Skip to content

Verified cloud accounts across 14 providers

How to structure cloud accounts for development, staging and production

Separate accounts are the cheapest isolation boundary a cloud platform offers. Here is how teams usually divide them and what each boundary actually buys.

Most teams start with one cloud account and add environments inside it: a few tags, a naming convention, maybe separate virtual networks. It works until the first incident, at which point the boundary that mattered turns out to have been a label rather than a wall.

What an account boundary actually isolates

Across every major provider, the account or subscription is the strongest boundary the platform gives you without buying anything extra. It separates three things that are hard to separate any other way.

  • Authentication and authorisation. Identities, roles and policies are scoped to the account. A permission mistake in one account cannot grant access in another.
  • Billing. Spend is attributed at the account level by default, so cost questions stop being archaeology.
  • Service limits and quotas. Quotas are usually per account per region. A runaway process in development cannot exhaust the capacity production needs.

Tags and resource groups give you reporting. Accounts give you containment. The two are not substitutes.

The three-account baseline

The common starting structure is one account per environment: development, staging and production. Development is permissive, because the cost of a mistake is low and friction there slows the whole team down. Production is restrictive, with a small number of human identities and most changes arriving through automation. Staging sits between them and should be configured as closely to production as budget allows, because its only real job is to catch the differences.

Where teams add a fourth

Two additions come up repeatedly. A separate account for shared services, holding things like container registries, artifact storage and DNS, stops those from living in whichever environment happened to need them first. And a sandbox account per engineer, or a pool of them, removes the temptation to experiment in development.

Agencies and multi-client work

When the work belongs to different clients rather than different stages, the boundary moves. One account per client is usually correct, because the questions you will eventually be asked are per client: what did this cost, who had access, what happens when the engagement ends. An account that can be handed over or closed cleanly answers all three. A shared account with per-client tags answers none of them well.

What to decide before you create the accounts

Three decisions are much cheaper to make at the start than to retrofit.

  1. Naming. Pick a scheme that encodes purpose and environment, and use it for the account, the networks and the roles. Renaming later is possible but rarely worth the churn.
  2. Who holds the root or owner credential. It should be one documented person or a shared vault, with multi-factor authentication enabled and no day to day use.
  3. How access is granted. Even with a handful of accounts, named human users per account becomes unmanageable quickly. Decide early whether access comes from a central identity provider or from per-account roles that people assume.

The practical consequence

Account separation is why teams often need more than one cloud account rather than a larger single one. The resources inside are the same; what changes is that a failure, a permission error or an unexpected bill stays where it happened. That containment is the reason the boundary exists, and it is the main thing a second account buys you.

Chat on Telegram
Scroll to Top