The Residual Estate: A Half-Day Audit of What Cloud Migration Left On-Site
Most cloud migrations are judged by what moved. A smaller, more awkward question is what stayed. For many UK SMEs, the answer is a residual estate: a line-of-business server under a desk, a network-attached storage device holding years of archived files, a firewall VPN that has not recorded a successful login since 2023, or a backup appliance that everyone assumes was decommissioned. These items rarely appear on a migration closure report, yet they continue to consume power, attention and risk.
The purpose of this audit is not to shame an organisation for having leftovers. It is to replace benign neglect with a documented decision. A device may legitimately remain on-site in 2026, but it should do so because someone has assessed its data, users, dependencies and cost, not because no one has had the time to switch it off.
Why leftovers survive the migration
Cloud programmes tend to be organised around cutover dates. The email system moves, the finance application moves, the file shares move. The project closes when the main users are working from the new platform. What remains is often the awkward edge of the estate: an application with no vendor-supported cloud version, a local print server, a building-management system, a research or design workstation, or a storage box that was used for one department's archive and then forgotten.
Ownership is usually the missing ingredient. The migration team owns the move; the business owns the process; the managed service provider owns the support contract. The device itself may sit in a grey area between all three. A half-day audit is enough to bring those items back into view and assign a named owner.
Set the boundary before counting
The audit should be deliberately small. Two people can complete an initial pass in half a day: one with technical knowledge of the network and one with authority over business processes. The technical person can read switch ports, DHCP leases and firewall rules. The business person can confirm whether an application still matters.
Draw a simple boundary around the physical and logical estate. Include servers, virtual hosts, NAS devices, firewalls, VPN concentrators, backup appliances, uninterruptible power supplies, local admin accounts and any device with a network address that is not a standard user laptop or printer. Do not attempt a full asset-management programme. The objective is a defensible list of remnants, not a perfect inventory.
Evidence over memory
Memory is unreliable in this area. People remember the systems they use and forget the ones they inherited. Use evidence instead: network scans, DHCP and DNS records, switch MAC address tables, firewall logs, cloud identity sign-in logs, finance asset registers and physical labels on equipment. A device that appears in none of these sources but is still powered on deserves particular attention.
Interviews are still necessary, but they should be short and specific. Ask who would notice if a device were switched off tomorrow. Ask which business process would stop. Ask where the data on it came from and whether a copy exists elsewhere. These questions produce better answers than a general request to list old systems.
Classify each remnant
A useful audit sorts each item into one of four categories. Retire now applies to devices with no active users, no unique data and no dependency. Migrate then retire applies where data or a function still matters but can be moved to a supported cloud service. Retain with owner and review date applies to a genuine local requirement, such as a manufacturing system or a latency-sensitive process. Investigate applies where no one can yet confirm the purpose or data content.
The classification should be recorded with evidence. For each item, note its location, function, data classification, user count, business owner, support status and decision. A NAS holding client records is not the same as a NAS holding old marketing images. A VPN concentrator with no multi-factor authentication is not the same as an isolated test switch.
The true cost is not the electricity
The power bill for a single server or NAS is rarely the main issue. The larger costs are patching, backup, monitoring, insurance declarations, physical access and management attention. An unsupported operating system on a device that holds personal data also brings UK GDPR obligations into scope. A forgotten VPN endpoint is an entry point that may not be covered by the same monitoring as the main cloud estate.
These costs should be estimated, not exaggerated. A simple annual figure covering support time, licences, backup media, power and known risk is enough to support a decision. The point is to make the trade-off visible. Retaining a device is a choice with a cost, just as retiring it is a choice with a migration task attached.
A decision register, not a to-do list
The main output of the half-day should be a one-page decision register. Each row should carry the item, location, function, data held, users, owner, decision, next action, review date and supporting evidence. The register should be signed by a business owner, not only by IT.
This distinction matters. A to-do list can remain open indefinitely. A decision register records that someone with authority has chosen to retain, migrate or retire an item, and has accepted the consequence. Retained items should have a review date, ideally quarterly, so that a temporary exception does not become permanent by default.
A half-day agenda
A practical agenda is straightforward. Spend the first thirty minutes agreeing the boundary and sketching the network. Use the next hour to gather evidence from scans, logs and asset records. Spend an hour on short interviews with process owners. Use the next hour to classify items and estimate costs. Finish with a decision session and a named action owner for each item.
The audit does not need to produce a full decommissioning plan on the same day. It needs to produce clarity. Retirement work, data migration and disposal certificates can follow.
What good looks like afterwards
A successful audit ends with no unknown systems. Retired devices are wiped and disposed of with a certificate. Unique data is archived in a supported location. Local accounts are disabled, firewall rules are closed and DNS records are removed. The asset register is updated. The remaining on-site estate has a named owner, a support status and a review date.
A small residual estate is not automatically a failure. An undocumented one is. The half-day audit gives UK SMEs a proportionate way to find what the migration left behind and to decide, deliberately, what should stay.