes
Rewriting the emails when nothing is reaching them. Reading sends without the trigger events beside them. Assuming a migration moved the flows because it moved the contacts. Loosening a consent filter to make the numbers recover. Backfilling everyone who missed a broken flow, including people who have since bought.
