Issue 14 · Equivalence risk
Claims-Made Migrations Lose the One Date That Matters
D&O and professional liability can share claims-made mechanics, but their entity, indemnification, and professional-duty structures are not interchangeable. A migration that keeps the claim date but loses the retroactive and reporting logic has preserved the wrong event.
One system
directors officers
The other
professional liability
D&O and professional-liability administration meet at claims-made chronology, where a notice, retroactive date, and reporting window decide which historical policy is entitled to answer.
For claims-made business, the most important historical date is often the one a transaction table was never designed to keep.
The cleanest claims-made migration failure is the easiest to miss. The policy record arrives on our side. The premium reconciles. The open claim count reconciles. The claim date lands in a well-named field. We all agree the conversion is complete. Then a lawyer asks whether a matter reported this year belongs to the policy issued three years ago.
For claims-made business, the date that matters is the one our transaction table was never designed to keep.
It may be the date a circumstance was reported. It may be the retroactive date carried across a change of insurer. It may be the last day of a reporting window after cancellation. The common feature is not that the date is obscure. It is that the date means something to us only against a particular wording, limit, deductible, and notice condition. Which is why D&O and professional liability share a migration conversation and not a product model. Both are commonly claims-made-and-reported. Both make notice a condition, not a courtesy we can waive. Both can attach great weight to a notice of circumstances.
Both commonly erode limits with defence costs.
Then their coverage structures diverge sharply.
D&O turns on corporate indemnification, entity coverage, and one limit shared across three sides. Professional liability turns on the work, the retroactive date, prior knowledge, and the firm’s tail.
Our data model sees a claim.
The policy sees a sequence.
The claim date is frequently the wrong event to key on
Take the simplest professional-liability chronology.
Design work in 2018. A new insurer in 2021. Cracking reported in 2023. A statement of claim in 2025. An occurrence-policy instinct looks for the policy in force in 2018.
A claims-made policy asks something else.
Was the claim first made and reported in a policy period?
Was the work after the retroactive date?
And did anybody report the 2023 facts as a circumstance? The answers can put the 2025 lawsuit onto the 2023 policy year. A notice of circumstances deems a later claim made in the year of the notice. Its limit, deductible, wording, and aggregate become the relevant ones.
The claim-received date is still useful to us.
It is simply not decisive.
The reverse sequence is worse.
If the firm knew of the circumstance in 2023 and said nothing, the 2023 policy has no notice. The 2025 policy sees facts known before its inception, and may apply the prior-knowledge condition.
The firm has not found a technicality.
It has created a gap between two policies that both have a plausible contractual answer.
The D&O version starts before litigation too.
A TSX issuer restates revenue, amends its MD&A, and meets a renewal quote with a specific-event exclusion. A valid circumstance notice fixes any later claim from those facts to the current policy year. When the secondary-market claim arrives under Ontario’s Securities Act, the date on the pleading is not the coverage history.
The policy year that received the notice is.
That is the date we lose when we map a claim record to a claim record.
What we need is not notice_date.
It is the facts reported, the addressees, proof of delivery, policy number and layer, period, wording, limit, deductible, and the link to later related claims. A timestamp without its relationships is a photograph of a control.
The retroactive date is not an inception date
The retroactive date is the earliest work a policy accepts. A claim can reach us flawlessly today and still fall outside, because the work came first. Full prior acts means no retroactive date at all.
Here is the quiet failure in a carrier change.
A new insurer offers a lower price, effective at its own inception.
If its retroactive date is that inception date, every earlier engagement has left the book.
The old policy expired.
The new policy has no contractual route back.
Call both of them continuous coverage and we have a tidy fiction.
The field alone is not enough.
It needs its origin and its continuity.
The declarations page. The prior policy reference. A full-prior-acts indication. The insured entity. Every interruption.
A merger, an acquired practice, a change of entity: all matter.
The question is not whether our customer record looks familiar.
It is whether the policy accepts that work by that insured.
I would call this behavioural equivalence, not data quality.
Give both environments the same pre-retroactive allegation.
The same post-retroactive allegation.
The same late-reported claim.
If they find different coverage paths, the conversion changed the book, whatever the completeness report says.
D&O has three payment paths, not one claim bucket
We normalise D&O beside professional liability because they share a trigger.
It should stop there.
Side A pays directors directly when the corporation does not or cannot indemnify.
Side B reimburses the corporation for indemnity it paid.
Side C covers the corporation’s own liability, normally only for a defined Securities Claim.
Three insuring agreements, one policy, usually one limit.
Not three policy types for a generic flag.
The Canadian corporate-law layer makes that operational.
Under the CBCA, indemnity depends on good faith, and on reasonable grounds to believe the conduct lawful where a penalty is enforced.
Advancement can require an undertaking to repay.
Derivative proceedings need court approval for both.
Our wording does not replace those rules.
It reacts to them.
Presumptive indemnification therefore belongs in the scenario pack.
It applies the Side B retention as though the corporation indemnified as far as law permits.
A solvent corporation cannot refuse indemnity and summon Side A by preference.
Financial insolvency and statutory inability are different states.
A platform that can’t tell them apart can’t say who pays first.
The shared limit adds another dimension.
An order-of-payments clause puts Side A loss ahead of entity loss when money runs short.
A Side A DIC layer covers individuals only, and drops down when an underlying insurer refuses, rescinds, or exhausts.
Defence costs erode the tower we sold.
A file storing only paid and outstanding misses the rule deciding which insured still has protection.
The exclusion is not the rule, and the carve-back is
The insured-versus-insured exclusion shows why a label is unsafe.
A field saying IVI = Y tells us almost nothing.
It bars claims by one of our insureds against another.
The coverage we actually sold lives in the carve-backs.
An independently maintained derivative action comes back into cover where no insured person solicited, assisted, or joined it.
Claims by a trustee, receiver, liquidator, or CCAA monitor carve back, because insolvency is when Side A may be a director’s last asset.
Employment and whistleblower claims carve back too.
The wording also decides whether employment exposure belongs in a separate policy rather than Side C.
Those are distinct tests.
Not friendly names for one exception queue.
The converted record keeps the wording version, the claimant’s capacity, the carve-back condition, and the facts applied.
Related-claims wording needs the same care.
It aggregates connected matters back to the first, fixing the retention and the limit year.
A new claim number can therefore be a new administrative file and an old insurance event.
The lesson for us is unglamorous.
Don’t automate the reading of a bare exclusion code.
Reconstruct the versioned wording, the carve-backs, the activating evidence.
AI finds notices, spots fact patterns, compares wording.
It shouldn’t decide that a shareholder acted independently, that a monitor qualifies, or that two allegations are related.
Those are coverage determinations.
They are not extraction tasks.
A tail is a reporting window, not a replacement policy
Cancellation closes our window while the exposure carries on.
The answer is an extended reporting period.
The word sounds reassuring, so our systems treat it as a coverage-extension flag.
Which is precisely how the result gets changed.
It extends only the time to report claims for pre-cancellation work.
It does not move the retroactive date.
It does not restore a used limit, add a new one, or insure new work.
Typical forms offer an automatic month or two, with purchased periods of one to six years, or unlimited.
The whole run-off shares the final year’s aggregate.
Run-off is related and not interchangeable.
A change-of-control clause converts the policy to run-off for the balance of the term plus a negotiated tail, answering for acts before closing and nothing after.
A dissolved firm needs run-off because its work stays actionable after the last policy expires.
A model saying tail = active, without purchase option, election date, period, boundaries, final aggregate, and pre-closure condition, is not storing coverage.
It is storing our optimism.
There is a Canadian reason to be exact.
No regulator approves commercial D&O wording, and none decides whether a claim is covered.
OSFI supervises federally regulated insurers for solvency.
FSRA and the AMF license and regulate conduct.
The coverage decision comes from the form, the endorsement, and the applicable corporate and insurance law.
Quebec’s insurance contracts also sit under the Civil Code rather than a common-law Insurance Act.
An American-looking tail field won’t supply what it left out.
Test the date chain before you move the book
The test isn’t whether the target lists historic claims.
It is whether it reproduces outcomes for late allegations.
So build the scenario set before converting a record.
A circumstance reported before a renewal exclusion. A claim reported after the period. Work either side of a carried-forward retroactive date. A new claim related to an old notice. A solvent company declining indemnity. A derivative claim with and without insured participation. A firm in run-off with defence costs already eroding its final aggregate.
Keep the old answer and its source.
Declarations. Wording and endorsements. Notices and proof of delivery. Policy and retroactive dates. Limit consumption. Corporate status. The trail of related matters.
Then compare the target at the same level.
Policy count, premium, open claims: reconciliation measures.
They are not evidence of behavioural equivalence.
The board hears that the migration reconciled.
Our question is narrower.
Can the new environment reproduce the outcome when the claim arrives late and the important event happened years ago?
It sounds fussy until the first late claim asks it.
What the issue establishes
- distinguish Side A, Side B, and Side C as shared-limit D&O insuring agreements
- identify insured-versus-insured exclusions and their derivative, insolvency, employment, and whistleblower carve-backs
- retain retroactive dates, notice-of-circumstance records, policy-year wording, and related-claims links
- distinguish claims-made notice from occurrence timing and broker notice from insurer notice
- test extended reporting periods and run-off for their reporting-only scope, preserved retroactive date, and shared final aggregate
- separate Canadian corporate indemnification rules from policy wording and insurance regulation
The unfashionable recommendation
Do not migrate claims-made books by balancing premiums and open claims; build scenario tests around prior acts, notice, reporting, and run-off before converting any record.
Asked in the meeting
- Does the date of the alleged error decide a claims-made claim?
- Usually not by itself. A claims-made-and-reported policy requires a claim to be first made and reported during the policy period, while the retroactive date sets the earliest wrongful act the policy will accept. A valid notice of circumstances can deem a later claim made in the policy year of that notice. Occurrence timing and claims-made chronology therefore answer different coverage questions.
- Which historical date does a claims-made migration have to preserve?
- The migration must preserve the date and evidence of every valid notice of circumstances, linked to the policy year, wording, limit, deductible, and affected facts. That notice can fix a claim reported years later to an earlier policy year. The claim-received date alone does not preserve the coverage outcome, because it does not show whether the matter was already deemed made.
- What does the retroactive date do after an insurer changes?
- The retroactive date sets the earliest professional work for which a claims-made policy responds. Work before that date is excluded even when the later claim is reported perfectly. A new professional-liability insurer that resets the retroactive date to its own inception can delete coverage for prior work. Full prior acts means there is no retroactive date.
- What happens when a firm knows about a problem but waits to report it?
- A known circumstance that is not reported can leave the firm between two policies. The expired policy does not receive the required notice, while the later policy can apply its prior-knowledge condition to facts known before inception. Reporting a circumstance during the policy period deems the resulting claim made in that policy year, with that year's wording, limit, and deductible.
- Does telling the broker count as notice to the insurer?
- Not automatically. Professional-liability policies require written notice to the claims address named in the policy, and notice to a broker is not necessarily notice to the insurer. In a D&O tower, the broker commonly circulates notice to each layer; missed notice to an excess insurer can still become a coverage defence. The migration record needs the recipient and delivery evidence, not merely an internal note.
- What are Side A, Side B, and Side C in D&O?
- Side A pays covered loss of directors and officers when the corporation does not or cannot indemnify them. Side B reimburses the corporation for indemnity it has paid. Side C covers the corporation's own liability, normally only for a defined Securities Claim in a public-company policy. The three insuring agreements commonly share one limit, so the payment order matters.
- Can a company simply refuse to indemnify and move a director to Side A?
- No. A presumptive-indemnification clause treats the corporation as having indemnified to the fullest extent permitted by law, so the Side B retention still applies when a solvent corporation merely refuses to pay. Side A is designed for legal inability to indemnify, such as insolvency or a statutory bar. The corporation's preference is not the coverage trigger.
- Why does the insured-versus-insured exclusion need its carve-backs?
- The exclusion bars claims by one insured against another to prevent collusive recovery and ordinary internal disputes. Its carve-backs can restore coverage for independently maintained derivative actions, claims by a trustee, receiver, liquidator, or CCAA monitor, and certain employment or whistleblower claims. The condition on a derivative carve-back matters because insured participation can defeat it.
- Does a tail give a firm new coverage after cancellation?
- No. An extended reporting period extends only the time to report claims arising from pre-cancellation work. It does not move the retroactive date, reinstate the limit, or cover new work. The run-off period shares the final policy year's aggregate. Typical forms offer a short automatic period, commonly 30 or 60 days, and optional periods of 12, 36, or 72 months or unlimited.
- What does run-off mean after an acquisition or a firm wind-up?
- Run-off preserves reporting for wrongful acts committed before the transaction or cessation of practice; it does not insure post-closing or post-cancellation work. A D&O change-of-control clause can convert a policy to run-off for the balance of its term and a negotiated tail. A dissolved professional firm needs run-off cover because its past work can remain actionable after its reporting window closes.
- Do defence costs reduce the limit in these policies?
- Usually yes. D&O and professional-liability policies commonly use eroding or wasting limits, so counsel and expert costs reduce the amount left for settlement or judgment. In D&O, Side A, B, and C also compete for a shared limit, and an order-of-payments clause can put Side A first. A migration must retain the applicable limit structure and aggregate consumption, not just paid indemnity.
- Why is Canadian D&O not just an American securities policy with different addresses?
- Canadian corporate indemnification begins with the governing corporate statute, including the CBCA or a provincial equivalent, while coverage is set by insurer-authored wording and insurance regulation governs the carrier. OSFI supervises federally regulated insurer solvency; FSRA and the AMF regulate provincial licensing and conduct. Ontario secondary-market claims arise under Part XXIII.1 of the Securities Act, which requires court leave.
- What should a migration test instead of counting policies and open claims?
- A migration test recreates coverage outcomes for a late claim, a reported circumstance, a related claim, a prior-acts allegation, a missed reporting deadline, a new insurer with carried-forward retroactive date, and a cancellation or merger with run-off. Each case compares the old and new policy year, wording, limit, deductible, retention, and notice evidence. Reconciled counts do not establish contractual equivalence.