Moving to the cloud removes a great many failure modes. It does not remove the concept of failure. Every major provider has had significant outages, and when one occurs it takes down a striking proportion of the internet simultaneously — including, often, the status pages and support channels people use to find out what is happening.
The useful response is not to avoid the cloud. It is to be clear-eyed about which risks you transferred and which you kept.
What you actually bought
Moving to a cloud provider transfers hardware failure, physical security and a large amount of routine maintenance to someone considerably better resourced than you. That is a real improvement and, for most small businesses, worth it on reliability grounds alone.
What it does not transfer is dependency. You now have a single provider whose problems become your problems, whose maintenance schedule you do not control, and whose account systems can lock you out. The failure probability is lower; the failure is more total when it comes.
The shared responsibility line
Every cloud provider publishes a shared responsibility model, and almost nobody reads it. The recurring surprise is data. Providers guarantee the availability of the platform. They do not, generally, guarantee to recover data you deleted, or data an attacker deleted using your credentials, beyond a short retention window.
This is why third-party backup of cloud services exists as a product category. If your position is that Microsoft or Google holds the only copy of your business data, you have a backup strategy with a significant gap in it — one that a compromised account or a departing employee can exercise. The reasons Microsoft 365 is not a backup solution are worth understanding properly before you rely on it as one.
What redundancy means in practice
The word covers several different things, and the distinctions matter.
Redundancy within a provider means running across multiple availability zones or regions. This protects against localised failures and is often a configuration setting rather than an architectural project. It does not protect against a provider-wide control plane failure, which is the kind that makes the news.
Redundancy across providers protects against that, and is genuinely expensive and complex. For most small businesses it is not proportionate.
Redundancy of data — an independent copy of your information somewhere your primary provider does not control — is cheap, proportionate, and the one most often missing. This is the one to prioritise.
Decide what degraded operation looks like
The proportionate question for a smaller organisation is not “how do we stay fully online” but “how do we keep functioning at reduced capacity”. If your email platform is down for six hours, can staff reach customers another way? If your cloud accounting is unavailable at end of month, is there an offline record of what was outstanding?
These answers are usually procedural rather than technical, and they cost very little to prepare. They are also the difference between a frustrating morning and a stopped business.
Know before your customers do
During a cloud outage your provider’s status page is often slow to update and occasionally unreachable. Independent monitoring of your own services tells you what your customers are experiencing rather than what the provider has acknowledged, and that gap is frequently thirty minutes or more.
Knowing early does not fix the outage, but it lets you communicate first, which materially changes how customers experience it.
The exit question
One last thing worth knowing before you need it: how would you get your data out? Confirm the export path exists, in a usable format, and that someone has run it once. Providers rarely make this difficult on purpose, but discovering the process during a dispute or a price change is a poor time to learn it.