The honest answer starts with a different question: if this system failed right now, how much work could you stand to redo?

That amount of work is your recovery point (how much data you can afford to lose, measured in time). Backup frequency is the recovery point turned into a schedule. If losing an afternoon would hurt, back up more often than once a day. If losing a week would not matter, daily is generous.

Set the rhythm system by system

One schedule for the whole office is a common mistake. Different systems change at different speeds, and losing them costs different amounts:

  • Laptops and desktops change all day as people work. Lose a laptop at 3 p.m. on a nightly schedule, and everything since last night goes with it.
  • File servers and NAS shares change through business hours, often by many people at once.
  • Line-of-business databases (accounting, practice management, dispatch, ERP) change with every transaction. A lost hour can mean re-keying orders or payments.
  • Microsoft 365 and Google Workspace change constantly, but each platform holds recent deletions for a while, which softens the gap between copies.
  • Archives of closed projects and past years barely change at all.

For each system, ask two questions. How fast does it change? What does it cost to redo the work since the last copy? The answers give you its rhythm.

What sensible schedules look like

Frequent backups used to be expensive, because each run copied everything. Modern backup copies everything once, then only the changed blocks on each later run. An hourly run on a laptop is usually small.

That changes the practical defaults:

System Sensible default Worst-case loss
Laptops and desktops File changes every hour while online About an hour of work
Servers A nightly image plus hourly file versions About an hour of files
Microsoft 365 and Google Workspace 3 copies a day Several hours, with the platform’s own bin catching recent deletions
Archives After each change, or weekly Very little, because little changes

The first three rows are the schedules we run by default. The Workstation Backup plan copies file changes every hour a laptop is online, over any internet connection, and keeps 90 days of versions. The Server Backup plan takes a nightly image of each server plus hourly file versions, and keeps 30 daily, 12 monthly and 7 yearly versions.

Why servers get two rhythms

A server needs two kinds of copy, each on its own clock.

An image backup copies the whole machine: the operating system, the apps, the settings and the permissions. It is what lets you rebuild a dead server in one step rather than reinstalling everything from memory. Images are heavy, so they run once a night, late, when nobody is working.

File versions are light. Taken every hour, they mean a deleted folder or a damaged spreadsheet comes back from the last hour, not from last night.

In a full server failure you use both: restore last night’s image, then replay the hourly file versions on top. Image backup or file backup? explains the difference in more depth.

What more frequent backups cost

More often is not free. The trade-offs are real, if usually small:

  • Bandwidth. Each run sends changed data. Offices with slow uploads and large files (CAD models, video, imaging) may need the first full copy seeded on a shipped encrypted drive, with small runs after that.
  • Storage. More runs mean more versions. Deduplication keeps the growth modest, because unchanged blocks are stored once.
  • Load. Backup agents ease off when someone is working hard, but an old, busy server may still feel a heavy run in the middle of the day.
  • Noise. More jobs mean more chances for a job to fail. That only helps you if a person reads the failures.

That last point matters more than the schedule itself. An hourly backup that quietly stopped three weeks ago is worse than a nightly one that someone checks every morning. Every job we run is watched, and a failed one is followed up by a person the same business day.

When daily is enough

Hourly is not always the right answer. Daily copies are fine for:

  • a system where everyone’s work is also saved somewhere else that is backed up more often
  • a reference library or archive that changes a few times a week
  • a server whose data can be rebuilt quickly from paper, email or a supplier’s records

Be honest about the last one. “We could re-enter it” often means one person giving up a weekend.

A note on laptops: “hourly while online” means exactly that. A laptop that spends a day offline on a plane or at a remote site copies nothing until it reconnects. If people work offline for long stretches, the work done there is only on that laptop until it gets back to a connection.

Frequency is not retention

Two settings get confused. Frequency is how often you copy. Retention is how long you keep each copy.

Hourly copies kept for a day protect you from this afternoon’s mistake. They do nothing for a file that was quietly corrupted last month. You need both: frequent copies for recent work, and long retention for late discoveries. How long should you keep backups? covers the second half.

Example: schedules for a 35-person office

Consider a 35-person property management company. It runs 35 laptops, a Windows server holding the accounting database and a shared drive, a NAS of scanned leases, and Microsoft 365.

Working through the two questions for each system:

  1. Laptops. Staff draft notices and spreadsheets all day, and losing a morning is a real cost. Hourly while online.
  2. Accounting database. Rent payments post through the day, and re-keying a day of them is slow and error-prone. A nightly image plus hourly file versions, captured in a consistent state.
  3. Shared drive. It lives on the same server, so it inherits the same schedule.
  4. NAS of scanned leases. New scans arrive a few times a week, and the paper originals exist. A nightly copy is plenty, with a locked copy off site.
  5. Microsoft 365. Three copies a day, with Microsoft’s recycle bin catching same-day deletions.

Now test the plan against a bad day. A software update corrupts the accounting database at 2:40 p.m. on a Thursday.

With a nightly schedule only, the office restores Wednesday night’s copy and re-keys most of a day of payments. With hourly file versions, it restores the 2 p.m. version and re-keys 40 minutes of work. The difference between those two afternoons is the whole case for frequency.

Signs your schedule is too slow

You do not need a formal review to spot a schedule that no longer fits. Watch for these:

  • Staff keep their own copies on USB sticks or in personal cloud folders “just in case.”
  • A recent restore brought back a file older than people expected.
  • The business added a system (a new practice app, a dispatch tool) and nobody changed the backup.
  • The last backup report you saw is months old, or you have never seen one.

Any one of these is a reason to recheck the rhythm against how the business works today.

Set your own in five steps

  1. List each system and who relies on it.
  2. Write down the longest stretch of lost work you would accept for each, in hours.
  3. Set the backup interval at or below that number.
  4. Make sure someone reads failed jobs on that system every working day.
  5. Restore one item from the newest copy and check that its timestamp matches the schedule.

Step 5 is the one people skip. It tells you whether the schedule you think you have is the schedule you actually have.

If you want a structured way to see where you stand, the backup readiness check scores 12 statements across six areas, including copies, off-site storage and testing, and suggests one next step for each weak spot.

A second pair of eyes

If you would like someone to look at your schedules with you, the free backup check takes 45 minutes, remotely. We review what is copied and how often, run one test restore, and send a written page of findings. Book a free backup check when you have the time.