Run-Off Data Is a Policyholder Protection System

The phrase “data remediation” has a harmless sound. It suggests a conversion weekend, a queue of exceptions, a few awkward source fields. Perhaps a consultant explaining that the remaining records are almost complete. That is not what the work becomes when an insurer is in distress.

Run-off is not the moment the policy system matters less. It is the moment every missing relationship becomes a policyholder question.

Which policy is still in force? Which legal entity owes it?

Has the claim been reported? Did the claimant receive notice? Does the transfer release the old insurer?

Which treaty responds when the estate pays?

A blank field stops being our data-quality defect.

It becomes the difference between a recoverable claim and an unprovable one. So I would stop calling the run-off file a legacy conversion.

It is a policyholder-protection system.

It has to carry a promise through a transfer, a rehabilitation, a liquidation, or a long solvent run-off.

The system we inherited may be old.

The obligation is not.

The Canadian setting sharpens the point.

OSFI prudentially supervises federally regulated insurers, while provinces license transactions and handle conduct, complaints, and automobile rates. So the first question in a crisis isn’t who owns the data. It is which authority needs which evidence, for which legal function.

The intervention starts long before the headline

Insurer failure doesn’t begin with a notice.

It begins earlier, on a supervisory capital ladder.

Canadian property and casualty insurers use the Minimum Capital Test, with a supervisory target well above the regulatory minimum. Life insurers use LICAT, with its own total and core ratio targets.

Those numbers measure different things.

Together they make one operational point.

Intervention begins while an insurer is still solvent.

At the middle stages, OSFI can require a remediation plan, restrict new business, block dividends and intercompany payments, and demand more frequent reporting. The insurer may still be issuing pink slips and paying claims. Its policy records may already be wanted for a recapitalisation review, a buyer’s diligence, a transfer proposal, and a regulator’s judgment about whether the book can be administered safely. The usual response is to assemble a data room.

Necessary. Insufficient.

A data room shows us that files exist.

It cannot establish that policy record, claim record, contact record, reinsurance record, and legal entity record still describe the same obligation. Take a commercial liability claim reported under one company name after an acquisition. The policy sits in one platform. The adjuster’s notes in another. The claimant contact in an archive. The treaty bordereau names an old cedent code. We can reconcile those four into one row. A liquidator has to decide whether the policyholder has coverage, whether the estate owes the claim, and whether a reinsurer owes the estate.

The row needs lineage, not tidiness.

So here is the minimum run-off record I would define before distress. The issued promise and term. The insured and claimant contacts. Claim status, valuation, payments, reporting position. The entity obliged to pay. The applicable treaty and recoverable. The notices already sent.

The exact data model will vary.

The links cannot.

The transfer test is the counterparty, not the extract

A buyer can consume the extract we send.

That does not mean the book transferred.

In Canada, transferring a block of policies needs the Superintendent’s approval, notice to affected policyholders, and access to the actuary’s report.

A federally regulated transfer also needs ministerial approval.

Those aren’t ceremonial gates.

They are where the adequacy of our policy record becomes part of the policyholder case for the transaction.

The legal test is novation.

Assumption reinsurance novates, substitutes, and releases the original.

Indemnity reinsurance does not.

It leaves the original insurer liable to the policyholder, with a recovery from a reinsurer.

That distinction should change the transfer workplan.

A project labelling every moved record “assumed” can hide an original obligation that never moved.

In liquidation the policyholder still claims against the failed estate.

The reinsurance sits on the asset side, waiting for a liquidator to collect it.

A field mapping doesn’t change that relationship.

Canada has no solvent scheme that commutes liabilities into finality.

A block can be transferred with approval.

Absent an assuming insurer, it runs off until the last claim.

Which is the unfashionable reason to spend on recoverable records before a transaction is urgent.

There may be no later mechanism to turn uncertain history into a clean exit.

I would make a transfer team prove one thing per cohort.

Who is the direct obligor before the transaction, who is it after, and what evidence makes that true?

The project plan can keep its migration counts.

The protection system needs its counterparty proof.

Liquidation changes the clock

Rehabilitation and liquidation aren’t two words for trouble.

Rehabilitation keeps our licence and our policies, under receiver control.

Liquidation follows a finding of insolvency, creates an estate, and switches on guarantee-fund protection.

Nothing is payable by a guarantee fund because a carrier is merely under stress.

In a Canadian property and casualty liquidation, in-force coverage continues only to the earlier of policy expiry or thirty days after the order.

Thirty days is a short interval in which to discover the address is stale.

Or that the broker record is incomplete.

Or that the policy was renewed in a system the contact centre can no longer query.

On claims-made, the reporting tail is the real loss.

The contact record therefore sits beside the policy.

Not in a customer platform somebody already retired.

The liquidator notifies the policyholders and claimants we knew about, publishes notice for unknown ones, and sets a bar date for proofs of claim.

A late claim is not automatically extinguished.

It ranks behind distributions already made.

The question isn’t whether we kept a notice file.

It is whether the estate can show who was known, how they were reached, and which obligation the notice described.

The governing architecture changes too.

Insurers are excluded from the Bankruptcy and Insolvency Act and the Companies’ Creditors Arrangement Act.

A federal insurer winds up under the Winding-up and Restructuring Act, with the Insurance Companies Act supplying OSFI’s control before the court order.

The order appoints a liquidator, vests the property in the estate, stays actions against the insurer, and ends the authority of directors and officers.

This isn’t a bankruptcy workflow with an insurance adapter.

Our source systems have to become estate evidence.

Protection funds are not a substitute for the file

PACICC and Assuris get described as though they regulate an insurer, or insure every promise in full.

Neither description survives contact with the process.

PACICC is the property and casualty protection mechanism, and membership is a condition of licence.

It covers most claims to a stated ceiling, and refunds most of the unearned premium to a limit.

Not portfolio-level numbers, and several lines sit outside entirely.

Assuris serves life and health, where the preferred outcome is a transfer to a solvent insurer.

Protection there is the greater of a dollar floor or a percentage of the promise.

Union of Canada Life, in 2012, is the domestic reference case.

Neither fund examines a balance sheet, licenses an insurer, approves a rate, or takes over market conduct.

Those roles stay split.

OSFI’s mandate is prudential, not conduct.

Ontario’s FSRA is the integrated provincial case, holding licensing, conduct, and rate filings together.

Elsewhere rate authority sits with a distinct Automobile Insurance Rate Board or a utility-style tribunal.

A record saying “regulator” has already failed a routing test.

The funds need the same relationships we needed.

PACICC pays a covered claim and takes an assignment, standing in the policyholder’s priority in the estate.

So it needs a defensible policy, claim, payment, and claimant record.

The estate needs the same record to work out what it owes above the protection limit.

A guarantee arrangement absorbs part of the failure.

It doesn’t remove the need to know what failed.

Reinsurance is where the record becomes money

Our estate’s largest asset is commonly reinsurance.

It is also the relationship a modernisation programme most often models as a finance extract with a few policy keys attached.

In run-off it is a claims asset with a route.

The insolvency clause in standard treaty wording makes the reinsurer pay the liquidator on the liability allowed in the liquidation.

No reduction because the estate can’t pay in full.

The reinsurer may still assert its own defences, challenge an allowance, claim permitted set-off, or offer a commutation.

A commutation buys finality with a lump sum.

It accelerates the estate and converts uncertainty into an estimate.

Nobody decides that from a balance alone.

It needs treaty wording and period, ceding entity, policy and loss links, claim allowance, paid and outstanding amounts, notice history, balances available for set-off, and the status of each collection dispute.

Mutual debts with one counterparty may be netted.

Triangular set-off across affiliates generally may not.

The entity key is not legal decoration.

It is part of the recoverable.

The distribution order gives the problem its consequence.

Liquidation costs and preferred claims come first.

Policyholder claims rank ahead of general unsecured creditors.

Subordinated debt and shareholders follow.

Interim dividends wait on recoverables we collect.

The final dividend waits for the final commutation.

Maplex General, from 1994, and Markham General, from 2002, remind us that Canadian estates run from a few years to well over a decade.

An old treaty code nobody can explain is not an archival irritation in that setting.

It is a delay to somebody else’s payment.

Can we administer our promises if writing stops next quarter?

The board conversation isn’t about a modern platform.

It is whether this company can administer its promises if writing stops next quarter.

A contingency question.

It deserves a minimum viable answer rather than a transformation programme with a reassuring horizon.

So I would bring a small test.

We select a closed policy, an open claim, and a claim with a reinsurance recovery.

Ask the operating team to trace each one, without heroics, from policy term to contact, notice, legal entity, claim decision, payment, treaty, recoverable.

Then have a second team repeat it from source records.

Where the paths disagree is the run-off backlog.

AI helps us with this archaeology.

It classifies scanned files, compares old endorsements, surfaces claimant correspondence, and finds a treaty reference buried in a claim note.

It should not decide whether a novation occurred, whether a late claim takes priority, or whether an allowance binds a reinsurer.

Those determinations carry legal and financial consequences.

Fluent ambiguity is unattractive when the insurer is no longer there to explain it.

The useful asset isn’t a clean conversion extract.

It is a recoverable record of the promise, the person, and the party that still has to pay.