How to automate customer reminders with correct dates and states

An automatic reminder should reflect the current state of the appointment or deadline it communicates. Scheduling a message is only one stage: changes, cancellations, preferences and delivery failures need handling. This guide shows how to design a flow the team can check before and after sending.

By Lyntev. Updated on .
About 5 minutes to read.

A request that stays connectedThe same reference follows the request and prevents duplicates.

Translation published on . Original Portuguese guide; source revision .

By the end, you can:

  • Define the event that triggers the reminder
  • Handle rescheduling, time zones and preferences
  • Test delivery, retries and recovery

Link each reminder to a specific event

Choose the purpose: reminding someone of a confirmed appointment, missing document or agreed deadline. Identify the record proving that state. A date written in a conversation may be an unconfirmed preference; it should not automatically trigger a message presenting the appointment as final.

Define the recipient, channel and sending condition. Distinguish service messages from promotional communications and check rules applying to the channel and customer relationship. Automation does not create permission to contact someone or turn an existing number in the database into a confirmed preference.

Store the date with the time zone that gives it meaning

An appointment at 10:00 needs an associated location or time zone, especially when the business and customer are in different countries. Google Calendar documentation distinguishes event dates, times and time zones. In implementation, maintain an unambiguous time reference and present the time in the context the recipient expects.

Also define what “one day before” means: exactly 24 hours earlier or the same time on the previous calendar day. These rules may diverge when the clocks change. For deadlines without a time, choose a specific policy instead of assuming midnight. Test clock changes and the creation of an appointment already inside the reminder window.

Follow fictitious rescheduling from start to finish

Fictitious appointment MAR-071 is confirmed for 19 November 2026 at 10:00 in the Europe/Lisbon time zone. The exercise’s rule sends a reminder 24 hours earlier: 18 November at 10:00. The appointment is then changed to 20 November at 11:00 before the first reminder goes out.

The flow invalidates the message linked to the earlier version and prepares a new one for 19 November at 11:00. Immediately before sending, it checks that the appointment remains confirmed and the version matches. If the earlier message was already sent, it cannot be erased from what the customer received: assess a clear change notification.

Follow fictitious rescheduling from start to finish
MomentRecordExpected result
Initial confirmationMAR-071, version 1Reminder for 18/11 at 10:00
ReschedulingMAR-071, version 2Version 1 invalidated; new scheduling
Before sendingCurrent version and stateSend only if the condition remains valid
CancellationCancelled stateDo not send the pending reminder

Prepare a short message that enables action

Include the service or sufficient reference, date, time and how to communicate a change. Avoid exposing sensitive information that may appear on a locked screen. If using a link, check that it opens the correct destination and that any appointment change requires appropriate verification rather than relying on a predictable identifier.

Fictitious example: “A reminder of your appointment on 20 November at 11:00, Lisboa time. If you need to change it, contact our team through the usual channel. Reference MAR-071.” Adapt wording to the service and approved conditions. AI can help prepare versions, but dates and state should come from the record; a stable message can be filled using rules.

Distinguish scheduled, sent and delivered

The provider may accept scheduling and fail later at sending. Twilio documents this difference and provides cancellation of messages still scheduled. In that service, preference changes do not automatically remove every scheduled message: the application should track and handle pending items according to the defined policy.

Retain the provider reference and track states the channel actually provides. Sent is not the same as delivered, and delivered does not prove reading or attendance. Read states, where available, depend on the channel and recipient options. Do not present “customer informed” solely on technical acceptance of the request.

Test changes close to sending time

Use controlled test recipients or simulated functions. Try cancellation before sending, rescheduling while work is queued, an invalid number, provider failure and repeating the same task. Define a key per appointment, version and reminder type to recognise repeats without preventing a legitimate new reminder.

If the provider has already received a message for immediate sending, there may be a point beyond which cancellation is impossible. Record that limit and the actual outcome. On failure, choose human review or a limited retry; do not automatically switch channels without checking that contact is appropriate and permitted.

  • No reminder sent for an already cancelled appointment.
  • Two simultaneous attempts do not create two messages.
  • The old version does not reactivate after a retry.
  • The time zone and notice rule produce the expected time.
  • Failures are visible to a person responsible, with state and reference.

Measure operation before attributing commercial results

Track planned, cancelled, sent, delivered and failed messages over the same period. If assessing missed appointments or lateness, define the appointment group and other factors that changed; a reduction observed after activating reminders does not, alone, demonstrate that automation caused the entire difference. Start with evidence of operation and clarity for the customer.

Worksheet for your project

Purpose and record triggering the reminder:
Required state and event version:
Recipient, channel and confirmed preference:
Rules applying to the contact and provider:
Date, time zone and notice rule:
Approved wording and variable fields:
Retry key and provider reference:
Cancellation, rescheduling and point of no return:
Check immediately before sending:
Observed states and failure handling:
Test cases and controlled recipients:
Person responsible, metrics and next review:

Sources and criteria

Content prepared with AI assistance and research using technical sources, with checks of the relevant examples and calculations. The references support the guide’s decisions; examples identified as fictitious let you try out the method.

Wellinton Oliveira

Written by

CEO of Lyntev

Technology with context, original analysis and sources you can consult.