Skip to main content

Automations Best Practices

Recommended practices distilled from across the Automations configuration guides, covering workflow setup, every node type, monitoring, and permissions. Each section below corresponds to its own detailed walkthrough.

Getting Started​

The Create dialog's choices carry consequences that extend well past the dialog itself.

  • Decide Trigger Type carefully at creation. Event-based versus Schedule-based is locked permanently once the workflow is created. It cannot be changed later, even from Edit mode on the Details tab.
  • Leave Set Date Range disabled unless the workflow truly needs a bounded window. The default is Disabled, running indefinitely. Enabling it requires both a Start and an End Date, and any run outside that window is skipped rather than queued.
  • Treat Submit as only the first step. Creating a workflow just opens the Designer with an unconfigured Trigger node. It still needs a real event or schedule, and at least one Action, before it does anything.

Workflow Builder​

The canvas offers both a reversible way to pause a workflow and an irreversible way to remove one. Know which is which.

  • Use the Status switch on the Details tab to pause a workflow, not Delete Workflow. Status is reversible. Delete Workflow permanently removes the workflow and every execution record tied to it, with no undo.
  • Save after every meaningful change. Each Save Workflow becomes its own versioned entry on the History tab, so a prior state can be reviewed later.
  • Don't look for a delete icon on the Trigger node. A workflow's own Trigger node cannot be deleted by design, only reconfigured, since a workflow without a trigger has nothing to start it.

Event-Based Workflows​

The CRM-Related trigger catalog is large because every module gets its own set.

  • Prefer "On [Module] Created or Updated" over two separate triggers when downstream steps don't care which happened. One node covers both cases instead of maintaining two parallel Trigger paths for the same workflow.
  • Use the Module filter chips or name search in the Trigger picker rather than scrolling. The catalog runs to 112 triggers across 12 pages, since every CRM module contributes its own Created / Updated / Created-or-Updated / Deleted set. Filtering to a single module, or searching a word from the trigger's name, is almost always faster than paging through from the top.

Schedule-Based Workflows​

A preset name is a convenience label. The cron expression underneath it is the actual schedule.

  • Use a Quick Preset instead of Custom Schedule whenever one of the standard intervals fits. One click, and less error-prone than assembling Frequency, Hour, and Minute by hand.
  • Re-check the Current Schedule summary right before clicking Save, every time Custom Schedule is opened. Confirmed directly: opening the Custom Schedule tab can silently shift an existing schedule, for example from midnight to noon, before any field on the tab is touched. A glance at the Current Schedule summary and Cron Expression catches this before it is saved.
  • Combine Date Range with a schedule deliberately. With Date Range enabled, the schedule keeps firing, but any run that falls outside the Start/End window is simply skipped, not deferred or queued for later.

Trigger Categories​

Every CRM module inherits the same trigger shape automatically, which is worth knowing before searching module by module.

  • Remember every CRM module automatically gets the same four-trigger set. Created, Updated, Created or Updated, Deleted. Learning this one pattern covers all CRM modules at once.
  • Filter by Messaging, not WhatsApp, to find On WhatsApp Message Received. WhatsApp has no separate module chip. It's grouped with SMS and conversation triggers under Messaging.
  • Don't assume a saved Webhook is automatically wired into the Trigger catalog. Webhook creation under Configuration → Integrations is a separate step from that webhook actually appearing as a selectable Trigger. Confirm the connection directly in your environment.

Action Categories​

A few action categories sit under a broader module filter than their name suggests.

  • Filter by Messaging to find SMS or WhatsApp actions. There is no separate SMS or WhatsApp filter. Those actions, along with generic conversation actions, all live under the single Messaging module filter.
  • Filter by System Configuration to find "Open URL" web actions. Web-related actions live alongside configuration and cache actions under System Configuration, with no separate Web filter.
  • Use Limit and Offset on Get All Records to page through a large module. Both Get All Records and Get All Records By Filter expose Filter Rules, Sort Rules, Limit, Search, and Offset together, so a large module can be paged a batch at a time instead of fetched whole and filtered afterward with a List node.
  • Don't treat a pending Add Lead to Campaign as a failure. It queues the lead immediately only if the campaign is already running. If the campaign is paused or not yet started, the lead waits pending by design rather than being dropped.

Condition Node​

The rule group's logic is more flexible than it first appears, and one field picker has a confirmed gap.

  • Use a nested Group to invert a single rule, not the NOT toggle on the whole group. NOT applies per group rather than per rule, so wrapping one rule in its own nested group is the only way to invert just that rule.
  • Use nested groups for compound logic like "A and (B or C)" instead of chaining multiple Condition nodes. Each nested group carries its own independent AND / OR / NOT, which is enough to express that logic in a single node.
  • Connect the Condition node to its upstream source before opening Select field. The dropdown only lists fields from a node it has a genuine connector line back to. A node placed on the canvas but not wired to its predecessor shows "No data" with nothing to select. See Dynamic Fields below.
  • Don't rely on a numeric greater-than comparison against a Formula field's output. Confirmed directly: the rule builder only offers string operators there, so a check like "is not empty" confirms a value exists, not that it has crossed a numeric threshold. Use a Script node instead when a real numeric comparison is required.

Delay Node​

A Delay is a pause, not a branch, and its fields combine additively.

  • Only set the time-unit fields a delay actually needs. Days, Hours, Minutes, and Seconds combine additively into one total wait. Unused fields can be left untouched.
  • Verify what Dynamic Mention mode actually lists before depending on it. Only typed-number (list) mode was confirmed to behave as expected in testing. Mention mode's available values were not captured live.

List Node​

What a List node can select depends entirely on what the node before it produced.

  • Only connect a List node after a step that actually returns an array, such as Get All Records. Select List Data Field is scoped to array-type fields only. A single record from a Trigger has no array field to offer.
  • Use First or Last instead of All when only one item from a list is ever needed. Simpler than looping through every result with the All pattern when a single item is the actual target.

Formula Node​

The live function picker is the authority on what's actually available, not the Toolbox card's subtitle.

  • Don't rely on MAX or MIN existing just because the Toolbox subtitle mentions them. Confirmed directly in the live function picker: no MAX or MIN function exists, only SUM, MULTIPLY, DIVIDE, ABS, NEGATIVE, and LEN under Mathematical Functions.
  • Use a Formula node instead of a Script node whenever a built-in function covers the transformation. It's the lighter option. Script's JavaScript editor is only necessary once a transformation exceeds the available String, Math, and Date functions.

Script Node​

The editor's autocomplete does reach workflow data, once it has something to match against.

  • Type a few letters of a field's name before expecting autocomplete to find it. An empty Ctrl+Space only shows generic JS globals. Typing casetype surfaces real fields like "Node Label - Case Type" and inserts (nodeId.fieldName) on accept.
  • Let Response Mapping auto-populate from the script's own return statement instead of pasting JSON by hand. Saving the node generates JSON Input and the Schema Editor directly from what the script returns, even inferring Field Types like Email and PhoneNumber from field names.

Dynamic Fields​

Several node fields depend on an "@" dynamic mention to pull in an upstream node's output, and that mention list is only ever as good as the connector line behind it.

warning

A dropped node doesn't auto-connect to the one above it. Without a drawn connector, every "@" field there shows "No data."

  • Check for a visible connector line first when an "@" mention field shows "No data." A solid line with a small colored dot at each end should run between the two nodes on the canvas. If it isn't there, draw it before troubleshooting anything else.
  • Use a literal typed value as a fallback on plain text fields when there's genuinely no upstream node to mention. Works on an Email action's to, subject, and text fields. Useful for the very first node in a chain, which has nothing upstream yet.
  • Build a saved filter on the module's own list screen for select-only fields with no upstream node at all. A Schedule-based Trigger has no preceding node to mention from, so a Filter field right after it (Record ID, Select List Data Field, or similar) needs a saved filter created separately and then selected by name, rather than a dynamic mention.

Automation Monitoring​

Monitoring and Registry Health Check answer different questions, and the Purge action deserves caution.

  • Start troubleshooting a specific failed run in Execution History, not Registry Health Check. Execution History shows exactly which node failed and the real data it had. Registry Health Check is for catalog and reference integrity, not single-run failures.
  • Run Registry Health Check before assuming a blank Trigger or Action node editor is a UI bug. A confirmed real case showed a blank Trigger editor was actually caused by a stale trigger reference the live catalog no longer recognized, not a display fault.
  • Treat the Purge button as destructive, and confirm its scope before using it. It exists in the interface, but its exact scope was deliberately left untested to avoid deleting real execution history. Confirm what it clears before relying on it.

Execution History​

A large tenant's execution log is too big to scroll. Search and the per-node detail are how it's actually navigated.

  • Search by workflow name before scrolling a potentially very large table. A real tenant was observed running into thousands of pages at ten rows per page.
  • Treat the main table's Error column and the expanded Step Logs' per-node message as complementary, not redundant. The top-level Error shows on the execution row. Step Logs pinpoint exactly which node produced it and what data it had at the time.
  • Don't be thrown by a step Status reading SUCCESS instead of COMPLETED. Both mean the node finished without error. Which one appears depends on the node type.
  • Open Data on a failing step rather than diagnosing from the Step Logs table alone. It opens the full Workflow Log Details, including the raw payload the node actually received, not just its pass/fail status.

Registry Health Check​

A clean set of top-level metrics doesn't guarantee every saved workflow is actually healthy.

  • Run Health Check periodically even when Schema Warnings and Endpoint Errors both read 0. A confirmed real case showed both metrics clean while the banner still read "Stale Registry Configurations Detected". The actual problem surfaced only in Saved Workflows Health further down the page.
  • Expand only the rows marked ERROR in Saved Workflows Health. PASS rows are not expandable by design. Only ERROR rows carry an error count badge and an expand arrow with detail underneath.
  • Cross-reference an internal module key against the Designer's displayed category name before concluding a module is missing. Confirmed mismatches exist, such as calls mapping to Telephony and config mapping to System Configuration.

Permissions​

The three Automations permission toggles don't form a ladder, and built-in roles can only be reviewed, not edited.

  • Don't assume the three Automations permissions form a simple hierarchy. Confirmed directly on the built-in User role: Basic Workflow Automation reads Yes while Manage System Workflow and Advanced Workflow Automation both read No, proving the three are independent toggles, not a ladder.
  • Expect built-in roles like System Owner to open read-only. Their Edit Role panel shows no Save, Cancel, or Delete. Only a role the organization created itself can actually be changed.
  • Confirm exactly what each of the three permissions gates before relying on it for access control. None of the three carries a description or tooltip in the interface, and exactly which screens each one restricts was not exhaustively tested.
  • Use Clear All when a role should have zero Automations access. One click sets all three toggles to No, instead of switching each individually.