Version Control Is Capital: Underwriting Proprietary Decision Rules in UK MGA Deals

Ask a UK managing general agent where its value sits and the answer rarely involves bricks. It involves a manual: a set of rating matrices, declinature rules, referral triggers and pricing judgements accumulated over years of delegated authority. Capacity providers, prospective investors and acquirers are told the same story. The difficulty is that when a diligence request lands, that story is often supported by a shared drive, three personal laptops and the recollection of a senior underwriter who is halfway out of the door.

That is an underwriting problem before it is a legal one. Proprietary underwriting know-how behaves like any other intangible asset: it has an owner, a value, a failure mode and a set of conditions under which it stops being worth what the balance sheet says it is. The task for 2026 is to price those conditions explicitly rather than assume them away.

The asset that fails the first data request

Most MGA valuations implicitly capitalise underwriting discipline. The multiple applied to a book assumes that the rules producing the loss ratio will travel with the business — to a new capacity provider, a new owner, or simply a new cohort of staff. Very little of the documentation in the market tests that assumption.

A typical diligence pack will contain a sample underwriting guide in PDF, a spreadsheet of rating factors and a description of referral authorities. It will rarely contain an assignment clause covering the guide, a version history reconciling the spreadsheet to the live platform, or evidence that the rules are actually applied on risk-by-risk decisions rather than loosely interpreted after the fact. The gap between the narrative and the evidence is where intangible exposure concentrates.

Ownership: assignment is not a synonym for access

Underwriting IP is created by people, and people create it under different legal conditions. A manual drafted by an in-house underwriting team belongs to the employer only if the contractual position says so and is properly documented. A rating model built by a contracted actuary may belong to the contractor. A set of decision rules embedded in a third-party policy administration platform may be licensed rather than owned, and may be exportable only in a form that is useless outside that platform.

Four ownership questions repay attention in any UK MGA review:

• Is there an executed assignment of intellectual property from every individual and firm that contributed to the manual, including former employees and contractors?
• Do the MGA–capacity provider agreements address ownership of underwriting materials created during the term, and do they survive termination?
• Where rules sit inside a vendor platform, what are the extraction and portability rights, and are they tested rather than assumed?
• Are there open-source components, third-party data licences or model inputs whose terms restrict commercial reuse in a sale?

None of these are exotic. All of them are routinely unresolved, and each one converts a claimed asset into a contingent liability at exactly the moment value is being realised.

Versioning: the manual with nine parents

Underwriting IP decays quietly. A rating matrix is amended for a new scheme, an override is introduced for a broker relationship, a declinature list is updated after a claims trend emerges. Within eighteen months there may be several versions in circulation, some live in the platform, some in spreadsheets used by regional teams, and some in the heads of people who no longer work there.

Version control is not an administrative nicety. It is the mechanism by which the asset can be evidenced, defended and transferred. An underwriter assessing this exposure should ask for the change log, the approval trail for each material amendment, and a reconciliation between the documented rules and the rules actually executed at the point of quote. Where that reconciliation cannot be produced, the appropriate response is not to decline the risk but to price the uncertainty it represents — and to require remediation as a condition.

Evidence: proving the rules actually bind

The most common weakness in an underwriting IP claim is not ownership but application. A manual that exists but does not govern behaviour has option value at best. The evidence test is straightforward: take a sample of bound risks across the period and demonstrate that the applied pricing, terms and referrals trace back to the documented rules, including any approved deviations.

This is also where the link to the loss record becomes credible. If the stated discipline cannot be shown to have operated, then the historical loss ratio is not evidence of underwriting skill — it is evidence of market conditions, and it should be underwritten accordingly.

Custody and leakage: what walks out of the door

Know-how is portable in a way that buildings are not. Senior underwriters carry pricing intuition, broker relationships and the tacit judgement that fills the gaps in any written manual. When they leave, the documented asset is what remains — and if the documentation is thin, so is the residual value.

Three practical controls matter. First, exit protocols that capture a structured handover of decision rationale, not merely a file transfer. Second, restrictive covenants and confidentiality terms that are proportionate, enforceable and actually aligned to the jurisdictions in which staff work. Third, a documented distinction between the rules the MGA owns and the judgement it merely employs, so that a departure is an operational event rather than a valuation event.

A five-test framework for 2026

Scores should feed directly into pricing and terms rather than sit in a diligence appendix.

What cover should actually respond to

Where insurance is used to protect this exposure — whether for the MGA itself or as part of a transaction risk solution — the wording needs to be specific about what is being insured. Legal defence costs following a misappropriation claim are one thing; the loss of value arising from an unenforceable assignment, or from the inability to extract rules from a platform on termination, is another.

Cover considerations include whether the policy responds to diminution in value of the asset rather than only to third-party claims, whether it addresses contractual warranties given on IP ownership, and whether exclusions for known defects at inception are drawn narrowly enough to be workable. As ever, the quality of the schedule matters more than the headline limit.

The conclusion for underwriters

The manual is the moat only when it can be produced. In 2026 the differentiator among UK MGAs will not be whether they claim proprietary underwriting IP, but whether they can hand an underwriter a reconciliation, an assignment and a change log without hesitation. Those that can will find the asset priced as capital. Those that cannot will find it priced as risk.