Cloud or local server: where a small company's systems should live

The choice between cloud and office server is not which is better: it is who is responsible for each failure mode, decided before the night the line goes down.

Somewhere in the office there is a machine, often a tower under a desk with a label taped to it, holding the shared folder, the accounts, sometimes the invoicing. And in a browser tab somewhere there is a rented service doing roughly the same job. Most small companies carry both for years without deciding anything, and the decision announces itself only on the day one of the two arrangements fails.

The honest framing is not better and worse. It is a division of responsibility. The cloud vendor carries the hardware: the failed disk, the dead power supply, the machine that needs replacing become someone else's late night, bought with a subscription. What you carry is the dependency: the account, the renewal, the terms, and the fact that your operations sit inside someone else's business model, so their continuity becomes your continuity.

The local server carries the mirror image: your control, your custody, and your single point of failure. The office loses power and the systems lose power with it. The roadworks crew cuts the line and the file server becomes a monument for two days. The disk dies, and if the backup was never tested, the only copy of the accounts dies with it. Nobody else is coming. That is what owning the machine means.

The failures arrive with a pattern of their own. The file server nobody could reach for two days because a crew cut the line outside. The cloud account where the only administrator was the employee who left, and the company had to negotiate with its own records. The disk that failed holding the sole copy of the accounts. Each one is a responsibility that was never assigned, a failure mode nobody wrote a name on.

The practice is to decide per system, not per office. Ask what genuinely needs to be reachable from outside, and what must survive the office going dark, and let each answer land where it belongs. Then put backup discipline behind whichever arrangement you pick, because the cloud is not a backup and neither is the local disk: copies live somewhere neither failure can reach. And write down, on paper, who holds the admin credentials and where recovery lives, so that one person's departure is not a change of regime.

We host our own systems and run our own funnel, and the lesson both taught us is the same: write the failure modes down while everything still works. Guidance, not guarantees: which service fits which system, the terms of any provider, and the recovery plan your operation actually needs are things to settle with whoever runs your infrastructure, per system, in writing. Decide who owns each failure before a failure decides for you.