Try the follow-up decision lab
Change the fictional lead state below and inspect why a reminder would be skipped, delayed or held for review. This calculator runs locally in your browser. It uses explicit rules, not an LLM. It never sends an email, writes a CRM record or connects your website to n8n.
Start with the due reminder. Select “Visitor replied”: the decision becomes skip. Now also select “Provider delivery outcome unknown”. The reply still stops the reminder. Reset the example, select only an unknown outcome, and inspect the review decision. A timeout is not evidence that an email failed.
Local decision calculator. No email addresses, provider connection or external requests. Values reset when you leave this page.
Fixed test clock: 29 October 2026, 12:00 UTC. This is a repeatable fixture, not the current time.
Eligible for a final production state check; no email has been sent
Inspect all 12 reproducible fixtures
- Due at the boundary: expected eligible, calculated eligible — Eligible for a final production state check; no email has been sent
- Future reminder: expected wait, calculated wait — The reminder is not due yet
- Permission withdrawn: expected skip, calculated skip — No recorded follow-up permission
- Reply before reminder: expected skip, calculated skip — A reply needs a conversation, not a reminder
- Meeting booked: expected skip, calculated skip — Booking provider handles meeting reminders
- Opt-out overrides due time: expected skip, calculated skip — Opt-out blocks follow-ups
- Cancellation stays stopped: expected skip, calculated skip — Cancellation does not restart follow-ups
- Duplicate delivery: expected skip, calculated skip — This reminder was already sent
- Timeout with unknown result: expected review, calculated review — Resolve the unknown provider outcome before retrying
- Invalid due date: expected review, calculated review — A valid due time and reference time are required
- String false is not a boolean: expected review, calculated review — Flags must be explicit booleans
- Missing preference: expected review, calculated review — Flags must be explicit booleans
“Eligible” is a decision in this calculator. A production system still needs durable state, atomic job ownership, a final check and a confirmed provider result.
Import the same example into n8n
The download is a three-node, inactive workflow with a Manual Trigger and two JavaScript Code nodes. It contains fictional fixtures and the same pure decision function used by the browser lab. There are no credentials, external HTTP requests, AI nodes, email nodes or scheduled triggers. An existing n8n account or local instance is required to import it; this download does not provide hosting.
Create an empty workflow in your n8n instance and use its import-from-file option to select the downloaded JSON. Inspect all three nodes before executing. Leave both Code nodes in Run Once for All Items mode. Execute the workflow manually, then open the output of “Decide and compare expected results”. Each of the 12 items has case, expected, action, reason and passed fields.
All passed fields should be true with the unmodified fixtures. The reference clock is fixed at 29 October 2026, 12:00 UTC so that the due-time tests stay repeatable. It is deliberately not the real clock. Change a fictional record in “Create fictional test records” to explore a failure; change its expected field only when the intended policy has changed.
- Manual Trigger
Start an explicit test run. There is no public webhook.
- Create fictional test records
Produce twelve items, each with a lead record and an expected action.
- Decide and compare
Evaluate the current flags and due time, then compare the result with the expected action.
Inactive, credential-free learning workflow. Inspect before running. No messages are sent.
The decision contract and its limits
The decision order matters. Explicit stop states come first: opt-out, reply, booking and cancellation. Missing permission and an already-sent reminder also stop the action. An uncertain provider result is held for review. Only then do valid timestamps determine whether to wait or mark the record eligible.
Inputs must contain real boolean values. The text “false” is not the boolean false. A spreadsheet or webhook that sends strings must normalize them using an explicit allowed mapping before reaching this rule. Unknown or missing values should reach review, rather than silently granting permission.
Eligible means a record passed these local checks. It does not authorize sending on its own. The example has no persistent state, authentication, rate limiting or concurrency lock. It cannot stop a second worker, validate a real provider event or recall an email. Those are separate production responsibilities.
| Output | Meaning | Production next step |
|---|---|---|
| skip | A known stop rule applies. | Record the reason and end the pending reminder. |
| wait | The valid reminder time is later than the reference clock. | Requeue and fetch fresh state when it becomes due. |
| review | Input is incomplete or delivery is uncertain. | Investigate; do not blindly retry the send. |
| eligible | The local decision checks passed. | Claim the durable job, recheck current state and use the authorized provider. |
What we tested and what you still need to test
The release includes 12 decision fixtures: exact due-time boundary, future due time, missing permission, reply, booking, opt-out, cancellation, duplicate delivery, unknown delivery outcome, invalid date, a string boolean and a missing preference. Additional checks exercise missing records, an invalid reference clock, stop-rule precedence and unchanged input records.
The downloaded JSON is also executed by a local Node.js harness that supplies the Code node’s $input.all() interface. This checks the embedded JavaScript and returned item shapes against the same fixtures. It is not a claim that this workflow was executed in n8n Cloud, connected to a CRM or tested with an email provider. Import and execute it in your n8n version before adapting it.
For a real deployment, replace the fixture node with a protected record lookup, and replace the fixed reference clock with a verified server time. Use provider-authenticated reply and booking events to update the same lead. Enforce a unique delivery key in the database, claim the job atomically and record the provider outcome separately. Test two workers attempting the same reminder, a reply arriving during send and a provider timeout after acceptance.
Do not add a production email node directly after the eligible result and assume duplicate protection is complete. A pure calculator has no durable memory. Keep a manual approval step until the connected workflow’s actual stop events, job ownership, recovery procedure and consent process have been tested.
Define the record before connecting the tools
List the fields your team needs to respond: the selected service, the prospect’s requirement, a contact route and the current owner. Then decide where the durable record lives. A CRM or database represents the lead’s business state; workflow execution history represents runs of the automation.
For a fictional agency, useful states could be new inquiry, conversation active, meeting booked and follow-ups stopped. Give each state a clear meaning. A clicked calendar link is not a confirmed meeting, and a sent email is not a completed discovery call.
Capture the inquiry before sending notifications
The website should validate the request through a protected backend before passing it to the workflow. n8n’s Webhook node can receive an event and start the process. Keep private workflow credentials on the server rather than exposing them in the browser.
Save or update the lead before attempting a team notification. If an email provider is unavailable, the inquiry should remain available for recovery. Record the notification outcome separately so a retry can send the missing notification without creating a second lead.
- Validate required fields and expected formats.
- Use an event identifier for repeat submissions.
- Save the business record before optional notifications.
- Keep credentials and administrative endpoints private.
Recheck the state before every follow-up
A scheduled reminder should look up the latest lead state immediately before sending. Someone may have replied or booked since the reminder was planned. Connect those events to the same durable record so the send decision reflects the current conversation.
Use follow-up preferences that match your contact process. If a lead opts out, stop pending messages and preserve the stop state through retries. Scheduling confirmations and marketing reminders are different purposes; letting the calendar provider manage meeting reminders can help avoid duplicate messages.
Handle booking events and duplicates explicitly
Treat a verified event from the scheduling provider as the source of booking status. Store the provider’s event identifier and map rescheduling or cancellation to the existing record. A cancellation does not automatically mean the visitor wants a new follow-up sequence.
An event can be delivered more than once. Design the receiving workflow so processing the same event again has the same business effect. This is called idempotency. For example, look up an existing booking before creating one and check whether its notification has already been sent.
Test the failure paths before launch
Test a repeated inquiry, a failed notification, a reply before a reminder and a booking arriving twice. Include a provider outage and a cancellation. These scenarios show whether the workflow can recover without losing the lead or contacting someone unnecessarily.
n8n provides separate test and production webhook URLs. Keep sample events in the test workflow until the connected providers, permissions and stop conditions are reviewed. Your launch checklist should name who receives failure alerts and who can pause a workflow when its behavior needs attention.
- Repeat events do not create duplicate records or emails.
- A reply, booking or opt-out blocks the next reminder.
- Failed notifications remain visible and can be retried.
- Each connected provider has a clear owner and recovery path.
Worked example: a reply arrives before the reminder
Imagine a fictional prospect who submits an inquiry on Monday and asks for a follow-up. Save lead-001 with follow_up_allowed=true, status=new and a stable submission identifier. The team notification then fails. The inquiry still exists; the notification job can be retried independently. An email provider accepting the retry does not mean the prospect has read it.
On Tuesday, a verified incoming reply updates lead-001 to conversation_active and records a stop reason. On Wednesday, the reminder job looks up the record again. It finds the reply state and records skipped rather than sending. A workflow that only checked permission on Monday would miss this change.
A repeated reply event should preserve the same stop state. A cancellation later in the conversation should not silently restart marketing. If your business wants to offer a new appointment, define a separate, reviewed rule and check the visitor’s preference again.
| Event | Durable record | Next action |
|---|---|---|
| Inquiry accepted | Save lead and submission ID. | Queue the team notification. |
| Notification fails | Keep the lead; record the failure separately. | Retry that notification with a stable delivery key. |
| Visitor replies | Set conversation_active and stop pending follow-ups. | Route the reply to the responsible person. |
| Reminder becomes due | Read current status and preferences again. | Record skipped; do not send. |
| Same reply delivered again | Recognize the provider event ID. | Acknowledge without repeating the business action. |
Map the n8n nodes and rehearse delivery failures
A starting capture path is Webhook → validate and normalize → save or update the lead → queue notification → respond with confirmed acceptance. Use the production webhook only after the workflow is published and tested. A separate scheduled path loads due reminders, rechecks the current record and sends only eligible ones. Store provider credentials in n8n credentials or the server environment, never in website code.
For duplicate protection, use a database uniqueness rule for the submission or provider event key. A check followed by an insert without that rule can race when two copies arrive together. For notifications, keep a delivery record tied to the lead, message purpose and provider response. Use the provider’s idempotency mechanism when supported. If a timeout leaves the send outcome unknown, investigate it before retrying; a blind retry may send twice.
There is also a small race between the last state check and the external send. Claim the job, recheck the stop state as close to sending as possible, and cancel queued jobs when a stop event arrives. Do not promise that a message already accepted by an email provider can always be recalled. Test simultaneous events and document the remaining limitation for the team.
- Capture
Validate the inquiry and save the durable lead record.
- Notify
Send the team notification as a separately tracked delivery.
- Recheck
Before a due reminder, inspect replies, bookings, preferences and stop reasons.
- Send or skip
Record the provider outcome or the reason the message was skipped.
Eight sample failure and stop-condition tests for a test workflow. It contains no real prospect email list and does not send messages.
References and further reading
The planning examples and checklists are FindSolution’s guidance. Official documentation explains the underlying tools.
Found an error or a point that needs clarification? Email boss@findsolution.pro with the page link so we can review it.
Plan your first workflow.
Bring your questions, current tools and the process you want to improve.
Book a discovery call