Scheduled cloud flow

What is a scheduled cloud flow?

A scheduled cloud flow is a cloud flow that starts on a timetable you set. You give it a start date, a time, and a repeat, say every weekday or once a week, and the recurrence trigger creates a run at each of those moments.

This type suits periodic work that has no reliable single event behind it: gathering open tasks, checking ageing records, or sending a weekly report. The schedule only says when the flow tries to start. It does not promise that one run finishes before the next begins, or that an external system is up, so time zones, long-running work, and missed runs belong in the design.

Scheduling and time zones

Pin down which time zone the business task belongs to. A run at 08:00 local time behaves differently around daylight-saving changes than a fixed UTC schedule. Pick the frequency from the actual need and data volume; re-reading every record every minute is rarely necessary. Filter on the relevant status and a time window, or keep a watermark so the next run carries on where the last one stopped, and prevent overlapping runs when the same records must not be processed twice at once.

A worked example

Every weekday at 07:30 a scheduled cloud flow finds contracts that expire within thirty days, groups them by account manager, and produces one summary per person. It records when each contract's reminder last went out, so a restart later the same day does not send a second message. Contracts with no owner go to a central exceptions list, and if the source is briefly unreachable the run logs the error and a managed retry follows.

Scheduled, automated or instant?

Use scheduled when time or a periodic check is the starting point, automated when a system event should start the flow, and instant when a person or calling app decides on purpose. A daily search can be simpler than thousands of event runs when only a morning summary is needed, while a daily schedule is far too slow for an alert that must go out right after an incident.

Combine the types where it helps: a scheduled flow can recover records an event flow missed, as long as deduplication stops both from doing the same work. Measure how many records each run finds, processes, skips, and rejects, since a status of Succeeded alone does not prove the expected data was there, and check whether a backfill is needed after holidays or outages.