Scheduled Automations: Never Forgetting Date-Driven Work
A guide to date-driven automation: renewals, payment reminders, dormant record alerts, recurring maintenance and time zone traps.
Event-driven automations run when something happens: a ticket opens, a stage changes. But a business’s most expensive losses come from things that should have happened and did not — and there is no event to trigger those. Only the calendar can.
Classic date-driven scenarios
- Contract renewal: a task for the owner at 60/30/7 days before expiry. Leaving the renewal conversation to the final week is the most common revenue loss there is.
- Payment follow-up: graduated reminders once an invoice is overdue — provided they never reach someone who has paid.
- Scheduled maintenance: by time or usage threshold, whichever comes first.
- Dormant records: a list of customers untouched for 90 days — the earliest signal of silent churn.
- Stale opportunities: an alert for quotes with no movement for 14 days.
- Trial expiry: contact before the end is far more effective than contact after it.
- Document validity: insurance, licence and certificate expiry dates.
How they differ from event-driven rules
Event-driven
Runs the moment something happens. The trigger is clear and there is no timing problem — “assign when a ticket opens”.
Scheduled
Scans all records at intervals and acts on those matching a condition — “every morning at 08:00, find invoices due today”.
The second kind is more powerful and riskier: one scan can affect thousands of records at once. A badly built scheduled rule can send thousands of messages in a single run.
Four rules for setting them up
- Do a dry run first. Before the rule performs any action, you should see the answer to “how many records will this affect?”. If you expected 12 and get 4,000, the condition is wrong.
- Set an upper bound. Cap the number of records processed in one run. That limits the damage a faulty rule can do.
- Prevent repeats. The same reminder must not reach the same record twice. Mark processed records, or add a “not sent in the last X days” condition.
- Log the outcome. When did the rule run, how many records did it process, how many errored? Without that log, problems cannot be diagnosed.
Time zones and day boundaries
This is where the most insidious errors live. If the server runs in UTC, “due today” does not match the local day, and records near midnight shift by one. The rule is clear: day boundaries must be calculated in local time, not the server’s. Send times matter too — a reminder arriving at 03:00 is worse than no reminder.
The risk of stopping silently
The most dangerous state for a scheduled job is to stop running. When an event-driven rule breaks somebody notices; when a scheduled one breaks, nothing happens — and nothing happening looks normal. So: the rule should leave a trace when it runs, and raise an alert if it has not run within the expected window.
Measurement
- Records processed per rule and how it changes over time. A sudden drop is the first sign a rule has broken.
- Reminder-to-action conversion: how many who received a renewal reminder renewed?
- Wrong-send complaints: reminders reaching people who have paid should be zero.
- Run duration: a lengthening scan means you are approaching a timeout as data grows.
Frequently asked questions
How often should they run?
As often as the work requires. Once a day is enough for payment reminders; a dormant-record scan can be weekly. Running unnecessarily often consumes resources and increases the risk of duplicates.
What about weekends and holidays?
A working-day check is needed in most scenarios. A collections reminder sent on a Saturday goes unanswered and has to be repeated on Monday; it is better for the rule to know the business calendar.
Do rules apply retroactively?
Usually not, and that should be deliberate. If a newly created rule runs against every matching historical record, the first run messages thousands of old records. Defining a start date is essential.
Next step: to see how date-driven rules are built, request a demo.