Keeping data current
Manual re-sync, scheduled refresh, and the history that tells you whether it worked.
A dataset imported from a database can be brought up to date manually or on a schedule.
Manual re-sync
Re-runs the original import — the same table or the same query — and replaces the dataset's rows. Everything built on top survives: relationships, calculated columns, measures, and every chart.
Scheduled refresh
Set a frequency (hourly, daily or weekly), a time, and a timezone. The timezone matters more than it looks: "6am" is only meaningful somewhere, and a schedule set in one zone and read in another is a classic source of "it ran at the wrong time".
Schedules are stored as your intent — frequency, hour, minute, weekday, timezone — not as a cron expression, so the form can always show you exactly what you chose.
Refresh history
Every attempt is recorded, successes and failures alike, newest first. This is the part that makes a schedule trustworthy: a schedule nobody can audit is worse than no schedule, because it invites you to assume data is current when it may not be.
A run that stays in a running state with no finish time means the job did not complete — that state is left alone deliberately rather than tidied up, because it is the signal.
What cannot be scheduled
Uploaded files. There is nothing to re-read — the file is a one-off snapshot. Attempting to schedule one is refused with a message that says so, rather than silently accepting a setting that could never run.