When the Policy Admin Lives Elsewhere: Underwriting Cloud Portability Across UK MGA Capacity Arrangements

The exposure that sits behind the login screen

A UK MGA can look well Capitalised, well governed and sensibly diversified and still carry a material concentration in a single cloud tenancy. Policy administration, quote-and-bind logic, claims workflow, document stores and management information frequently sit with one hosting provider, often selected by the MGA's software vendor rather than by the MGA itself. Capacity providers reviewing delegated authority arrangements have become adept at interrogating claims authority, pricing discipline and bordereaux quality. Far fewer ask a simple question: if this relationship ended, how quickly could the data and the systems that act on it be moved elsewhere?

The answer matters because the end of a capacity arrangement is rarely clean. Run-off, novation, renewal rights and portfolio transfers all depend on records that remain accessible, intelligible and admissible. Where those records sit in a cloud environment the MGA does not control, portability stops being an IT matter and becomes an underwriting one.

Why lock-in is rarely what the contract says it is

Most hosting agreements contain an exit clause. What they do not always contain is a route to a functioning exit. The distinction is important. Portability has at least four dimensions, and capacity providers should test each separately.

Technical portability concerns whether data can be extracted in a usable format and at a usable cadence. A nightly database dump is not the same as a managed migration. Proprietary policy administration schemas, embedded rating logic and document management indices often travel poorly.

Operational portability concerns whether the MGA has the internal capability to run the book elsewhere. A team that has only ever operated inside a vendor's platform may find that knowledge of the underlying data model sits with the vendor, not the MGA.

Contractual portability concerns termination assistance, transition periods, data deletion obligations and the cost of cooperation. These are frequently drafted in general terms and priced not at all.

Regulatory portability concerns the MGA's ability to demonstrate, to the FCA and to its capacity providers, that it can meet its obligations on outsourcing, operational resilience and data protection irrespective of which provider hosts the data. Under the FCA's operational resilience expectations and the PRA's outsourcing rules for insurers, impact tolerances are tested against disruption, not against commercial exit.

The regulatory dimension capacity providers tend to miss

UK GDPR and the Data Protection Act 2018 impose obligations on the controller that do not disappear when a processor changes. International transfers, sub-processor chains and retention schedules all have to survive migration. An MGA that has committed to UK data residency may find that its vendor's disaster recovery sits in another jurisdiction, or that a group-level platform consolidates data outside the UK on a timetable the MGA does not control.

There is also the question of regulatory access. The FCA expects firms to be able to produce records promptly. If the policy administration platform is operated by a third party under a contract that limits inspection rights, or if access depends on a vendor's cooperation, the MGA's ability to respond to a supervisory request may be constrained. For capacity providers, that is a governance risk with a direct bearing on delegation.

A practical underwriting framework

The following questions are designed to be asked before signing, and revisited at each renewal.

1. Where does the data actually live? Identify primary hosting, backup and disaster recovery locations, including any sub-processors. Confirm whether the MGA has a contractual right to approve changes to those locations.
2. What is the exit format? Ask for a sample extract in the format the MGA would receive on termination. Assess whether it is complete, machine-readable and sufficient to reconstruct the book without vendor assistance.
3. What does the vendor commit to on exit? Transition assistance, duration, cost, deletion certification and the treatment of derived data such as management information and model outputs should be specified rather than implied.
4. Who holds the operational knowledge? Test whether the MGA's own staff can administer the book outside the vendor platform. Where they cannot, treat the vendor as a key-person dependency.
5. How is the exit funded? Establish whether a termination assistance budget exists, whether it is ring-fenced, and how it would be triggered. An unfunded exit promise is a contingent liability, not a mitigation.
6. What happens to decision rules? Where algorithmic or rules-based underwriting is embedded in the platform, confirm ownership and exportability. A model that cannot be extracted is not an asset the MGA can take to a new capacity provider.

Pricing the exit rather than assuming it

Capacity providers do not need to become technology auditors. They do need to recognise that portability risk has a financial shape. It can be expressed as a migration cost, a delay to run-off, a loss of renewal rights or, in the worst case, an inability to evidence claims handling. Each of these can be reflected in capacity terms, in the scope of delegated authority, or in the MGA's own capital and governance arrangements.

The more useful discipline is to treat cloud concentration as a named exposure in the underwriting file, alongside claims authority and pricing autonomy. An MGA that can demonstrate a tested exit route is a better counterparty than one that cannot, and the difference is usually visible in the questions the MGA is willing to answer.

The quiet test of a durable relationship

Capacity arrangements are increasingly judged on how they end as much as how they begin. A binder that assumes continuity of a hosting provider is a binder with an unexamined dependency. Asking where the data lives, how it moves and what it costs to retrieve is not a distraction from underwriting. It is underwriting, and in 2026 it belongs in the file before the signature, not after the notice.