Deduplicate campaign events by storing a stable event key and tracking whether a specific downstream action has already been completed. A retry of an event should not create a second sales task. Two different events can still be meaningful activity even if your business rule permits only one open task for a buyer. Design the event rule and the sales rule separately.
The setup in practice
Imagine a vendor serving fleet maintenance shops. Its integration receives an authorized click event twice after a network timeout. The first delivery created a follow-up task before the connection failed. The second delivery repeats the event. A day later the prospect clicks a different link about installation. The first duplicate is a technical replay. The later click is a new event. Neither click independently proves the viewer's identity or buying authority.
Cortex Signal Flow can pass click events to a CRM, and the brief names HighLevel among connected CRMs. Verify whether the configured path exposes a stable event ID, event occurrence time, and delivery status before designing a deduplication rule. Built-in email tracking in HighLevel is a separate capability from importing a campaign event through an integration.
The outside evidence and its limits
HighLevel's documentation separates native email and trigger link behavior. Confirm how the configured CRM integration carries external click events and which metadata is available before implementing a replay rule. See the HighLevel documentation on email events.
Set up a controlled walkthrough
Give the CRM owner one timeline containing occurrence time, receipt time, event key, action key and action status. Keep the second delivery of the same event as a diagnostic if permitted. The seller should see the legitimate engagement history without receiving a second instruction to call the same buyer.
Walk through a failure after task creation and a failure before task creation. The first needs reconciliation to avoid a duplicate. The second may require a retry to avoid dropping the work. A stable event record and a separate action record let the integration distinguish those two situations.
Stable key
Prefer a documented immutable source event ID with source and tenant context. If there is no source ID, examine collisions and replay behavior before using a composite key. To close the review, write down what the decision owner is expected to do. Never use contact ID alone because that would collapse distinct events for the same person.
Ask whether the stable key check outcome should change a task, a deal stage, or only the evidence stored about the account. Keep these choices separate in reporting. This allows the operations owner to review a meaningful event without overstating the buyer’s progress. If they disagree about facts, improve the definition or show where further confirmation belongs before expecting the data to train a model. An approver can check that the rule behaved as specified.
Before treating the stable key check as ready for routine use, agree how to handle evidence that arrives after the initial decision. Identify which current state can change and which original snapshot remains frozen. This preserves the difference between the latest information and the evidence available when the earlier action was approved.
Separate event processing from business action
An event record represents an observation from one source. An action record represents what a particular workflow decided to do about it. Use separate keys when several workflows can legitimately react to the same event. For example, an authorized click might update an internal activity view and independently supply context to an existing task. A unique action rule can prevent a second task without erasing the event. The system owner should document which parts of the integration can verify an earlier action and what happens if that confirmation is unavailable.
Task identity
Write a seller-facing version of the the task identity check rule and an analyst-facing version. Link the task to the account, action type, and decision window. Existing tasks can absorb new relevant context while the original event remains in history.
Apply the rule while the source evidence is still available. Verify the owner can see which touch came first without opening multiple tasks. Date the observation used for the task identity check and the response it prompted. Even if a seller corrects the record after discovery, the original evidence should be available within permitted access to explain why the workflow took its earlier path. Do that before treating the reviewed sample as representative of the full queue.
When comparing replay-aware configuration with previous configuration, verify how the task identity check was handled in both views. Different rules can explain an apparent difference even if the underlying cases behave similarly. Keep the rules visible with the results and show any material exception when the two groups cannot be judged on equivalent terms.
Partial failure
Simulate a task creation success followed by a timeout before the event processor receives a reply. A blind retry can duplicate work.
Review the partial failure check with the person who actually takes over the account. For the partial failure check, compare the activity that arrived with the change the seller actually made. A reliable record needs both, because an incoming activity is often legitimate even when the requested downstream action is not appropriate. Inspect the result where sellers work. Put the next step in the operating notes. Ask how the connector finds the prior action on retry or route ambiguous states to reconciliation.
Read this checkpoint from the buyer side as well as the operator side.
Consider how the partial failure check behaves when the buyer or operator supplies incomplete information. Define a useful next action for that state. That might be gathering a missing input, holding the case, or keeping an aggregate observation. The approved action should match the evidence that exists at the time.
Two cases to work through
The system creates task T-19 and times out. On retry, searching the action key returns T-19, so processing completes without another task. A second event on the same opportunity lands a day later and adds history, but the open-task rule leaves T-19 in place.
In another test the retry occurs after a contact becomes suppressed. Processing can acknowledge the prior record without starting another send. Confirm the connector does not bypass the latest suppression state just because an older event is being replayed.
Unresolved link
Store unresolved account or person relationships according to policy. A weak association should not acquire certainty just because it crosses a CRM boundary. The review record for the unresolved link check should retain this disposition. Prevent unresolved events from automatically creating contact-level follow-up tasks.
Ask whether the unresolved link check outcome should change a task, a deal stage, or only the evidence stored about the account. Keep these choices separate in reporting. This allows the operations owner to review a meaningful event without overstating the buyer’s progress. Keep those meanings distinct when exchanging the record with analytics or a campaign operator, because a filtered report can turn the same blank into an apparent negative.
Connect the unresolved link check to the report on replay test. Establish whether an exception changes eligibility, leaves the result unknown, or just produces a note. Removing a record and counting it as a failure have different effects on the reported percentage. Describe the handling clearly enough for a second reviewer to reconstruct it.
Exercise the connector's failure modes before handoff
Test a response timeout, a temporary service error, and an invalid client mapping. The first might require reconciliation because the action may already exist. The second may be safe to retry if no action occurred. The third needs correction before another attempt. Use marked records and keep permitted traces showing which CRM client and task received the event. HighLevel's native click trigger is useful context, but it is the configured connector's own behavior that determines how an external event should be retried.
Replay review
Give the replay review check a dated correction path. Run tests with the same ID, similar but distinct IDs, delayed events, and a suppressed record. Inspect both activity and tasks under the correct client.
For this case, the expected next step is clear. Confirm replay cannot bypass the latest suppression state or create a cross-client task. Date the observation used for the replay review check and the response it prompted. Even if a seller corrects the record after discovery, the original evidence should be available within permitted access to explain why the workflow took its earlier path. The rule should be precise enough to explain the difference to the owner.
Put a failure example beside the standard example for the replay review check. For the first example, require the reviewer to explain why the rule should stop. For the second, ask why it should advance. This exposes definitions that sound precise in a meeting but do not lead to consistent decisions under ordinary conditions.
What the replay test denominator includes
The fictional comparison below uses scripted workflow retry cases as its unit. Write the count that produced exactly the expected number of tasks beside each denominator before doing any division. The result describes this specific measure. A separate measure is needed if the decision owner wants to estimate a different outcome.
| Illustrative group | Count meeting rule | Eligible count | Rate |
|---|---|---|---|
| Replay-aware configuration | 29 | 32 | 90.62% |
| Previous configuration | 22 | 32 | 68.75% |
The first result is 29 out of 32, or 90.62 percent. The second is 22 out of 32, or 68.75 percent. On unrounded rates the first is 21.88 percentage points above the second. If you present a relative change, keep the comparison denominator explicit. Using 68.75 percent as the reference gives a difference of about 31.8 percent in relative magnitude.
An unchanged case set gives an operational comparison, not evidence that prospect behavior changed. Investigate each failed case because an erroneous task can be more disruptive than a delayed diagnostic event.
Record the snapshot date beside the replay test result. For this hypothetical comparison, use the same written review standard for both groups. If a reviewer changes the comparison rule, keep the original replay test calculation and clearly identify the revised version.
When presenting replay test to a sales manager, show which underlying cases still need action. A pass can mean the system recorded the intended result under test conditions. It does not tell you that every prospect in a later production queue will respond or that a downstream milestone happened.
Assign ownership for actions in an uncertain state
A queue can mark an event processed when it has only attempted the CRM write. Ask whether the system can distinguish attempted, confirmed, rejected and awaiting reconciliation. A human operator can resolve an ambiguous task when the integration cannot safely confirm what happened. Save the decision and the original action key. A later event may supply additional context, but the existence of the older retry should never become the reason for a second task.
The replay test counts leave 3 cases that did not meet the rule in the first fictional group and 10 in the second. In actual data, inspect those remaining cases before calling them all failures. For example, some may be awaiting a decision, while others may be complete and outside the rule. Keep a reconciliation showing the final and unresolved dispositions. That distinction matters more than rounding the percentages to another decimal place.
Review the comparison of replay-aware configuration and previous configuration with the sales operations owner. Inspect one record that changed a seller action and one that only changed the evidence visible in the CRM. This distinction helps the team avoid treating every logged click, task or label as progress in the buyer’s actual decision.
For more on the signals behind this decision, read the Cortex guide to buyer intent data.
If the team cannot reproduce replay test, create a smaller check using one agreed input and a fixed observation window. Assign a date to examine the result. The sales operations owner can then distinguish a missing measurement capability from an ambiguous commercial outcome and give the client a more specific next step.
Rehearse the handoff with a seller and an administrator
Release only after the replay cases produce the intended tasks and preserve legitimate event history. Bring a retry example that created uncertainty about whether a task already exists. Review the stable event and action keys, who can reconcile the failure and what the seller should see. A short test can save time spent unwinding duplicate work.
Visit Cortex Marketing AI to see how sales outcomes can inform which buyers to prioritize.
Bring an anonymized CRM example, your intended disposition, and the action that should follow. If you want to scope that work, call Cortex Marketing AI to discuss a consultation.
