Microsoft Azure in Princeton, NJ

With the map and the claiming: every tenant, subscription, resource group and account inventoried — including the ones under personal logins — and ownership established in the organisation’s name as the first act, before any redesign. Scatter is the normal archaeology of funded work; the risk is not the scatter itself but the unowned pieces: the production workload under a departed fellow’s account, the billing nobody reviews. The inventory prices and prioritises the claiming (custody risks first, cosmetic sprawl later), so the estate becomes governable in stages, with the dangerous exhibits fixed in the first one.

It is the single most urgent finding the map can produce, and it is common enough here that the programme has a routine for it: the workload is claimed into the organisation’s tenant under controlled conditions — access replicated, data reconciled, the move validated by its real users — with the personal account’s cooperation arranged while it is still available. The risks of leaving it are concrete: the account holder’s departure, incapacity or mere unreachability becomes an institutional incident, and any diligence (a funder’s, a partner’s, an acquirer’s) will price the discovery accordingly. Fix it while it is a procedure rather than an emergency; the window for the easy version is exactly as long as the person’s goodwill and availability.

Yes — with their endings designed at their beginnings. A credit is a legitimate subsidy for real work; the discipline is tenancy and sizing: the workload built on the governed tenant from day one (so the credit funds usage, not a separate universe), sized to its real needs rather than the experiment’s ambitions, and its post-credit running cost estimated in writing before it becomes load-bearing. The monthly meter report then keeps the transition boring: when funding ends, a budget line changes — the workload does not suddenly bill an unwatched account at an unexamined size. Sprawl is not caused by credits; it is caused by credits accepted without an exit design.

Certainly — placement is decided workload by workload, by arithmetic and obligation: a specialist platform that genuinely serves a workload better stays, governed inside the same register and policy; a system whose data terms forbid the tenancy stays off cloud entirely. What this programme refuses is the platform-first answer in either direction. The landing zone’s job is to make the organisation’s Azure ground excellent for what belongs there — and to keep the map honest about what does not, so the estate’s shape is a set of recorded decisions rather than a history of accidents.

By a rhythm rather than a rescue: the meter audited monthly in the report — idle resources named and closed, sizes fitted to measured use, storage tiers matched to how often data is actually read, commitments (reservations, savings plans) bought only against proven usage patterns, never against a forecast’s enthusiasm. Budgets and alerts are set per project so a runaway experiment announces itself in days. The consolidated tenant is what makes any of this possible: one bill, one register, one team accountable for the line. Unconsolidated sprawl cannot be cost-managed; it can only be cost-discovered, usually by an auditor.

The exit is an engineering requirement from the design stage: data held in exportable forms, the estate’s configurations and policies documented as your records, the register current — so departure is a migration project like any other, executed from knowledge. Sovereignty that depends on never leaving is tenancy with good marketing, and this programme’s documentation discipline exists precisely so that your organisation’s choices remain your own, including the choice to go. Ask any Azure partner for their exit documentation on a current client estate; the fluency of the answer is the diligence.

The mapping and claiming engagement is defined and priced first — its findings usually pay for themselves in the first meter audit. Consolidation and migration are priced per wave from the map, with the destination’s running costs shown before commitment. Operations are a standing rhythm sized to the estate, reported monthly with the waste named as it is removed, so the service line defends itself in the same report that carries the platform’s bill. The comparison is the sprawl’s own ledger: unwatched subscriptions, full-rate post-credit billing, and the incident priced in a departed person’s account. One governed tenant is the cheaper estate; it is also, not coincidentally, the defensible one.