How much downtime could your business afford?

Every backup decision rests on two numbers: how long the business can be without a system, and how much recent work it can lose. Published averages for the cost of downtime cannot give you either one, because they are averages of other people’s businesses. Your own payroll and sales figures can.

The two numbers

The US standards body NIST defines the recovery time objective as the length of time a system can be in recovery before it harms the organisation’s work. The recovery point objective is the point in time to which data must be recovered after an outage. Put plainly: how long you can be down, and how far back you can afford to go.

NIST has a third term, maximum tolerable downtime: how long a part of the business can be disrupted before the harm is significant. Start there. Your recovery time has to fit inside it.

What one hour costs

Pick one system, such as the server, the booking system or email, and imagine it gone at 10 am on a Tuesday. Then work through four steps. The figures in brackets are an invented example to show the working. Use your own.

1. Wages. Count the people who cannot do their job without that system, and multiply by what an hour of their time costs you, super included. (Eight people at $45 an hour is $360.)
2. Income for that hour. Divide a year’s revenue by the hours you trade in a year. ($1.2 million over 2,000 hours is $600.)
3. How much of it is gone for good. Some work can be caught up later, and some customers will not wait. Estimate the share that is lost. (If half can be caught up, $300 is lost.)
4. Add them. That is the cost of one hour. ($360 plus $300 is $660.)

Multiply by the hours in your working day and you have the cost of a day. In the example that is $5,280 for eight hours, and $15,840 for three days. The figure is rough. It only has to be good enough to set beside the cost of a backup.

Some costs will not fit in the sum: a tender deadline missed, a penalty clause in a contract, a customer who goes elsewhere and stays there. List them beside the number. Any one of them can outweigh the arithmetic.

Do it for each system

One number for the whole business hides the detail that matters. business.gov.au suggests starting an emergency plan by listing your most important or profitable products and services and how each one is delivered. Do the same here: for each service, write down the systems it needs, then run the sum for each system.

• A booking or point-of-sale system may stop income within minutes.
• Email and phones may leave staff working but customers unable to reach them.
• The accounting system may cost almost nothing for a day, unless that day is payroll or the end of the quarter.
• An archive of old jobs may not be missed for a week.

Timing changes the result, so use the worst realistic day for each system when you estimate.

From cost to recovery time

For each system, look at the cost after an hour, a day and three days, and mark the point where you would no longer accept it. That is your maximum tolerable downtime. Set the recovery time comfortably inside it, because time passes between a system failing and someone starting the restore.

The number then tells you what kind of backup you need, and what it is worth spending. If the current setup would take three days to restore and a better one would take one, the better one saves two days per outage. In the example that is $10,560. Compare it with the extra cost of the faster setup over the years you expect to keep it. A system that has to be back within a few hours usually needs a copy that is already in the office or already running somewhere else. One that can wait several days can be restored from a slower, cheaper copy.

Working out the recovery point

The recovery point is about work you would have to do twice. Suppose the backup runs once a night and the server fails at 4 pm. Everything entered since the night before is gone. Ask two questions.

• What would it cost to redo? Hours lost, times people affected, times the hourly cost. (Eight people losing a full day is 64 hours, or $2,880.)
• Can it be redone at all? Invoices can be rebuilt from emails and bank records. An order taken over the phone, a photo from a site visit or a signed form may exist nowhere else.

If a day’s work can be recreated at a cost you accept, a nightly backup is enough. If part of it cannot be recreated, that system needs copies made through the day. The Australian Signals Directorate’s advice follows the same reasoning: how often you back up depends on how often the files change and how important they are. Its Essential Eight asks for backups to be made and kept “in accordance with business criticality and business continuity requirements”. These two numbers are how a small business states those requirements.

Write it down, then test it

Put one line per system on a page: the system, the longest acceptable outage, the most work you would agree to lose, and who decided. Give it to whoever runs your backups and ask whether the current setup meets each line.

Then check the answer by timing a real restore. The steps are in A backup you have never restored is a guess. If the restore takes longer than the number on the page, either the backup changes or the number does. Either is fine, provided the owner made the choice.

For how the copies should be arranged, see the 3-2-1 backup rule. If you would like help setting the targets and building the backups to meet them, that is the disaster recovery work described under Data Management.