Sector context, because a generic design fails somebody

Regulatory pressure is the most obvious divergence and the most commonly underestimated. A financial institution has to demonstrate segregation of duties, retain evidence of who could reach what and when, and satisfy a supervisor that outsourced arrangements have been assessed. That obligation does not merely add a compliance workstream at the end; it changes how identity is designed, how logging is retained, how administrative access is granted, and how much of the configuration has to be reproducible on demand. A university carries real obligations too, but they are different ones — safeguarding, academic licensing eligibility, and the management of a user population that turns over substantially every year.

Connectivity is the second, and in this region it is decisive rather than incidental. An energy operator with remote sites on constrained or expensive links cannot run the same collaboration design as a professional services firm with a handful of well-connected offices. The correct answer for one is cloud by default; for the other it may be local caching, offline tolerance, and a deliberate decision to keep specific workloads on site. Both are competent designs. Applying either one in the wrong context produces a platform that technically works and practically frustrates everybody who uses it.

Change appetite and recovery tolerance vary at least as widely. A telecom operator running availability-critical infrastructure treats an unscheduled change as a serious event, with change control and maintenance windows to match. A growing fintech may reasonably prefer to move quickly and accept more frequent small disruption. Neither posture is wrong, but a delivery plan built for one will be rejected outright by the other — either as recklessly fast or as impossibly slow. The same applies to recovery: four hours of downtime is an inconvenience in some organisations and a production loss measured in barrels or transactions in others.

Then there is procurement, which quietly determines what is achievable at all. Public sector buyers must document options considered, justify the selection, and be able to answer questions about it years later. Budget cycles constrain not only how much can be spent but when, and a design that assumes capital availability in the wrong quarter simply does not happen. A private company can often decide in a fortnight what a ministry will take two budget rounds to approve. A design that ignores that reality is not a plan; it is a document that sits unexecuted while the underlying problem gets worse.

Identity is the first of them. Compromised credentials and over-privileged accounts cause more damage in practice than any exotic threat, and they are addressable with controls most organisations already own and have already paid for. Enforcing strong authentication everywhere, removing standing administrative rights, and running a joiner-mover-leaver process that actually completes will do more for a bank, a ministry, and a consultancy than any of them expects. It is unglamorous work and it eliminates a large share of realistic attack paths.

Recoverability is the second. Nearly every organisation has backups; considerably fewer have proven restores, and the difference only becomes visible during an incident. A job silently failing for weeks, retention shorter than anybody believed, or a repository sitting on the same network the attacker just encrypted are not exotic scenarios. They are the ordinary ones. Agreed recovery objectives, an immutable copy, and a tested restore with written evidence are what separate a bad week from a permanent loss.

The third is documentation, which sounds administrative and behaves like insurance. Environments outlive the people who built them, particularly where staff rotation is high. Design decisions, as-built configuration, and runbooks are what allow an organisation to keep operating a platform after the person who understood it has moved on — and what make an audit, an incident review, or a change of provider survivable rather than traumatic.