Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

CRM & Marketing Automation › Troubleshooting

CRM Workflow Not Running

Why a CRM workflow stops enrolling or acting on records: values spelled differently, records that bypass the trigger, suppression, a paused clone and a leaver.
Troubleshootingintermediate15 min readUpdated

First, read the run as trigger, condition, action and exit with one test record

Start from one record that should have been handled and was not: an enquiry that was never assigned, a contact whose region field was never set, a deal whose stage change raised no task. The workflows hub describes every workflow as four parts, and a silent failure is a failure in one of them. Open the record's enrolment history and ask the four questions in order. Did the event that should have started a run actually happen to this record, in the system the workflow watches? If it did, did the record pass every condition as the workflow spells it? If it entered, did each action run, and what did it do? If it ran, is the result still there now, or has something changed it since?

The answers narrow the diagnosis faster than anything in the workflow's settings. A record that never entered points at the trigger or the conditions. A record that entered and produced nothing points at an action. A record that entered, produced the right result and now shows the old value points at a second writer. The five whys method carries the first finding back to its cause: the lead was not assigned because it never entered; it never entered because the county did not match; it did not match because the new form writes free text where the old one wrote a picklist value; it writes free text because the agency rebuilt the form from a template; the template was never checked against the workflows that read the field. The fix lives at the last why, and the prevention lives one step further back.

Two readings save time before the table. First, compare the enrolment count for the last period with the same period before the problem started; a count that fell to nothing on one day usually coincides with a change that can be found in the change log, the form's history or the leaver list. Second, check whether the workflow runs for some records and not others. A workflow that still handles web form submissions but ignores imported and synced records is telling you where the fault is before you open anything else.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Records from one form or source stopped entering on the day that form or source changedThe raw value on a missed record against the value the condition expectsThe enrolment criteria name a value the record never carries as spelled, so nothing matchesStandardise the value into its own clean field and route on that, with unmatched values sent to a review queue
Web form records enter; imported, synced or bulk-updated records never doWhich event the trigger listens for, and whether that path raises it or fills the field at creationThe record arrived by import, sync or bulk update on a path that never fires the trigger, or with the field blank at creationTrigger on the change that every path makes, or enrol the batch deliberately after a sample
Returning contacts never enter a second timeThe record's enrolment history for an earlier runThe contact was enrolled once before and the workflow allows one enrolment per recordSet re-enrolment deliberately for the workflow's job
Records enter and leave in the same moment with no actionThe entry conditions against the record's customer, opportunity and consent fieldsA suppression or customer condition at entry removes the record before any action runsKeep the gate, and fix the field or the copy of the field it reads
Nothing has entered since a date nobody can explainThe workflow's status and the inventory's change log for that weekThe workflow is paused, left as a test clone, or was edited without a change log so nobody can say what changedRestore from the version before the edit, retire the clone, and log every change
Records enter, the assignment or alert step runs, and nobody receives anythingThe user, owner or recipient each action namesAn assignment, alert or sender step still points at a user who has left, so the run fails or reaches nobodyMove the slot, repoint alerts to the owner field and deactivate the user last
The action ran but the field shows the old value againThe record's history for who wrote the field after the workflowA second workflow or the sync writes the same field back, undoing the first workflow's actionGive the field one owner and make the other writer read-only

Fix, verify and prevent each cause

The criteria name a value the record never carries as spelled

Routing, scoring and segment conditions read a handful of fields, and the tutorial on standardising CRM field values explains why those fields drift: each form, import, lead form and integration writes its own spelling of the same country, title or sector, and a routing rule that expects one spelling misses the rest. The damage is quiet, because the record falls through to a fallback owner or to nobody and no error is raised.

Fix: keep the raw value, write a standardised value to its own clean field from a mapping table built from the values already in the CRM, and point the condition at the clean field; send anything the table cannot map to a review queue with a fallback owner, so a lead is never stuck because its county was spelled a new way.

Verify: re-submit the missed record's values through the form or the import path and watch it enter; then read the review queue after a day to see which spellings still arrive.

Prevent: fix the spelling where it enters by giving every form you control a picklist, check the field mapping of every lead form and integration that writes the field, and keep the mapping table beside the field documentation so a reviewer can extend it without editing the workflow.

The record arrived by a path that never fires the trigger

A workflow built on a form submission watches for that event, and the tutorial on importing contact lists is explicit that an import is a trigger of a different kind: it creates many records at once, and each workflow either fires for all of them or for none, depending on which event it listens for and whether the fields it reads are filled at creation or afterwards. A record created by a sync or a bulk update behaves the same way.

Fix: decide what the workflow should do with records that arrive without the event, then either trigger on the change every path makes, such as the clean field being set, or exclude the batch by its tag and enrol it deliberately after a sample has shown what the actions will do.

Verify: load a small tagged sample through each path, open each record and confirm it entered once and received the right action.

Prevent: list the workflows every import, sync and bulk update will meet before it runs, as the importing tutorial does, and record the decision for each one on the import.

The contact was enrolled once before

Many workflows allow one enrolment per record, so a contact who entered a sequence months ago and now does the same thing again is ignored. The nurture sequences guide treats re-entry as a rule to write rather than a default to accept: allow it after a minimum interval, route repeat entrants to a shorter variant, or require a decision. The same question applies to a routing or field-update workflow, where a returning enquiry usually should run again and a welcome task usually should not.

Fix: set the re-enrolment behaviour to match the workflow's job and write the rule beside the workflow in plain language.

Verify: take a record with an earlier run in its history, raise the trigger again and confirm it enters or is skipped as the rule says.

Prevent: make re-entry a line in every inventory row, so the next owner knows the behaviour was chosen rather than inherited.

A gate removed the record before any action

The automation governance guide describes the rails every workflow should carry: a global suppression check and the customer, open-opportunity and not-a-fit conditions that keep people out of sequences meant for someone else. A gate that works is not a fault, even when it is the reason a record did not enter; the fault is a gate reading a field that is wrong or a copy of a field that never updated. The governance guide also records the clearest failure of the opposite kind, a suppression list that stopped syncing and let customers into prospect sequences. When the gate failed that way and a customer was mailed as a prospect, the diagnosis for customers receiving prospect messages finds the gate, status or import that let them through.

Fix: keep the gate, and repair what it reads: the customer flag on the record, the consent field, or the copy of either that the marketing platform holds. Where the missing action is an automated email rather than an assignment, task or alert, the diagnosis for an email flow not sending covers the email platform's own entry rules, consent field, priority and authentication. Never remove a suppression check to make a workflow run, because the ICO's direct marketing guidance, as the site's guides state it, expects an objection to marketing to be honoured in every tool, and the GDPR gives people an absolute right to object to direct marketing, as the site's GDPR glossary entry states it.

Verify: a seed record with the right fields enters; a seed record marked as a customer or objector is removed at entry and stays out.

Prevent: keep both seed records permanently and check them after every change to the sync, the suppression list or the entry conditions.

The workflow is paused, a clone, or edited without a log

The governance guide names clones as the commonest source of orphans: a copy made for a test or a campaign, without a marker or an expiry, becomes the live version by accident or stays running by neglect, while the original was paused and forgotten. An edit made in a hurry has the same effect when nobody logged it, because the drop in enrolments cannot be traced to the change that caused it.

Fix: read the platform's edit history and the inventory's change log for the week the count fell, restore the version before the change where one exists, retire the clone with a reason and date, and reactivate the right workflow with a seed record in front of it.

Verify: the enrolment count returns to its earlier level and the seed record receives the actions in order.

Prevent: every workflow gets an inventory row before activation, every clone gets a test marker and an expiry, every edit gets a log line with the person and the reason, and the quarterly audit retires anything past its date.

A step points at someone who has left

The tutorial on reassigning CRM records when someone leaves lists what a leaver leaves behind: a slot in a round-robin group that keeps receiving new enquiries, alerts and notifications addressed to their inbox, and automated emails sent in their name. None of it raises an error. A routing workflow that assigns to a deactivated user, or an alert that goes to a mailbox nobody reads, looks exactly like a workflow that stopped.

Fix: take the leaver out of every assignment rule and round-robin group, name who receives what they used to, repoint alerts to the record's owner field rather than a fixed name, and re-sign templates that carry their name.

Verify: send a test lead through each affected route, as the lead routing QA checklist describes, and confirm it reaches a current person with a task and a notification.

Prevent: make workflow ownership and routing membership steps in the leaver process, deactivate the user record last, and run the leaver's filters again a week later for anything still pointing at them.

A second writer undoes the action

A field-update workflow can run perfectly and still appear to have done nothing, because another workflow or the sync wrote the old value back. The guide to syncing the CRM and the marketing platform sets the rule: a field two systems may write has no owner, and its value is whatever was saved last. The standardising tutorial adds the workflow version of the same fault, a standardising rule that overwrites a manual correction or triggers itself on its own write. If the history shows two records for one person, each holding its own value, the fault is a duplicate pair rather than a second writer, and the diagnosis for duplicate contacts in the CRM finds the entry point that created it.

Fix: read the record's history for who wrote the field after the workflow, give the field one owning system and one owning workflow, and make every other writer read-only or hand it a request field of its own.

Verify: change the value by the workflow, wait for the next sync and the next run of every other workflow that touches the record, and confirm the value holds.

Prevent: keep the field ownership table beside the inventory and treat a new writer of any shared field as a reviewable change.

Prevention: an inventory row, a seed record and a weekly volume check for every workflow

Every cause above was a change nobody connected to the workflow: a form rebuilt, a list loaded, a field renamed, a clone left running, a colleague leaving, a sync switched on. The governance guide's inventory is the record that makes the connection visible, and its row for each workflow should name the trigger and conditions, the actions and the user or field each one points at, the exits and the re-enrolment rule, and the dependencies: the fields, forms, lists, templates, integrations and other workflows it relies on. A dependency list is what lets whoever changes a form see which workflows to re-test before the change ships rather than after the leads stop. When the change is a whole new website and not one form, the tutorial on moving a website without breaking the CRM and its workflows builds that list for every form, trigger and link at once.

Seed records make the test cheap. Keep permanent test contacts that match each branch, as the workflows hub describes for a launch: one that should enter, one that should be gated out as a customer, one that should be gated out as an objector, one that arrives by import with the batch tag. After any change to a form, a field, the sync or the workflow itself, raise the trigger for each seed and read what happened in the enrolment history. The CRM fundamentals guide asks integrations to run as their own named users for a related reason: an audit log that shows which system changed a field is what turns a flip-flopping value from a mystery into a lookup.

Then watch the count. The governance guide's volume alarm, a workflow that enrols or sends far more or far fewer records than usual in a day, is where most quiet failures become visible first, and a scheduled report comparing each workflow's enrolments with its recent average does the job where the platform offers nothing. Read it weekly in minutes and act on the outliers the same day. Put the workflows that sales and marketing would both notice if they stopped at the top of that list: routing, the hand-off task, the customer flag on a won deal. Tie the rest of the routine to the governance guide's cadences, with a monthly spot check of suppression and consent conditions using the seed records, and a quarterly walk of the inventory that asks whether each workflow's purpose still holds, its owner is still present and its dependencies still exist.

Finally, keep the leaver process and the change process honest. A workflow's ownership transfers explicitly when someone leaves, the reassigning tutorial's handover covers routing, alerts and sender names in the same pass, and no edit to a live workflow ships without a log line and a seed test. When several of these habits are missing at once, the lead management audit scorecard rates capture, routing, follow-up and reporting in one pass and is the better place to start than a single diagnosis. The person who owns the lifecycle stage field, the person who owns the sync and the person who owns each workflow should all be named in that pass, because a workflow nobody owns is the one that fails silently for longest. When a symptom does not yet point clearly at a workflow, the marketing automation troubleshooting hub matches it to the page that owns it.

A tool hire firm finds why new enquiries stopped being assigned (illustrative)

Illustrative example — not a real result

Brimsolvay Tool Hire

Brimsolvay Tool Hire, an invented plant and tool hire company with 9 depots, notices that trade enquiries from its website have stopped reaching depot managers. The routing workflow shows as live and was last edited 14 months ago. The operations lead opens the enrolment history of 20 enquiries from the past week: 17 never entered the workflow, and 3 entered but produced no assignment.

The 17 share a cause. The web agency rebuilt the enquiry form 11 days earlier and the new form writes the county as free text, while the routing condition reads a picklist value the old form set; "N. Yorks" and "North Yorkshire" both fall through to a fallback owner who left the company in the spring. The 3 that did enter were assigned to that same former manager, whose round-robin slot was never moved and whose user record was deactivated, so the task went nowhere.

The firm maps the free-text county to a clean field with a review queue owned by the hire desk, moves the leaver's slot and repoints the alert to the owner field, and re-enrols the 20 enquiries in two batches with the introduction email switched off. It adds an inventory row for the routing workflow listing the form and the two fields it depends on, keeps a seed enquiry that is re-submitted after every form change, and sets a weekly report comparing enrolments with the average of the previous 8 weeks. The following month the queue receives 6 new county spellings, all from one partner's referral form, and no enquiry waits more than a day.

Common mistakes

  • Reading the workflow's settings instead of a record's history. The settings say what should happen; the enrolment history says what did.
  • Re-enrolling the backlog before fixing the cause. Weeks of missed records hit every owner and every sending step at once.
  • Removing a suppression or customer gate to make the workflow run. The gate was right; the field it reads was wrong.
  • Changing a form without listing the workflows that read its fields. The form works and the routing stops.
  • Deactivating a leaver's user before moving their slots, alerts and sender names. Everything pointing at them breaks at once.
  • Letting two workflows or a sync write the same field. The workflow runs and the value changes back.
  • Treating the fix as finished when the seed record enters. Without a weekly count, the next silent failure waits for a complaint.

Frequently asked questions

The workflow shows as live and nothing looks wrong in it. Where do I look first?

At one record rather than at the workflow. Pick a record that should have entered in the last few days and open its enrolment history. If it never entered, the fault is at the trigger or the conditions: the event did not reach the workflow, the record did not match the criteria as spelled, it was enrolled before, or a gate removed it. If it entered and nothing happened, the fault is at an action: a step that failed, a step pointing at someone who has left, or a later writer that undid the change. The workflow's own settings rarely explain a silent failure; the record's history usually does.

Should I just re-enrol everyone the workflow missed?

Not until the cause is fixed and you know what the actions will do to a backlog. A routing workflow re-enrolled across weeks of missed records sends every lead to an owner at once, each with a task due today, and an introduction email to all of them in the same minute. Fix the cause, test with one seed record, then decide deliberately which missed records should enter, in what batches, with the sending steps paused or the backlog handled by hand where that is kinder to the people on the receiving end.

How do I stop this happening again after a form or field change?

Treat every workflow as a dependency of the fields and forms it reads, and make the check part of the change rather than part of the clean-up. The inventory row for the workflow lists the fields, forms, lists and integrations it relies on, so whoever changes one of them can see what to re-test. A permanent seed record that is re-submitted after each change, and a weekly count of enrolments against the recent average, catch the cases the inventory misses.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Direct marketing guidance — Information Commissioner's Office (UK)
  2. Regulation (EU) 2016/679 (GDPR) — EUR-Lex

Related