First, find the message, the workflow and the customer's status in each system at the time of the send
Get the message before changing anything. The send log names the workflow or campaign that sent it and the time it went, and the message itself says what kind of workflow it came from: a nurture step, a demo invitation, a trial reminder, a first-order offer. The automation governance guide sets the incident routine that starts here: pause the workflow, identify from the send logs who else was affected, decide with the owner and, where relevant, legal or compliance colleagues whether a correction is needed, fix the cause, log the incident with the cause and the fix, and check whether the same cause exists elsewhere in the inventory. Pause first, because the workflow that mailed one customer is still running for the rest.
Then establish the customer's status in each system at the time of the send, not now. In the CRM, read the lifecycle stageA CRM field recording where a contact stands in the relationship, such as subscriber, lead, opportunity or customer. and its history: when it was set to customer, by whom or by which workflow, and the won date on the deal. In the marketing platform, read its copy of the stage and that field's history, and whether the record that was mailed is the record the CRM shows. Then search both systems for every record holding the person's address and every contact at the customer's organisation. The order of events narrows the cause before the table is opened. A deal won before the send, with the CRM saying customer and the marketing platform saying lead, points at the sync or at a closed-won process that never touched the contact. A platform that said customer and sent anyway points at a sequence with no gate or no exit. A mailed record that is not the one carrying the flag points at a duplicate or an import.
Ownership is the third question, and it is the one the complaint usually exposes. The RACI matrix framework lists the decisions in a piece of shared work against the roles involved and insists on exactly one accountable role per row. Written for this problem the rows are the customer flag in the CRM, the sync that carries it, the entry conditions on every prospect sequence, the merging of duplicate records and the rules an import applies on a match; the columns are sales operations, marketing operations, the sequence owner and whoever owns data protection. Fill the accountable column first. A row with nobody in it is where the gap was, and a row with two names is where the next one will be. The guide to syncing the CRM and the marketing platform already uses the same matrix for field ownership, so the rows can be added to a table that exists rather than started from nothing.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| The marketing platform showed the person as a customer and sent anyway; the entry condition passed weeks earlier | Whether the sequence checks customer and open-opportunity status at entry only or before each send | No customer or open-opportunity condition at entry, or one checked at entry only and never again | Gate on customer and open-opportunity status at entry and before every send |
| The deal shows won; the contact's lifecycle stage in the CRM still reads lead | What the closed-won process changes, and on which records | Nothing changes on the contact when a deal is won, so the person is still a lead to every workflow | Build the closed-won workflow that sets the customer stage and exits prospect sequences |
| The CRM says customer, the marketing platform says lead, or caught up after the send | The sync field table for lifecycle stage, its direction and the sync schedule | Customer status set in the CRM but never reaching the marketing platform, or reaching it hours late | Give the stage one owner, carry it in one direction and read the sync error queue |
| The record that was mailed is not the record that carries the customer flag | Every record holding the address, and the contact's colleagues at the same company | A second record for the same person that carries no customer flag | Merge under written rules with the CRM merging first |
| Customers reverted to lead on the day of a list load | The import batch tag, the match rule and what it was allowed to change | An import that reset existing customers to new leads on match | Restore the stage from the deal history and change the match rule to fill empty fields only |
| The customer booked a demo or signed last week and still received the next steps | Which exits the sequence has: the goal event, a disqualifying event, a higher-priority workflow | A prospect sequence with no exit on the goal, so the demo booked last week is still being encouraged | Wire exits on the goal event, any demo booked by any route and an open opportunity |
Fix, verify and prevent each cause
No customer condition, or one checked only at entry
The governance guide's rails are the conditions that apply to every workflow regardless of what its designer remembered, and two of them own this cause. Current customers do not receive prospect sequences, and contacts with an open opportunity do not receive automated sales nudges; both are CRM conditions, both depend on fields kept accurate by a named owner, and both are checked before every send rather than only at entry, because status changes mid-sequence. A condition checked once lets a lead who becomes a customer at the third step receive every step after it, and a sequence with no condition at all lets a customer in from the first form fill.
Fix: add the customer and open-opportunity conditions at entry and before every send in every sequence written for prospects, exit anyone already enrolled who meets them, and make a workflow that bypasses the gate a reviewable change, as the guide does for anything shared.
Verify: a seeded customer contact submits the sequence's entry form and enters nothing; a seeded lead that is marked as a customer part-way through the sequence receives nothing further.
Prevent: record the gate in the inventory row's trigger and conditions field, so a sequence is not activated without it, and use the guide's monthly spot check of suppression and consent conditions with seed contacts to catch a gate that an edit quietly removed.
Nothing changes on the contact when the deal is won
The closed-won workflow tutorial starts from the observation that when a deal is marked won a dozen things should change at once, and in many CRMs they change by hand or not at all. The first of them is the lifecycle stage set to customer on each contact and the company associated with the deal, because every later condition and report reads it. Where the won status lives only on the deal, the contact is still a lead to every workflow that reads contacts, and the demo nurture carries on exactly as the tutorial's common mistakes describe. The tutorial builds the one workflow that does the rest in order: one agreed trigger on the deal, the customer stage on contacts and company, pipeline workflows stopped, the deal's contacts exited from every prospect sequence with the customer stage added as a suppression condition, prospecting audiences cleared, the onboarding hand-over, the won outcome sent to the ad platforms, refreshed customer exclusion lists, and a reversal path.
Fix: build that workflow, or repair the step that sets the stage on the contact rather than on the deal, and run it for the won deals whose contacts still read lead, after checking how many would enter.
Verify: a dummy deal with marked test contacts is marked won, and every row of the tutorial's table checks out: stage set, sequences exited, audiences cleared, task created; then it is reopened and the reversal checks out too.
Prevent: keep the closed-won workflow at the top of the governance guide's weekly volume list, because the customer flag on a won deal is the workflow sales and marketing would both notice if it stopped, and re-test it with the dummy deal after any change to stage definitions or the deal pipeline.
Status set in the CRM but never reaching the marketing platform
The sync guide gives lifecycle stage one owning system, the CRM, one accountable team and one direction, from the CRM to the marketing platform, with the marketing platform requesting a change through its own field and a CRM workflow rather than writing the stage itself. Three failures of that contract produce this cause. The field is not in the sync at all, so the platform's copy is whatever it was when the record first crossed. The sync carries it on a schedule, so a deal won this morning is a lead to the platform until tonight, and a step scheduled for this afternoon goes out. Or the platform recalculates the stage from its own score and writes the lead value back, which the guide names as the field most often written by both sides. When the stage arrives and then changes back, the diagnosis for CRM values changing back after a sync owns that reading.
Fix: put lifecycle stage in the field ownership table with the CRM as its owner and the marketing platform read-only, carry it in the sync, read the error queue for records the stage failed to reach, and where the platform needs to promote a lead give it a request field of its own.
Verify: set a seeded contact to customer in the CRM, confirm the platform shows customer after the next sync, and confirm it still does after the platform's own workflows have run; then check that the seeded contact is out of every prospect sequence in both systems.
Prevent: follow the guide's watching routine: the error queue read on a schedule by a named person, the counts of customer-stage records on each side compared for a gap that grows, and a sample of records checked field by field for a stage that differs or keeps changing. A new synced field, or a change to the sync's direction, stays a reviewable change.
A second record with no customer flag
Duplicates arrive from predictable places, and the CRM fundamentals guide lists them: the same person using a work and a personal address, a list import that did not match on email, a record created by hand instead of by search, a form that creates a new record on every submission. Each duplicate splits the person's history, and the guide's consequence is the one that matters here: workflows may fire for both records, and the one sales is working may not be the one marketing is nurturing. The customer flag sits on the record sales touched; the record a webinar form created last month has no flag and meets every entry condition. The sync guide adds the cross-system version: a merge in the CRM does not merge the marketing platform's copies unless the surviving identifier is passed across.
Fix: find every record holding the person's addresses and every contact at the customer's company, merge under the rules the fundamentals guide asks to be written in advance, with the CRM merging first and the surviving record's identifier passed to the marketing platform, and exit the merged record from every prospect sequence.
Verify: one record remains in each system, it carries the customer stage, and deduplicationRemoving double counts when one conversion is reported by more than one source. of the marketing platform's copies shows no second reachable record for the address.
Prevent: match on email at every entry point, forms, imports, integrations and manual creation, with the exception requiring a deliberate choice, and keep the weekly pass over the merge queue that the guide's hygiene routine asks for.
An import that reset customers to new leads
The tutorial on importing contact lists opens with this failure: a list import takes minutes and can undo months of CRM hygiene, with existing customers reset to new leads among its first examples. The tutorial's rule on a match is the safe default: the import may fill empty fields and add the person to the new campaign, but never overwrites original source, created date, lifecycle stage, owner or consent. A match rule that overwrote on match turned every customer on the file into a lead, and the workflows that read a new lead did the rest, which the tutorial's step on checking which workflows an import will trigger exists to prevent.
Fix: find the batch by its tag, restore the lifecycle stage for every matched record from the CRM's field history or the won deals associated with it, exit those contacts from the sequences the import triggered, and change the match rule before any further load.
Verify: every record in the batch that is associated with a won deal shows customer in both systems, and none of them is enrolled in a prospect sequence.
Prevent: make the fill-empty-fields-only rule the standing default, include rows that match existing customers in every sample, list the workflows the import will meet before the full load, and never load a file without a batch tag, so a bad import can be found and reversed.
A prospect sequence with no exit on the goal
A sequence that invites the lead to book a demo should stop the moment the demo is booked, and the nurture sequences guide names the exits that make it stop: the goal event itself, a disqualifying event such as a deal closed or a salesperson taking the record, and a higher-priority workflow entering. Its own description of the failure is the symptom on this page: a lead already in a sales cycle receiving automated encouragement to book a demo they booked last week, and, after the opportunity closes, a lead who should have left every prospect sequence and did not. The exit has to fire on the goal reached by any route, not only on the sequence's own link, because customers book by phone, from the pricing page and through the account manager.
Fix: wire exits on the goal event however it arrives, on an open opportunity and on the customer stage, and route a reply to a named person who exits the lead from the automated steps.
Verify: a seeded lead raises the goal event part-way through the sequence through a route other than the sequence's own link and leaves at once; a seeded lead with an open deal is pulled out when the deal is created.
Prevent: write the exits into the sequence plan beside the goal before the build, and treat an exit that fires only on the sequence's own link as a plan with a blank in it. The same missing exits are one of the reasons a sequence's entrants never reach its goal, and the diagnosis for leads stalling in a nurture sequence reads the sequence from that side.
Prevention: a customer gate at every send and a seeded customer record that every change has to pass
Every cause above is found the same way, by a customer who received something meant for a stranger. The cheapest prevention is to be that customer first. Keep a permanent seeded contact that is a customer in the CRM, with a won deal behind it, and that must stay out of every sequence written for prospects; keep a second with an open opportunity and a third that has objected to marketing. Run them after every change to the sync, to a prospect sequence's entry conditions, to the closed-won workflow, to the match rules on an import, or to the scoring model, and read the enrolment history rather than the workflow's settings, as the workflows hub describes for any launch. A seed that enters is a gap found in minutes; a customer who complains is the same gap found in months.
The gate is the other half. The governance guide's customer and pipeline suppression is a condition on every prospect workflow, checked at entry and before every send, and it is only as good as the field it reads. That is why the responsibility matrix from the first section belongs beside the automation inventory and the sync's field ownership table: the customer flag has one owning system and one accountable team, the sync that carries it has an owner who reads the error queue, every prospect sequence has an owner who answers for its entry conditions and its exits, the merge queue has a named reviewer, and an import has a named person who settled its match rule on a sample before the full load. A suppression listA list of addresses or contacts excluded from every send, such as unsubscribes, hard bounces and spam complaints. for customers is what the closed-won tutorial calls this arrangement, an exclusion that every prospect workflow checks, and it fails whenever one of those rows has nobody accountable.
Watch two counts weekly, as the sync guide watches record counts and the governance guide watches volumes. The number of customer-stage contacts in each system should differ by a gap that is stable and explained, and a gap that grows means the stage has stopped crossing. The number of customer-stage contacts enrolled in any prospect sequence should be none, and the report that lists them, run against the platform's enrolment data rather than against the CRM, is the earliest warning this symptom gives. Add both to the weekly review in minutes, beside the volume alarms, and act on an outlier the same day.
Incidents still happen, and the routine for them is the governance guide's: pause, identify the affected records, decide on a correction with the owner, fix, log and check the inventory for the same cause. Where the customer, or the account manager on their behalf, asks not to receive marketing at all, the request is honoured in every tool rather than in the one that sent the message. The ICO's direct marketing guidance, as the site's guides state it, expects an objection to marketing to be honoured in every channel it covers, and the GDPR, as the syncing guide states it, gives people an absolute right to object to direct marketing, which makes a message a person asked not to receive a data protection matter as well as a marketing one. From that point the diagnosis for unsubscribed contacts still receiving email owns the problem, because it starts from an opt-out rather than from a customer status. None of this is legal advice; whoever owns data protection should hear about every incident.
The same fault has a paid-media version, and the closed-won workflow already reaches it: the new customer is removed from prospecting audiences and the customer exclusion lists are refreshed. When customers keep seeing retargeting ads for what they bought, the diagnosis for Meta retargeting ads reaching recent buyers traces the exclusion, the purchase event and the windows, and the guide to using customer lists across ad platforms keeps one definition of a customer for every platform so that a person excluded on one network is not bought again on another. Whether prospecting should exclude customers at all is a decision for the business, and the existing customer exclusion test is the design that answers it with matched regions rather than opinion. Lead nurturingKeeping in useful contact with leads who are not yet ready to buy until they are. and prospecting ads read the same flag, which is one more reason to give it one owner. The glossary entry for lifecycle stage gives the short definition this page assumes, and the marketing automation troubleshooting hub lists the neighbouring symptoms, such as leads stalling in nurture or values reverting after a sync, that often arrive in the same week as this one.
A fleet telematics provider finds how a client re-entered prospecting (illustrative)
Casperling Telematics
Casperling Telematics, an invented provider of vehicle tracking for commercial fleets, hears from an account manager that a client of 4 years with 260 tracked vehicles has received a 3-step sequence inviting them to book a first demo. The marketing operations lead pauses the sequence and pulls its send log: 1,140 contacts received the first step in the past 30 days, and 38 of them belong to companies with a won deal.
The client's CRM record shows the customer stage, set 4 years ago. The marketing platform's copy of the contact shows lead. The closed-won workflow, built 2 years earlier, sets a won flag on the deal and creates the onboarding task, but never writes the lifecycle stage on the contact; the 38 affected contacts all belong to deals won since that workflow replaced a manual step. A second record for the client's fleet manager, created 5 weeks earlier when she registered for a webinar under a shared depot address, carries no status at all, and the sequence's entry condition, checked only at entry, let both records in.
The provider moves the customer flag to the contact with the CRM as its owner and the sync carrying it to the marketing platform in one direction, and extends the closed-won workflow to set it on every contact and the company of the deal. It adds a customer and open-opportunity condition checked before every send in all 6 prospect sequences, exits the 38 contacts and 11 others found by a report of customer-stage contacts enrolled in any prospect sequence, merges the two records with the CRM merging first, and writes a 5-row responsibility matrix naming one accountable owner for the flag, the sync, the entry conditions, the merge queue and import match rules. A seeded customer contact is added to every system and run after each change. In the next quarter's audit the weekly report of customer-stage contacts in prospect sequences shows 0 on 12 of 13 weeks, and the 1 exception, caught on a Tuesday, is traced the same day to a partner list loaded without its batch tag.
Common mistakes
- Removing one customer from one sequence. The rule that mailed them mailed others; pause the workflow and count from the send log.
- Reading the status now instead of at the time of the send. The flag may have arrived an hour after the message went, which is a different cause from a flag that never arrived.
- Setting the won status on the deal only. Every workflow reads the contact, and the contact still says lead.
- Letting two systems write the lifecycle stage. The value is whatever was saved last, and a customer can be a lead again by morning.
- Checking the customer condition at entry only. People become customers part-way through a sequence.
- Merging in one system. The marketing platform keeps both records and mails the one without the flag.
- Importing with overwrite on match. Customers become new leads and the welcome and nurture workflows do the rest.
- Treating the complaint as a sequence problem alone. If the customer asked not to be contacted, it is an opt-out to honour everywhere, and the ads side needs the same exclusion.
