What is recovery point objective?

Backup frequency is the number in disguise. Nightly backups mean up to a day of work can disappear.

Cost rises steeply as the target shrinks. Continuous replication costs far more than a nightly copy, and the gap is not linear.

Different data deserves different answers. Losing a day of analytics is annoying; losing a day of bookings means ringing customers to ask what they had agreed.

The number should come from what the loss costs rather than from what sounds reassuring in a meeting.

Say your booking database is backed up every night at 2:00 am, and the disk fails on Wednesday at 4:40 pm. Everything entered since the backup is gone, which is 14 hours and 40 minutes of work. If your front desk took 35 bookings and 12 reschedules that day, you have 47 changes to rebuild from memory, sticky notes and confirmation texts.

Now put a price on it. Rebuilding takes one person a full day at $25 an hour, so $200, and say you miss three bookings entirely at $180 each. That's $740 for one bad Wednesday. Hourly backups would have cut the loss to 40 minutes of entries, perhaps three bookings. If the hourly option costs $30 a month more, it pays for itself the first time. At $400 a month more, nightly is the right answer. Either way, you've chosen the number instead of inheriting it.

You have one whether or not anyone wrote it down. For each system holding something you can't recreate, find two facts: how often it's backed up, and when the last good backup finished. Most cloud tools state the schedule in their docs or on a security page. For a local server, open the backup software and read the history tab. A schedule that says nightly next to a history that stops in March means your real figure is months.

Don't forget the things nobody backs up because they don't feel like data. Think of the paper diary at your front desk, or the spreadsheet on one person's laptop.

A loose target is fine when the source still exists somewhere else. Call summaries that also sit in your email can be rebuilt. Unique data is where tight targets are worth paying for, and bookings and payments come first.

GreetKeeper publishes no recovery targets, because we offer no service level agreement and would rather not print numbers we have not committed to.

The part you can control is where calls go when something fails, and that rule lives at your carrier.

Set it to your own number and test it, which protects the thing your customers notice.