The loss ratio is a basis, not a number

FROM THE ARCHIVE

The loss ratio is a basis, not a number

Why insurers need to understand the losses, premium, period and reinsurance behind every ratio—and how AegisNow connects that understanding to better decisions.

When an insurance leadership team sees a loss ratio of 58.5%, the immediate question is often whether the business is performing well. My first question is different: 58.5% of what, measured how, and at which date? A precise percentage can conceal an imprecise definition. Before we use it to price a product, assess a portfolio or settle with a reinsurer, we need to understand the basis behind it.

One book. Three answers. Different business decisions.

The illustrative example in our loss-ratio paper shows why this matters. For the youngest accident year, the paid loss ratio is 30.8%, the reported loss ratio is 58.5%, and the estimated ultimate loss ratio is 83.6%. These figures describe the same book on different loss bases. They are not three competing versions of the truth.

Paid losses reflect payments already made. Reported losses, in this example, combine paid amounts with case reserves on known claims. Estimated ultimate losses also reflect expected future development, including claims incurred but not yet reported and further development on known claims. The 25.1 percentage-point difference between reported and estimated ultimate changes the conversation about the book.

Crucially, 83.6% is an estimate, not a promise about the final outcome. An actuary must assess the method, assumptions and uncertainty. A low paid ratio on a young portfolio does not, by itself, demonstrate strong underwriting performance.

Original line chart for accident years 2022–2025. For 2025, paid is 30.8%, reported is 58.5% and estimated ultimate is 83.6%. — **Figure 1. Four accident years: paid, reported and estimated ultimate**The original four-year chart, valued at 31 December 2025. Its widening gap illustrates why the age of a portfolio matters. The ultimate line represents estimates, not certain future losses.

The basis changes the pictureIllustrative youngest accident year from the supplied paper. The premium denominator is held constant; the loss measurement changes. These are not client results.

Paid loss ratio: 30.8%Payments already made
Reported loss ratio: 58.5%Paid losses plus case reserves
Estimated ultimate loss ratio: 83.6%Including expected future claim development

Reported to estimated ultimate: a 25.1 percentage-point gap. The ultimate estimate remains subject to uncertainty.

Make the definition travel with the number

I want a loss-ratio result to carry its essential context: the portfolio and period, valuation date, loss measure, claim-expense treatment, premium measure, and treatment of recoveries and reinsurance. An approved ultimate estimate should also identify its reserving run and accountable reviewer.

For example, “accident-year reported loss and allocated claim-expense ratio, on earned premium, net of reinsurance, valued at 31 December 2025” is substantially more useful than “loss ratio: 58.5%.” The underlying records should explain any additional treatment of salvage, subrogation and unallocated expenses.

This is a practical management control. Underwriting, finance, actuarial and reinsurance teams can use different views when their purposes differ. They should be able to reconcile those views without reconstructing the calculation from disconnected spreadsheets.

AegisNow architecture: five sources feed a central loss-ratio calculation, which supports seven insurance functions. — **Figure 2. Five ledgers, one governed calculation, seven uses**The architecture from the paper connects claims, policy, reserving, reinsurance and expense information through an explicitly defined calculation basis.

Three avoidable errors with material consequences

First, do not confuse portfolio growth with profitability. Written premium and earned premium answer different questions. Comparing losses with a rapidly growing written-premium denominator can produce a misleading performance signal. Select the denominator appropriate to the purpose and state it explicitly.

Second, aggregate the underlying amounts before dividing. In a simple example, a segment with premium of 100 and no losses has a 0% ratio. A second segment with premium of 1,000 and losses of 800 has an 80% ratio. The portfolio loss ratio is 800 divided by 1,100, or 72.7%—not the 40% obtained by averaging the two percentages. This example excludes expenses.

Third, keep clean business in the population and unresolved data visible. A segment with valid positive premium and no claims belongs in the analysis. A claim that cannot be matched to a policy belongs in an exception process, with its financial impact disclosed. Missing data is not the same as zero loss.

From historical experience to a responsible pricing decision

A historical loss ratio is a starting point for pricing, not a rate instruction. The CAS Basic Ratemaking text explains why prospective analysis considers loss development, trend, current premium rate levels, expenses and profit provisions. The actuary must select assumptions appropriate to the business.

Fixed and variable expenses should not be treated as interchangeable. Credibility also needs an explicit method and a defensible complement. A single claim-count threshold or an automatic assumption of “no change” is not a universal answer for every portfolio.

For leadership, the useful output is a recommendation with an explanation: what changed, which assumptions drive the result, how sensitive it is, and who has reviewed it. That makes challenge possible before a pricing decision reaches customers.

When the ratio moves money, the contract must control

A loss ratio may influence a sliding-scale commission, profit commission or aggregate protection. At that point, a definition error can become a settlement dispute. The treaty wording must determine the relevant premium, losses, expenses, periods, limits, development provisions and adjustment schedule.

Consider an illustrative treaty review. The team first identifies the covered business and contract period. It then reconciles subject premium and eligible losses, applies the agreed treatment of outstanding claims and other reinsurance, calculates the contractual adjustment, and obtains approval before settlement. Each step should retain supporting records.

An approved ultimate loss estimate does not automatically become the contractual settlement basis. Nor does every treaty use earned premium or the same treatment of IBNR. The actuarial view and the amount currently payable can differ for legitimate reasons; the system should explain the difference.

For delegated authority business, cumulative bordereaux require particular care. Adding successive cumulative snapshots can double-count amounts. The data model must distinguish a current position from movements during the period and reconcile corrections and exclusions before reporting.

Reconcile reporting views instead of assuming they are equivalent

US statutory reporting, US GAAP and IFRS 17 serve different reporting requirements. A reconciliation should identify changes in expense recognition, the revenue measure, discounting, risk adjustment and the treatment of reinsurance where relevant. A different ratio does not automatically mean a different economic outcome.

Under IFRS 17, insurance revenue excludes investment components and is presented separately from insurance finance income or expenses. It should not simply be relabelled “earned premium.” Combined-ratio comparisons also require the company’s metric definition and accounting policies; there is no single numerical conversion that fits every insurer.

The examples in this discussion principally concern P&C loss-ratio analysis. Life, annuity and health portfolios require measures suited to their products and reporting purposes. A P&C loss ratio should not be presented as a complete measure of life-insurance profitability or as the regulated medical loss ratio used for health-insurance rebates.

Illustrative waterfall: 93.0% statutory combined ratio, plus 6.0 and minus 2.0 points gives 97.0% GAAP; minus 6.0, minus 2.3 and plus 0.9 points gives 89.6% in the paper’s IFRS 17 example. — **Figure 3. Explaining the accounting bridge, step by step**The original waterfall is a simplified worked example under the paper’s assumptions. It is not a standard conversion between accounting frameworks or a statement of any insurer’s results.

AI should make the evidence easier to examine

I see a valuable role for AI in finding mismatched records, highlighting unusual movements and preparing explanations for expert review. It can reduce the time spent searching for the reason two reports disagree. The benefit depends on access to consistent, traceable information.

The calculation itself should remain reproducible. A generated explanation is not a substitute for the underlying ledger, an approved actuarial assumption or a contractual interpretation. If required inputs are missing, the workflow should identify the gap rather than silently substitute a target ratio and present it as observed experience.

Actuaries retain responsibility for methods and assumptions; finance and reinsurance professionals retain their respective approval responsibilities. Effective automation gives those people better evidence and more time to exercise judgment.

How Genellipse helps turn a ratio into an accountable decision

Our AegisNow approach connects policy, claims, reserving, reinsurance and expense information around an explicit calculation basis. The objective is a shared, governed foundation for insurance decisions: consistent source records, clearly labelled views, controlled assumptions and a traceable path from a result to its supporting evidence.

We start with a partner’s actual problem—for example, a recurring difference between underwriting and finance reports, a slow treaty reconciliation or uncertainty about a young portfolio. Together, we agree the business definition, identify authoritative data and build a reference calculation that the responsible professionals can validate.

GeneData and our integration services can help connect existing administration and finance systems, map identifiers and expose data-quality exceptions. AegisNow provides the insurance workflow context. Cortex AI OS can support governed agent workflows for evidence preparation and exception investigation, with access controls and human approval boundaries defined for the implementation.

We then validate representative cases: a loss-free segment, a late-reported claim, a reserve revision, a policy crossing reporting periods and a treaty adjustment. Applicable accounting and contract requirements are agreed with the insurer’s specialists. Success is measured through reconciliation time, unexplained differences, manual rework and evidence completeness—not an unsupported promise to improve underwriting results.

This is partnership with accountability throughout delivery. Our role is to connect technology, data and professional expertise so that the organization can operate and sustain the capability with confidence.

FROM PERSPECTIVE TO PRACTICE

Explore AegisNow capabilities

Five questions for the next portfolio review

1. What exactly is in the numerator and denominator, and how are the periods matched?

2. Is the result paid, reported or estimated ultimate—and what is the valuation date?

3. What expenses, recoveries, reinsurance and data exclusions explain the difference from other reports?

4. Can the team reproduce the result, identify the approved assumptions and explain its uncertainty?

5. Is this the correct basis for the decision we are about to make, including any contractual payment?

For me, the standard is straightforward: every important insurance number should come with enough context to support a responsible decision. That is how we move from a dashboard that looks precise to an organization that understands its risk.