Every few months the same event replays with a different logo: a major online platform - this month it was one of the world's most widely used code-hosting and collaboration services - suffers a worldwide outage, and for several hours the businesses that depend on it discover exactly how much of their day requires it. Tasks halt mid-flow. Teams that could be doing something else instead refresh a status page. The outage ends, the backlog clears, and most businesses conclude nothing except that it was annoying. The more useful conclusion is that this will happen again, to some service you depend on, and that the difference between a lost afternoon and a shrugged-off inconvenience is a plan that costs almost nothing to make and cannot be made during the outage itself.
The starting point is an honest dependency map, which most small businesses have never drawn. List the external services the business genuinely runs on - email, file storage, accounting, payments, booking, communications, whichever industry-specific platforms apply - and for each one ask the simple question: if this were unreachable for four hours on a busy day, what stops? The answers are usually surprising in both directions. Some services turn out to be load-bearing for everything, including things nobody associated with them, because other tools quietly depend on them for logins or data. Others turn out to matter far less than their invoices suggest. You cannot decide where a fallback is worth having until you can see which failures actually stop revenue and which merely stop comfort.
The second element is a degraded mode: a known, written answer to what we do while it is down. Often the answer is genuinely simple, which is exactly why it is worth writing down before it is needed. If the card terminal's network is out, can you take payment another way, and does the person at the counter know it? If the booking system is dark, is there a paper or spreadsheet fallback and a way to reconcile afterwards? If email is out, does the team have another agreed channel, or does coordination simply dissolve? A degraded mode is not a second infrastructure - it is a one-page answer per critical service, plus the small preparations that make the answer true, like an exported customer contact list that is refreshed occasionally and reachable without the service that is down.
What this planning is not is a reason to flee the cloud. Major providers, outages included, are still more reliable than almost anything a small business could run itself, and the periodic public failure is the visible cost of an arrangement that is invisibly saving money every other day of the year. The point is narrower: reliability is high, not total, and the gap between high and total is your problem, not the provider's. Their status page will tell you, politely, that they are working on it. What happens in your business during those hours was decided - or not decided - well in advance. An afternoon's outage at a service you depend on is a coin flip you do not control; whether it costs you an afternoon is one you do.