You have paid your policyholder in full and the accident was the other driver's fault. Walk me through what your claims system now has to do about that.

Indemnify first, argue later

The insurance contract obliges you to put your policyholder back where they were, and it does not make that obligation conditional on somebody else eventually paying. So you settle the claim, and in doing so you acquire the insured's right to recover from whoever caused the loss. That is subrogation: you step into your policyholder's shoes and pursue the third party in their place, having already discharged your own obligation.

Nothing about that sequence is optional, and getting it the wrong way round is the commonest modelling error in this area. A system that waits for the recovery before recognising the payment is not a claims system; it is a debt-collection system with a policy attached. The design consequence is that a claim carries two economically opposite flows against the same file, and they belong to different parties, different periods and different balance sheets.

Four flows, not one net figure

A claim of this kind ends up carrying at least four kinds of transaction, and the temptation to collapse them into a single net cost is what destroys the file's usefulness afterwards.

• Indemnity paid. What went out to or on behalf of the insured, at the amount agreed under the terms in force at the date of loss.
• Handling expense. Adjusters, engineers, solicitors. It is a real cost of the claim, it is not indemnity, and on a subrogation file it can be substantial because you are now litigating.
• Subrogation recovery. What the third party or their insurer eventually pays you. It reduces the net cost of the claim; it is not revenue, and it is not a reversal of the payment.
• Salvage. Where the loss was total and you paid for the whole item, the damaged item is now yours. Disposing of it produces proceeds, and often disposal costs, and both belong on the claim.

Keeping these separate is what lets you answer every question anybody will later ask of the file. Gross incurred cost is what the actuary needs for pricing and reserving, because a rate has to be set for the risk of the loss occurring, not for the risk of the loss occurring and nobody being at fault. Net cost is what the accounts need. Recovery ratios are what the recoveries team is managed on. A single netted figure answers only the middle one of those.

The recovery is an asset with its own risk

Before the money arrives, an expected recovery is a receivable, and receivables have credit risk and collection risk. That is why insurers hold a recovery estimate separately from the outstanding case reserve rather than simply reducing the reserve by what they hope to get back: reducing the reserve asserts that the eventual net cost is lower, which is a statement about the loss, whereas holding a recovery asset asserts that a third party owes you money, which is a statement about a counterparty. If the at-fault party is uninsured or insolvent, only the second of those statements turns out to have been wrong.

The practical implication is that a recovery estimate needs its own ageing and its own write-off policy. A file where the recovery has been anticipated for four years and never chased is a file whose net cost is misstated, and nothing about the claim itself will reveal it.

The excess belongs to the insured

This is where a well-built system separates itself, and it is the part candidates most often miss. If your policyholder bore an excess of five hundred, they were not made whole; they are out of pocket by that amount, and the right of recovery you inherited covers their loss as well as yours. Most jurisdictions and most policy wordings resolve this by requiring the insured's uninsured loss to be reimbursed out of what is recovered, and the widely applied convention on a partial recovery is that the insured's excess is returned before the insurer retains anything.

So a partial settlement is not a proportional split of a single number. It is an allocation against a priority order, and that order has to be modelled explicitly, because it will differ by wording and jurisdiction and because getting it wrong produces the most avoidable kind of complaint: a customer who discovers you recovered from the other driver and kept their excess. Any system that treats a recovery as a single credit posted to the claim's ledger cannot express this at all.

flowchart TD
    A[Loss notified] --> B[Coverage confirmed and reserve set]
    B --> C[Insured paid net of excess]
    C --> D[Right of recovery acquired]
    D --> E[Third party pursued]
    E --> F[Partial recovery received]
    F --> G[Insured excess reimbursed first]
    G --> H[Balance retained and cession restated]

Watch the last two steps rather than the first five: the recovery arriving is the easy part, and the allocation that follows it is where the money is actually decided.

The recovery does not stay inside the claim

A recovery on a claim that was ceded to reinsurers changes what the reinsurers are owed, so the cession has to be restated in the same proportions the original payment was allocated in, and the reinsurer's share of the recovery has to be passed on. That is only possible if the original payment stayed sliceable by treaty and layer, which is the reason recoveries and reinsurance are the same data-modelling problem wearing two hats.

It also disturbs the reserving history. A recovery received three years after the accident lands in a development period a long way from the one the loss belongs to, and if it is booked in the period it arrived in, the triangles show cost falling in a period where nothing happened. The transaction has to carry the accident date it belongs to as well as the date it was received, and the actuary has to be able to build both paid-gross and paid-net views from the same transactions. Where salvage and subrogation are material, most insurers build them as their own development patterns, because recoveries develop on a different clock from the losses.

Closure is not the end

A claim is routinely closed when the insured has been paid and reopened later when the recovery is settled, or worse, kept artificially open for years so the recovery has somewhere to live. Neither is satisfactory. The cleaner model treats the recovery as its own long-lived object linked to the claim, with its own status, its own owner and its own ageing, so the claim can close on its own timetable while the pursuit continues. What must never happen is that closure discards or freezes the evidence: the original engineer's report, the liability admission, the photographs and the payment record are what make the recovery provable, and a subrogation claim abandoned because nobody could evidence it years later is a pure, self-inflicted loss.

Recovery is not the reversal of a payment. It is a separate asset arising from a right you bought when you paid your policyholder, and any design that nets the two together loses the gross cost the pricing of the product depends on.