Resilience 10 min read

When the System Goes Down: Continuity Planning You Will Actually Use

The event that stops your business is unlikely to be a flood. It is more likely to be a tool being down for six hours on your busiest day.

In short

Skip the disaster plan and prepare for the interruptions that actually occur: a tool being unavailable, a key person unreachable, a supplier failing, or data being lost. For each of the three or four things that would genuinely stop you, write down how the work gets done manually, who decides to switch, and how clients are told. One page per dependency is enough. A plan short enough to be found and read during an outage beats a thorough one that nobody opens.

Key takeaways

  • Plan for outages, not disasters. The common events are unglamorous and far more likely.
  • Identify what would stop delivery. Usually three or four dependencies, not twenty.
  • Write the manual fallback down. During an outage nobody has time to work it out.
  • Name who decides to switch to the fallback, or the decision waits for someone unavailable.
  • Test the restore, not the backup. Untested backups fail at exactly the wrong moment.

Most continuity plans prepare for the wrong thing

Business continuity planning has an image problem. The term suggests fire, flood and catastrophe, so small businesses either produce a document about evacuation routes, or reasonably conclude that the whole subject is for organisations with more to lose.

Meanwhile the events that actually interrupt small businesses are mundane and frequent:

None of these is a disaster. All of them can stop delivery for a day or more, and each has occurred to businesses of every size in the last year. A plan for these is short, cheap, and gets used, which is more than can be said for the version about floods.

Prepare for the Tuesday when the tool is down, not for the fire. The Tuesday will happen, probably this year.

Find what would actually stop you

Start with a list of dependencies and a single question: if this were unavailable for a day, what would stop?

DependencyQuestion to ask
Systems and toolsWhich one, if unavailable, would halt delivery rather than inconvenience it?
PeopleWho, if unreachable for a week, would stop work? This is your key person list
Suppliers and subcontractorsWhich have no alternative you could reach this week?
DataWhat would be genuinely unrecoverable if it were lost tomorrow?
MoneyHow long could you operate if incoming payments stopped for a month?

Most small businesses find three or four dependencies that genuinely qualify, and a longer list of things that would be irritating but survivable. Sorting one from the other is the useful part, because it makes the planning finite. A plan covering twenty dependencies will not be written; a plan covering four will.

Sort the qualifying items by how quickly the pain arrives. Something that stops delivery within two hours needs a different level of preparation from something that becomes a problem after a fortnight.

Write the manual fallback

For each dependency that matters, document how the work continues without it. This is the core of the plan and the part that gets skipped, because during normal operation it feels hypothetical.

The fallback matters most for automated processes, and the reason is uncomfortable. Automating a process removes the need to know how it was done by hand, so within a couple of years the knowledge has gone: the people who did it manually have moved on, and nobody currently employed has ever performed the steps. This is a known cost of automation, which is why a documented manual fallback belongs in the automation project itself, as noted in automation maturity.

A workable fallback page contains four things:

1

Page per critical dependency. Anything longer will not be read by a stressed person on the day, and a plan that is not read on the day has no function.

Who decides, and how clients are told

Two elements make the difference between a plan and a document, and both are about people rather than procedure.

Who decides to switch. Outages create a period of waiting: the tool might come back in ten minutes, so nobody wants to start the disruptive manual process. That wait routinely consumes half a day. Naming one person who can call it, with a stated trigger such as two hours of downtime or an unknown recovery estimate, converts a drifting judgement into a decision. Without it, the decision waits for whoever is most senior, who may be exactly the person unreachable.

How clients are told. Drafted in advance, when nobody is panicking. Who contacts them, what is said, and what is not. The principle is the same as complaint handling: silence during an interruption does more damage than the interruption, and clients who are told early are consistently more tolerant than clients who discover it themselves. The sequence in a complaint process that protects the relationship applies almost unchanged.

Backups: test the restore

Most businesses have backups. Considerably fewer have restores, and the difference only becomes apparent at the moment it matters most.

Four checks, twice a year:

That last point connects directly to access management: recovery access is exactly the kind that gets held by one person and never reviewed.

Keeping it short enough to use

The failure mode of continuity planning is thoroughness. A forty-page document is written once, filed, never updated, and not opened during the incident it was written for, because nobody can find the relevant section while a client is waiting.

Four rules keep it usable:

RuleReason
One page per dependency, no preambleReadable in five minutes under pressure
Stored somewhere reachable when systems are downA plan inside the failed system is not a plan. Printed, or on a phone, or in a separate service
Reviewed twice a year and after any real incidentDependencies change faster than plans do
Rehearsed once, cheaplyThirty minutes talking through a scenario finds the gaps a document review misses

The rehearsal is the highest-value half hour in the whole exercise. Talking through a plausible scenario out loud, with the people who would be involved, reliably surfaces two or three assumptions that do not hold: a credential nobody else has, a client contact that lives in one inbox, a fallback that depends on the system that is down.

None of this requires a formal programme. It requires knowing which three or four things would stop you, writing a page on each, and testing the restore. The Mayim Ops assessment scores resilience alongside documentation and systems, because continuity is mostly a documentation question with a deadline attached.

Frequently asked questions

What should a small business continuity plan include?

For each dependency that would stop delivery: how the work continues manually, who decides to switch to that method, how clients are informed, and how you return to normal afterwards. One page per dependency, covering the three or four that matter, is more useful than a comprehensive document.

What is a manual fallback?

A documented way of performing a process by hand when the system that normally does it is unavailable. It matters most for automated processes, because automation removes the knowledge of how the work was done before, and the people who remember eventually leave.

How often should you test backups?

Test the restore, not the backup, at least twice a year. A backup that runs successfully but cannot be restored is common and is discovered at the worst possible moment. Restore something real to a separate location and check it opens and is complete.

What are the most likely business interruptions?

A critical tool or service being unavailable for hours, a key person suddenly unreachable, a supplier or subcontractor failing to deliver, and loss or corruption of data. These are far more likely than the catastrophic events continuity plans traditionally address, and they are the ones worth preparing for.

How long should a continuity plan be?

One page per dependency. Long plans are not read during an incident, which is the only time they are needed. If it cannot be found and understood by a stressed person in five minutes, it does not function as a plan.

Know what stops if something stops

The assessment scores resilience, systems and documentation together, and identifies the dependencies that would halt delivery if they failed.

Start your assessment

No credit card. No sales call required.