Issue 21 · Equivalence risk
A Rating Engine Must Remember the Deductible It Priced
Premium, deductible, retention, limit, and aggregate are one economic decision spread across quoting, policy issuance, and claims. Splitting them into separate services is how a coherent risk transfer becomes a set of contradictory fields.
One system
premiums
The other
deductibles limits
The rating engine and claims system meet at the bound deductible, retention, limit, and aggregate, where the price calculation must become the payment rule.
The deductible is not a claims setting that happens to appear on a quote; it is part of the price of the promise from the first calculation.
The rating defect that costs most is not a wrong rate. It is a correct rate attached to the wrong promise. Consider a commercial liability quote that moves from a $1M to a $5M limit, applies a deductible credit, and binds on estimated payroll. The rating service has just made four connected judgments.
Which exposure creates loss.
How much loss the insured retains.
What one occurrence can cost us.
How much the policy can pay in total.
Turn those into a few declaration fields at issuance, and let claims receive them as independent settings, and we have split one economic decision into several administrative facts.
The deductible is not a claims setting that happens to appear on a quote. It is part of the price of the promise from the first calculation.
The same holds for the limit, the aggregate, the sublimit, and the retention where a product uses one. They do not become contractual after we charge the premium.
They made the premium possible.
That is an unfashionable thing to say inside a quoting programme. Product cards want a clean price and a short list of reassuring choices.
Aggregates and retentions are untidy.
They ask the buyer and the operator to understand that a stated limit is not one pot of money.
But hiding them isn’t simplification.
It moves the product model out of sight until a claim needs it.
A rate prices an exposure under terms, and the terms are half of it
A rate is not premium.
A commercial general liability rate may run per $100 of payroll, or per $1,000 of gross receipts. Property uses insured value per $100. Commercial auto is commonly per unit. Premium is what happens when the rate meets the exposure base, the class, and the terms. A restaurant and a machine shop can report identical revenue and produce different premium, because their classifications differ.
The arithmetic is the easy part.
The rate starts with expected loss per exposure.
Expected aggregate loss is expected claim count times expected claim size. That separation matters, because frequency and severity don’t move together. Fewer collisions can sit alongside more expensive repairs on sensor-laden bumpers. Then loss adjustment expense, fixed and variable expenses, and a provision for capital enter the rate. The final premium is not a generic reflection of risk. It is a calculation about a particular exposure, under a particular promise.
Limits are in that calculation before the declarations.
An increased limits factor converts a basic-limit rate to a higher-limit rate, using a ratio of limited expected values. The second million costs less, because fewer losses reach it.
The tenth costs less again.
That is why a tower exists at all.
It is also why a $5M selection is not a harmless presentation choice after the pricing work has finished.
The deductible derives differently and weighs the same.
It removes small indemnity from expected loss.
It avoids handling cost in the hundreds and low thousands. And it restores the insured’s incentive to prevent, mitigate, and not report small losses. Arrow’s deductible theorem reaches the familiar shape: nothing below a straight deductible, everything above it, subject to its assumptions. So a deductible credit states something about loss and behaviour. A claims service that can’t identify the priced deductible is not administering the rate we bound. This is where our architecture gets pleased with itself.
A rate table carries a deductible code. The quote layer renders a friendly label. The policy system prints a number. Claims keeps its own configuration table.
Those representations can all be sensible.
Independent authorities cannot.
The question isn’t whether every application holds the field.
It is whether the records establish one bound term and its rule.
The limit is more than a large number on a certificate
Per-occurrence and aggregate limits bound different variables.
A $2M per-occurrence limit caps one occurrence.
It does nothing about the number of occurrences.
A $2M each-occurrence, $4M aggregate policy isn’t repeating itself.
The aggregate caps the sum.
A product model storing only the displayed limit has removed the part that makes frequency insurable.
The construction example is deliberately plain.
A contract requires $5M per occurrence.
The contractor’s $5M aggregate is already gone on unrelated work.
The certificate still looks compliant.
A per-project aggregate endorsement restates it for that project.
That changes the promise, not the typography.
Rating and issuance need to know it exists.
Claims needs to track the resulting pool.
A certificate does neither job.
In property the same error arrives from the other direction.
A $500M property programme can carry far smaller sublimits for flood, earthquake, sewer backup, by-law coverage, and increased cost of construction.
In British Columbia, earthquake deductibles of ten per cent are ordinary.
At a $40M location that leaves the insured carrying millions before the policy pays.
The headline limit does not define the coverage.
Neither is its price.
Canada reminded us of this after the 2013 floods.
Overland water moved into named, sublimited endorsements.
A generic North American configuration can keep the word “flood,” the headline property limit, and a deductible amount, and still lose the relationship among peril, endorsement, sublimit, and provincial rules.
That is enough data to look complete.
It is not enough to pay a claim correctly.
Layered liability sharpens the requirement again.
Above roughly $5M of need, capacity is assembled as a primary, an umbrella, and excess layers.
Attachment, maintenance of underlying insurance, and drop-down decide when a layer responds.
Most excess wordings follow the primary.
Not all of them.
A non-concurrent excess layer can cover a loss at $5M and not at $15M.
A rating engine that prices higher limits and passes only a total limit downstream has preserved neither the economics nor the contract.
Retention is a payment sequence before it is an amount
I have never seen a deductible and a self-insured retention behave as one instrument.
The retained dollars can be identical.
The operating arrangement is not.
Under a deductible policy our obligation is first-dollar.
We adjust and pay the claimant, then bill the insured back up to the deductible.
That leaves an unsecured receivable underneath.
Large-deductible arrangements therefore need credit underwriting and collateral, a letter of credit or a trust.
We also control the file from the beginning, because handling the small end affects what reaches our own money.
Under a self-insured retention, the insured pays and handles the retained layer before our obligation attaches, often through a third-party administrator.
No receivable below the retention.
No control there either.
Calling both “deductible” in the catalogue isn’t harmless shorthand.
It deletes the payment sequence that tells finance whether to expect a receivable, claims whether it owns the file, and risk whether collateral belongs in the programme.
Claims-made lines make an equally good test.
Directors and officers, professional liability, cyber, and medical malpractice are generally aggregate-only, with a retention per claim.
Defence costs can erode the limit, so every dollar to defence counsel reduces what remains to settle.
D&O separates Side A, where no retention applies because the company cannot indemnify, from Side B and Side C, which carry one.
A “limit” field without its erosion and insuring-agreement rules is a numerically accurate way to be wrong.
Quebec adds a boundary our shared model cannot cross.
Civil Code provisions make defence costs our burden in addition to the amount of insurance.
So an inside-limits structure isn’t freely available there.
The calculation needs the governing policy and the local rule.
Not a default flag copied from another province.
The invoice has a provincial edge of its own
The Canadian seams don’t stop at claims.
We each maintain our own rates and classification wording.
No shared American-style loss-cost manual exists here.
Automobile rates are the exception.
They are filed and approved provincially, through FSRA in Ontario and the Automobile Insurance Rate Board in Alberta, and set by public insurers in British Columbia, Saskatchewan, and Manitoba.
Shared mathematics does not make a generic configuration sufficient.
The invoice also separates premium from charges on it.
Premium tax varies by province and by line.
Ontario adds retail sales tax on certain insurance premiums, and Quebec applies its own tax on premiums.
Broker commission, commonly a tenth to a fifth of commercial premium, is deducted from premium rather than added to it.
A quote service calling all three “price” shows a persuasive total and leaves accounting unable to explain it.
On auditable business the price has a second life.
Deposit premium runs on estimated payroll or receipts.
At expiry, an audit reconciles T4 summaries and general ledger accounts and recalculates premium on actual exposure.
Uninsured subcontract cost is generally added to the contractor’s payroll and rated at the contractor’s class.
The bound deductible and limit don’t vanish in that recalculation.
Change either one while calling it a pure exposure audit and we have changed the original bargain.
Premium then changes meaning again on our books.
Written premium is recognised at inception.
Earned premium accrues over the term, pro rata by day.
IFRS 17 replaces the older unearned-premium presentation with a liability for remaining coverage, while the Minimum Capital Test uses premium volume as a proxy for insurance risk.
The same terms that change premium reach the numbers we explain to finance and to our supervisor.
Test a bound policy, and consume the aggregate while you are there
The migration test starts with a policy that exposes the seams.
Quote a limit using the increased limits factor.
Pick a deductible or a retention, and prove who pays first.
Add an aggregate, then consume it with two occurrences.
Add a sublimit, or an eroding-defence-cost rule.
Audit the exposure after expiry.
Run the claim through the actual jurisdictional configuration.
At each point, compare four records.
The rating inputs and calculation. The bound terms and endorsements. The billing and audit record. The claims rule and remaining capacity.
They have to agree economically and contractually.
A matching premium total is necessary, not sufficient.
A test that never exhausts an aggregate, invokes a retention, or applies a sublimit is testing the pleasant part of the product.
AI helps find these relationships in old rating worksheets, forms, claim files, and configuration tables.
It spots a deductible code meaning different things in different systems.
It should not infer whether the bound arrangement is first-dollar or self-insured, whether defence erodes a limit, or whether an aggregate has been restored.
Those are contract and control decisions.
Fluency is not a reconciliation.
The upward question is short.
Can we demonstrate that the deductible and limit we bound are the terms the rating engine priced and the claims system will apply?
Better than asking whether the quoting journey is modern.
It makes the journey answer for the promise.
A price is only credible if the loss the customer retains travels with it.
What the issue establishes
- A rate is a charge per exposure unit, while premium is the resulting amount after the exposure base and modifiers are applied.
- Increased limits factors convert a basic-limit rate to a higher-limit rate using limited expected values, so the limit selected is an input to price rather than a display field.
- A deductible changes expected indemnity, claim handling incentives, and therefore premium; it must travel from rating through issuance and claims.
- A deductible and a self-insured retention can leave the same dollars with the insured but differ in who pays and controls the claim first.
- A per-occurrence limit caps severity, while an aggregate limit caps the sum of claims and must be administered separately.
- Sublimits, defence-cost erosion, and provincial rules can materially change the amount available despite a headline limit.
- An auditable policy starts with deposit premium on estimated exposure and requires the same limit and retention terms when actual exposure is reconciled.
- Provincial taxes, policy rules, and automobile rating regimes prevent a generic North American configuration from being a sufficient Canadian product model.
The unfashionable recommendation
Do not optimize quoting conversion by hiding retentions and aggregates behind simplified product cards; the product model must carry the same financial structure into claims.
Asked in the meeting
- What is the difference between an insurance rate and premium?
- A rate is a charge per unit of exposure, such as payroll per $100, receipts per $1,000, or insured value per $100. Premium is the amount produced when that rate is applied to the account's actual or estimated exposure and then modified for classification, limits, deductibles, experience, and other applicable terms. A rate cannot be audited by itself; the exposure base completes the calculation.
- Why does a higher liability limit change the price if the rate already exists?
- A basic-limit rate prices only the layer of loss it was designed to cover. An increased limits factor converts that rate to a selected higher limit using a ratio of limited expected values. Higher layers cost less per dollar because fewer losses reach them, but they still require distinct capital and pricing. The selected limit is therefore a rating input, not a document value added after price is set.
- How does a deductible change the premium?
- A deductible removes the insurer's expected payment for losses in the retained band and restores the insured's incentive to prevent, mitigate, and avoid reporting small losses. Claims handling also has a fixed cost, so eliminating small claims changes expense as well as indemnity. The premium credit reflects that altered loss and expense expectation; it is not a cosmetic discount applied after the coverage has been priced.
- Is a deductible the same thing as a self-insured retention?
- No. Under a deductible, the insurer normally adjusts and pays the claim from the first dollar, then seeks reimbursement from the insured for the deductible amount. Under a self-insured retention, the insured handles and pays the retained layer before the insurer's duty attaches. The retained dollars can be similar, but the payment order, claims control, receivable, and collateral implications are different.
- Why are per-occurrence and aggregate limits not interchangeable?
- A per-occurrence limit caps the amount payable for one occurrence, so it bounds claim severity. An aggregate limit caps the sum paid during the policy period, so it bounds claim frequency as well as severity. A policy with a $2 million each-occurrence limit and a $4 million aggregate can pay two full-limit occurrences, subject to its wording, but the two figures do different work.
- What does a per-project aggregate do that a liability certificate does not?
- A liability certificate can show a $5 million per-occurrence limit even when unrelated claims have already consumed the policy's general aggregate. A per-project aggregate endorsement restates the aggregate separately for a designated project, preventing other work from eroding that project's pool. The endorsement changes the availability of insurance for the project; the certificate alone does not establish that protection.
- Why can defence costs reduce the amount available to settle a claim?
- Claims-made policies such as directors and officers, professional liability, cyber, and medical malpractice commonly use eroding limits. Defence counsel fees reduce the same aggregate limit that pays settlements or judgments, so a defence-heavy claim can consume substantial capacity before resolution. Commercial general liability more often provides defence outside the limit, while Quebec liability insurance rules place defence costs in addition to the amount of insurance.
- Why is the headline property limit not the amount available after a flood or earthquake?
- A property program can carry a large stated limit while applying smaller sublimits to flood, earthquake, sewer backup, by-law coverage, or increased cost of construction. Earthquake and flood deductibles can also be a percentage of insured values at the affected location rather than a flat dollar amount. After the 2013 Alberta and Toronto floods, overland-water coverage commonly became a named, sublimited endorsement in Canada.
- Why must the deductible and limit be retained when an auditable policy is reconciled?
- An auditable policy begins with deposit premium calculated from estimated payroll or receipts and is recalculated at expiry using actual exposure. The bound deductible, limit, classification, and applicable modifiers remain part of the original pricing structure during that reconciliation. Recalculating exposure with a different retained layer or limit produces neither the bound price nor a defensible additional or return premium.
- What makes a Canadian insurance configuration different from a generic North American one?
- Canadian commercial insurers maintain their own rates and classification wording, while personal and commercial automobile rates are filed and approved provincially or set by public insurers in British Columbia, Saskatchewan, and Manitoba. Provincial premium taxes vary, Ontario charges 8% retail sales tax on certain insurance premiums, and Quebec charges 9% tax on insurance premiums. Policy rules also differ, including Quebec's treatment of defence costs.
- What should a company prove before a rating and claims migration goes live?
- The company should prove that a bound policy's exposure base, class, limit, deductible or retention, aggregate, sublimits, defence-cost treatment, and endorsement terms are the same inputs used by rating and the same rules used by claims. Test cases need both ordinary and exhausted-limit outcomes. Matching a premium invoice or copying declaration fields alone does not establish behavioural or contractual equivalence.
- Why is hiding deductibles and aggregates on a simplified product card risky?
- A simplified product card can improve a quoting journey while concealing terms that determine what loss the insured retains and how much capacity remains after a claim. The deductible affects expected loss and premium, and the aggregate can be eroded by earlier claims. If those terms are not explicit in the product model, issuance and claims can apply a financial structure that the quote did not visibly price.