Audit each GoHighLevel sub-account as its own production system before changing a shared workflow pattern. Start with a read-only inventory, map every location-specific dependency, trace real enrollments and exits, classify defects by consequence, and approve a test and rollback plan. Similar workflow names or a common snapshot do not prove that the locations still behave the same.
This guide is a practitioner framework for review and change control. The examples are hypothetical. It does not claim that Edex Growth has delivered a particular multi-location result, and it does not replace the business’s approval, privacy, messaging, or compliance review.
Start with account boundaries, not the workflow canvas
In HighLevel, agency access and sub-account access are different scopes. Current HighLevel documentation says agency permissions can control which sub-accounts a user can enter, while sub-account permissions control access to modules such as Workflows, Calendars, Conversations, and Opportunities. Confirm the account name, location identifier, your authorized role, and the requested scope before opening a workflow.
Use least access that still permits the audit. A person who only needs selected locations should not receive agency-wide authority merely to compare workflows. Record whether the audit is read-only, whether draft edits are allowed, and who may publish.
Before inspecting logic, capture:
- the agency and exact sub-account;
- the business owner and technical change approver;
- the location timezone, business hours, and primary market;
- the approved communication channels and sending identities;
- whether the workflow came from a snapshot, a copy, or a local build;
- the last known accepted configuration and where its evidence is stored;
- other people who may be editing the same account.
Do not assume that a fix in one location should be copied to the others. A calendar, user, phone number, custom field, or pipeline stage can look equivalent in the interface while referring to a different account-specific record.
HighLevel’s current agency permission guide and sub-account permission guide describe the two access layers. Verify them again before publication because permission controls can change.
Build a multi-location audit matrix
Create one row per workflow per sub-account. Do not begin with a global statement such as “all locations use the same follow-up.” Make that a conclusion only after the rows agree.
| Field | Location A | Location B | Location C | Evidence |
|---|---|---|---|---|
| Workflow name and status | Record exact name and draft/published state | Record | Record | Workflow list and version history |
| Entry triggers and filters | Record event and every filter | Record | Record | Trigger configuration |
| Re-entry and opportunity behavior | Record settings and exceptions | Record | Record | Workflow Settings |
| Sender and time window | Record name, email, number, timezone | Record | Record | Workflow Settings and channel configuration |
| Data dependencies | Fields, tags, pipeline, stages | Record | Record | Action and condition settings |
| People and calendars | Assigned users, teams, calendars | Record | Record | Referenced records |
| Exit and suppression | Reply, booking, DND, opt-out, removal | Record | Record | Settings and execution evidence |
| External calls | Endpoint purpose, owner, expected response | Record | Record | Webhook configuration and receiving logs |
| Last accepted test | Date, scenarios, approver | Record | Record | QA record |
| Defects and drift | Severity and owner | Record | Record | Audit log and evidence link |
The matrix is an audit control, not a migration plan. It reveals where apparently standard locations differ before anyone makes them more alike.
Inventory workflows and their entry conditions
For each location, list every published workflow, every draft that could later replace it, and every inactive workflow that still explains historical records. Group workflows by business job, not only by folder or name. Examples include new-inquiry intake, appointment reminders, missed-call recovery, estimate follow-up, booking exits, and staff alerts.
Open each workflow and record:
- every trigger and its filters;
- whether multiple triggers can admit the same event;
- whether another workflow listens for the same event;
- the workflow’s published or draft state;
- Allow Re-entry and Allow multiple Opportunities;
- Stop on Response and any explicit removal actions;
- timezone, time-window, and sender settings;
- branch, wait, goal, and exit behavior;
- the record state expected when the workflow finishes.
HighLevel’s current Workflow Settings overview explains that re-entry, multiple-opportunity, response, time-window, and sender controls affect enrollment and execution. Appointment and invoice triggers have documented re-entry exceptions, so the setting label alone is not enough. Inspect the actual trigger type and test the destination account.
Trace conditions, branches, waits, and cross-workflow dependencies
A workflow can be internally correct and still conflict with the account around it. Build a dependency map for every action that reads or changes shared state.
Check:
- If/Else branches for missing catch-all paths;
- Wait actions for timezone, event, timeout, and contact-state assumptions;
- workflow-to-workflow add and remove actions;
- tags or custom fields used as entry signals elsewhere;
- pipeline changes that trigger another workflow;
- appointment events that can occur more than once;
- messages that continue after reply, booking, cancellation, or opt-out;
- contacts already waiting in the published version when a later step changes.
HighLevel’s current Version History and Restore guide says restoration requires the workflow to be in Draft with no enrolled contacts. A restored revision becomes a new draft, and the published workflow remains unchanged until that draft is published separately. Version restoration is therefore not a guaranteed rollback mechanism for a workflow with active enrollments. Production recovery may require a separately approved containment and record-reconciliation procedure. The guide also says contacts already in a workflow can use changes to upcoming actions, while a changed Wait does not alter an existing wait already in progress. The feature is documented as account or Labs dependent. Confirm availability and behavior in the location rather than promising it as a universal rollback button.
Protect identity, ownership, and duplicate rules
Separate the contact from the opportunity. One contact may legitimately have more than one job or opportunity, while repeated form submissions may be duplicate events for the same job. Define the business identity rule before interpreting the automation.
For each workflow, ask:
- Which fields identify the contact?
- What makes an opportunity new rather than an update?
- Can the same contact have several active opportunities, and in which pipelines?
- What happens when two triggers fire close together?
- Does a retry create another task, message, or opportunity?
- Does assignment replace an existing owner or only fill an empty owner?
- Which owner must remain protected during imports, reactivation, or re-entry?
- Which state proves that the current event has already been processed?
Do not “clean up” contacts, opportunities, owners, or tags during discovery. A destructive merge or rename can remove the evidence needed to diagnose the actual defect. The duplicate opportunity guide explains the business-rule questions in more detail.
Verify every location-specific dependency
Use this checklist before changing any branch or action:
Records and routing
- Custom fields exist in that location and have the expected object type.
- Tags are current, distinct, and still used intentionally.
- Pipelines and stages exist and mean the same thing locally.
- Assigned users, teams, and fallback owners are active.
- Calendars point to the intended service, team, availability, and timezone.
- Forms, surveys, funnels, and pages write the fields the workflow reads.
Communication
- From name, from email, and from number are approved for the location.
- The sending number and email domain are connected and eligible.
- DND and opt-out state stop the affected channel.
- Replies, bookings, cancellations, and manual takeover have explicit exits.
- Templates and custom values resolve to local information.
Integrations
- Webhook URL, method, authentication owner, and payload fields are documented without copying secrets into the audit.
- The receiving system has a matching test and production boundary.
- A retry cannot repeat an irreversible side effect.
- Failure responses have an owner and a visible state.
- Snapshot-linked assets are identified before a refresh or push.
HighLevel documents Custom Webhook as an outbound workflow action with configurable methods, authentication, headers, and payload. Validate the destination and business response, not only a successful send. Review the current Custom Webhook guide before using its current controls.
Treat snapshots as a source, not proof of sameness
A snapshot can carry workflows, funnels, calendars, forms, custom fields, custom values, and other supported assets. HighLevel also documents that contacts, appointments, conversations, integrations, and several account connections are not snapshot content. A snapshot-loaded workflow can therefore depend on local state that the snapshot did not create or reconnect.
HighLevel’s snapshot creation guide lists current included and excluded asset categories. Its snapshot refresh guide says updates are selected and pushed to linked sub-accounts, and warns that deleting steps through a refresh can remove contacts waiting on those steps. A multi-location release plan must identify those contacts and dependencies before a push.
Use the GoHighLevel snapshot import checklist when the audit includes loading or reconnecting snapshot assets.
Classify defects by consequence
Use severity to decide what can wait and what should stop a release.
| Severity | Definition | Example | Required response |
|---|---|---|---|
| S0: stop | Active or imminent behavior could create unauthorized communication, data exposure, destructive change, or high-consequence wrong action | Opted-out contacts can receive a message, or a webhook exposes protected data | Pause the affected path, preserve evidence, notify the authorized owner |
| S1: critical | Core business state or customer handoff can be wrong with no dependable recovery | A booked contact remains in sales follow-up, or the wrong location receives the record | Block the change or launch; diagnose, repair, and retest related paths |
| S2: material | The process still works, but duplicates, delays, or manual recovery are likely | Two owner alerts or an opportunity in the wrong non-final stage | Assign an owner and correction window; test before broad rollout |
| S3: maintainability | No current customer-facing failure is proven, but future changes are risky | Unclear names, unused branches, or undocumented custom values | Add to controlled cleanup after functional behavior is protected |
Severity describes consequence and urgency, not how easy the fix appears. A one-click sender change can be S0. A large naming cleanup may be S3.
Test normal and exception paths before publishing
Write expected results before running a test. Use fresh, controlled test contacts and channels owned by the testing team. Do not use a real customer’s contact record merely because it already has the right fields.
| Scenario | Expected entry | Expected state change | Expected communication | Expected exit or owner |
|---|---|---|---|---|
| Valid new inquiry | One enrollment | Contact and one intended opportunity state | Approved acknowledgement only | Assigned local owner |
| Repeated event | Identical retry follows the idempotency rule; a separate legitimate inquiry follows the approved identity and opportunity rule | Retry creates no duplicate side effect; legitimate inquiry creates or updates only the approved record | Retry creates no duplicate message; legitimate inquiry uses the approved sequence | Existing owner protected unless the business rule explicitly reassigns it |
| Missing required field | Exception branch | No false completion state | Internal alert if approved | Named exception queue |
| Existing active opportunity | Defined business rule | Correct record updated | No competing sequence | Current owner retained unless approved |
| Reply or opt-out | No new ineligible send | Channel status preserved | Suppressed | Human owner or closed path |
| Booking or cancellation | Event-specific enrollment | Appointment and pipeline state agree | Correct reminder or suppression | Booking owner or exception owner |
| Webhook failure | One visible failed attempt | No false success | No repeated external side effect | Retry or manual recovery owner |
| Outage or unavailable dependency | Safe fallback | Pending or failed state | No misleading confirmation | Monitored recovery queue |
Then inspect the contact record, opportunity, conversation, appointment, Enrollment History, Execution Logs, and receiving system. Current HighLevel documentation says Execution Logs and Enrollment History show how contacts moved through a workflow and why they completed, failed, or were removed. The current logging guide also documents path highlighting and snapshot-removal entries.
The existing workflow testing guide covers acceptance evidence and defect retesting. This article adds the cross-location comparison and change-control layer.
Release one controlled change at a time
Before a production edit:
- name the affected locations and exact workflows;
- save the current version or other recoverable baseline available in the account;
- record active contacts at waits or high-risk actions;
- define the defect, expected correction, and affected scenarios;
- assign editor, reviewer, approver, and monitor;
- make the smallest change in one location or safe test copy;
- run normal and exception tests;
- publish only after approval;
- monitor real executions for the agreed period without exposing customer data;
- expand to another location only after comparing its dependencies again.
Rollback means more than restoring the canvas. It may also require stopping new enrollment, reconciling records changed during the release, suppressing retries, and notifying the human owners of affected work. Write those steps before the change.
Safe audit and release checklist
Before editing
- Account and authorization confirmed
- Workflow inventory complete for every in-scope location
- Triggers, filters, settings, and cross-workflow signals mapped
- Identity, opportunity, duplicate, and owner rules approved
- Fields, tags, pipelines, calendars, users, senders, and webhooks verified locally
- Snapshot source and local deviations recorded
- Active enrollments and waits reviewed
- Defects classified by consequence
- Recovery and rollback owner named
Before publishing
- Expected results written for normal and exception tests
- Fresh controlled contacts prepared
- One scoped change reviewed
- Related workflows retested where they share a trigger or state
- Logs and receiving-system evidence captured
- DND, reply, booking, cancellation, and opt-out exits checked
- No QA-only restriction left in the release
- Version, approver, and acceptance date recorded
After publishing
- First executions reviewed by a named person
- Duplicate messages, tasks, and opportunities checked
- Failed or waiting executions monitored
- Location-specific drift added to the matrix
- Rollback remains available until the acceptance window closes
- Handoff notes identify where staff should look when behavior differs
Audit before you standardize
The useful outcome of a multi-location audit is not identical screenshots. It is a defensible record of what each location does, why it differs, which dependencies are local, and how a change will be tested and recovered.
If the account already has conflicting workflows, unclear ownership, or location drift, GoHighLevel Audit and CRM Cleanup is the relevant commercial path. Edex Growth can review the evidence and prioritize repairs, but scope, access, and approval still belong to the account owner.
Sources and editorial scope
Product behavior was checked on November 15, 2026 against current HighLevel documentation for agency and sub-account permissions, workflow settings, execution history, workflow version history, snapshots, and Custom Webhooks. Account availability and behavior can differ by plan, permissions, Labs status, feature rollout, and configuration. The audit matrix, severity framework, dependency checklist, test matrix, and release checklist are original editorial frameworks, not HighLevel guarantees or claims of a specific client delivery.
