A little perspective goes a long way.

An unexpected age-to-age factor is a signal to investigate, not a complete explanation. Start with the data and the claims process.

Three directions to read

• Origin periods: differences in underwriting, exposure or initial loss mix.
• Development ages: settlement speed, reporting delays and tail behaviour.
• Calendar diagonals: inflation, operational changes or a claims-system conversion affecting several origin years at once.

A useful diagnostic pack

Show paid, incurred, case outstanding, claim counts and average claim amounts. Compare results with and without separately identified large claims. Document whether recoveries and allocated expenses are treated consistently. Reconcile the same valuation date across all exhibits.

Imagine incurred development accelerates while paid development barely changes. A change in case-reserve practice is one possible explanation; faster payment is another hypothesis that the paid data can challenge. Talk to claims colleagues rather than forcing both triangles to tell the same story.

Selection and communication

If recent factors are distorted, alternatives may include different averaging windows, segmentation or a method less dependent on immature experience. Record what changed, the reason, the impact on ultimate losses and what future observation would invalidate the selection.

The deliverable is not just a factor table. It is a reviewable chain from data issue to assumption to financial effect, with remaining uncertainty made explicit.

Reading sources

• CAS · Exam 5 readings and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

Reserving estimates unpaid claims for a specified valuation date. A strong close process links the accounting totals, claim-development analysis, selections and explanatory memo.

A workable sequence

1. Reconcile paid claims, case reserves and recoveries to the agreed extract.
2. Construct triangles by the appropriate origin and development periods.
3. Investigate new large losses, changing case adequacy and unusual calendar-period movements.
4. Compare methods and select ultimate claims with documented judgment.
5. Calculate unpaid amounts and explain movements to finance and reviewers.

Original Bornhuetter–Ferguson illustration

Earned premium is 10m, the selected expected loss ratio is 60%, and 60% of ultimate claims are expected to have been reported. Reported claims are 3m.

Expected unreported claims = 10m × .60 × .40 = 2.4m. BF ultimate = 3m + 2.4m = 5.4m. If paid claims are 2m, total unpaid claims are 3.4m. If case reserves are 1m, the remaining 2.4m is the model’s IBNR component on this simplified basis.

Do not subtract reported claims when the question asks for total unpaid claims: reported includes both paid and case reserves.

What is routine?

Extracts, reconciliation, exhibit refreshes and close commentary repeat. A claims-system conversion, revised segmentation or new reserving method is project work. “Routine” still involves judgment when experience changes.

Reading sources

• CAS · Exam 5 readings and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

Before fitting a loss distribution, establish what the contract pays. “Reinsurance” can mean a proportional share, a per-risk excess layer, an occurrence layer or aggregate protection. Those structures do not respond to the same loss measure.

Original layer example

A 1m excess of 1m per-loss layer attaches after the first 1m and pays up to another 1m. For a covered 3m loss, recovery is min(max(3m − 1m, 0), 1m) = 1m. The cedant retains 2m, assuming no other covers, exclusions or relevant limits.

An occurrence treaty would require aggregation according to its wording before applying the layer. Reinstatement premiums and aggregate limits can change the economic result.

What the analyst actually checks

Build gross-to-ceded-to-net reconciliations, validate attachment and exhaustion logic, and test a handful of hand-calculated losses. Examine whether cedant data are ground-up, limited or already net of another cover. Trend and development assumptions must match the basis used in the model.

A reinsurer prices risk accepted from cedants and monitors accumulation. A cedant analyses purchase options, retained volatility, cost and counterparty exposure. These are related but different assignments.

Annual treaty renewals create deadlines. New cover design, catastrophe model changes and portfolio optimisation are project work; recovery tracking and data reconciliation also recur.

Reading sources

• Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

A lower validation error is useful only if the prediction targets the right insurance quantity and uses information available at the time of the decision.

Set the target and exposure first

For a claim-count model, policy counts alone can be misleading when observation lengths differ. An exposure offset can model expected count as exposure × frequency rate. A severity model fitted only on positive claims answers a different question from a pure-premium model across all policies.

A leakage example

A field populated after a claim investigation predicts claims very well in a random split. It is unavailable when the policy is quoted. Remove it from a prospective pricing model, even if accuracy falls. Split and preprocess data in a way that respects the intended deployment timing.

The validation pack

Compare benchmark and candidate models; examine calibration and stability by meaningful segment; inspect out-of-time performance; test rating-plan implementation and document limitations. Review the effect of sparse classes, changed exposure mix and extreme predictions.

A deployable model needs data definitions, version control, implementation tests, an owner and monitoring triggers. A model that fits well but cannot be reconciled with the production rating engine is unfinished.

Routine monitoring asks whether actual experience and input distributions have changed. A new model or recalibration is a project requiring actuarial, business and applicable regulatory review. No metric alone authorises use of a rating variable.

Reading sources

• CAS · Generalized Linear Models for Insurance Rating
• FSRA · Active Ontario major auto filing guidance

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

A rate indication estimates the change needed for a defined portfolio under selected assumptions. It is not automatically the increase customers receive.

The analyst’s deliverables

A reproducible indication workbook or pipeline typically connects:

1. Earned premium and exposure reconciled to source systems.
2. Premium adjusted to a common rate level.
3. Claims developed to ultimate and trended to the future exposure period.
4. Large-loss treatment, catastrophe provision and relevant reinsurance costs.
5. Fixed and variable expenses, with a stated profit or return assumption.
6. Sensitivities and a recommendation explaining the main drivers.

A deliberately simple example

Suppose projected loss and loss-adjustment expense is 76 per exposure, current premium is 100, variable expenses are 25% of premium, and the selected margin is 5%. With no fixed expense in this toy model, required premium is 76 ÷ (1 − .25 − .05) = 108.57: an 8.57% indication.

That result does not by itself answer whether to implement 8.57%. Product teams still consider regulatory requirements, changes by segment, distribution, retention and effective dates. An implementation can differ from the indication; the decision and its expected consequences should be documented.

Routine versus project

Refreshing data, validating trends and explaining actual-versus-expected results is recurring work. Replacing the indication methodology or redesigning the rating plan is a project with testing, review and implementation dependencies.

Reading sources

• CAS · Basic Ratemaking

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

A defined-benefit pension valuation estimates benefits promised by the plan. The result depends on the purpose and applicable jurisdiction; “the pension liability” is not always one number.

Three questions to separate

• Going-concern funding: what contributions support the plan on its continuing basis under applicable rules?
• Solvency or wind-up measures: what does the specified settlement or termination basis require?
• Sponsor accounting: what expense and obligation should the sponsor report under its accounting standard?

Do not copy a discount rate from one basis into another without checking the purpose and requirements. OSFI’s actuarial-report guidance concerns federally regulated plans, not every Canadian pension plan.

The analyst’s working file

Reconcile membership and asset data, validate service and pension amounts, implement benefit provisions, calculate results and prepare gain/loss explanations. Missing retirement dates or an incorrectly coded early-retirement provision can matter more than an extra decimal place in a discount factor.

A defined-contribution arrangement shifts the work toward contributions, member outcomes, investment options and design; it does not require the same employer benefit-liability projection as a DB promise.

Routine versus project

Regular valuations and annual accounting work recur. A plan merger, benefit redesign, annuity transaction or asset strategy review is a project. Ask which valuation purposes and jurisdictions the role covers.

Reading sources

• OSFI · Defined-benefit actuarial reports
• CIA · Standards of Practice

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

Scope: Ontario private-passenger auto. A filing is the evidence package supporting proposed rates and a risk-classification system. Do not apply this process unchanged to another province or product.

The package behind the percentage

FSRA’s active major-filing guidance calls for detailed actuarial support. A useful working package traces raw experience through development, trend, premium adjustments, expenses and the proposed rating algorithm. It also reconciles the actuarial exhibits to the filing summary and explains changes from the previous approach.

In a team, an analyst may own exhibit updates, checks, assumption comparisons and a response log. Product, compliance and the signing actuary have separate responsibilities; a junior analyst does not personally decide regulatory approval.

Illustrative review scenario

A company proposes 12%. A reviewer questions whether the selected severity trend reflects a temporary claims mix change. The next step is to isolate mix, examine alternative periods and explain the selection with reproducible exhibits. A supported assumption may be defended; an unsupported assumption should be revised. A different implemented change may then need documented business and regulatory consideration.

There is no universal “this percentage always passes” shortcut. Reducing a number merely to obtain approval is not a substitute for a defensible analysis.

What the work feels like

The first submission is a project milestone. Follow-up questions can become short, deadline-driven tasks: rerun a sensitivity, reconcile an exhibit or explain a modelling decision. Good version control prevents answers from being based on different workbook versions.

Reading sources

• FSRA · Active Ontario major auto filing guidance

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

A life valuation team turns policy data, assumptions and model results into a reconciled explanation of insurance liabilities. Reporting basis matters: pricing profit, IFRS results and regulatory capital are different measures.

Work that commonly appears in a close

An analyst validates policy counts and movements, checks model inputs, compares actual experience with expectations and produces a bridge from opening to closing balances. New business, expected unwind, claims, lapses, assumption changes and model changes need distinct explanations.

Under IFRS 17, fulfilment cash flows and the contractual service margin have different meanings. The CSM represents unearned profit for applicable groups; it is not a bank account and should not be described as immediately distributable cash. The applicable measurement model determines the detailed treatment.

Original diagnostic scenario

A liability rises after a model update even though policy counts barely move. Split the effect into data changes, assumption changes and code changes using controlled comparison runs. A successful total reconciliation alone will not identify which change caused the movement.

Routine and project boundaries

Scheduled runs, validation and reconciliations recur. Product onboarding, assumption reviews, reporting changes and model migrations are projects. A junior role can involve considerable checking and documentation before it involves changing valuation assumptions.

Ask which reporting bases the team owns, who reviews the results, and whether the analyst also supports capital or experience studies.

Reading sources

• IFRS Foundation · IFRS 17 key terms
• OSFI · LICAT

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

A life pricing model projects premiums, benefits, expenses, lapses, investment effects and capital requirements over the product’s life. A quoted premium is an output; assumptions and product design determine whether that output is useful.

The work behind a new product

An analyst implements benefit rules, creates model points, validates cash flows, runs sensitivities and prepares profitability exhibits. Reviewers need to see the selected profit measure, discount basis, expense allocation and capital treatment. A product approval memo explains the result and the material risks.

A scenario worth checking

An acquisition expense is paid at issue while premiums arrive over many years. Higher early lapse can leave less premium to recover the initial cost. On another product, lapse can release costly guarantees. The direction is product-dependent: “higher lapse is good” is not a general rule.

Test mortality, expenses and lapse jointly where there is a reason for dependence. A collection of one-variable shocks can miss an interaction.

After launch

Pricing work continues through actual-versus-expected studies, distribution monitoring and repricing decisions. An experience study must distinguish changes in the risk population from changes in underlying experience.

Pricing assumptions are also not automatically valuation assumptions. Their purpose, margins, governance and reporting rules can differ. Ask how the pricing model connects to product implementation and in-force monitoring.

Reading sources

• Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

“You will build models” can describe both an occasional project and a recurring reporting process. Ask for examples of deliverables and the calendar, not only a list of software.

A realistic task map

In a pricing team, recurring tasks can include exposure checks, indication refreshes and portfolio monitoring. A project could replace a rating variable or automate a reviewed workbook. In reserving, data reconciliation and close exhibits recur; revised segmentation or a method comparison can be a project.

These are illustrative work patterns, not claims about the daily schedule at a particular employer.

What makes an analyst’s work reviewable

A reviewer should be able to identify the source extract, valuation or effective date, assumptions, code or workbook version, reconciliations and explanation of changes. A correct answer without that chain may be difficult to approve or reproduce.

For an automation task, success is more than fewer clicks. Run the new process against the reviewed old process, explain differences, document exceptions and assign an owner. Saving time while hiding a data-quality issue is not an improvement.

Questions for the team

• Which deliverables repeat monthly, quarterly or annually?
• Who reviews a new analyst’s analysis, and what does that review involve?
• How much work is an independent project versus supporting an existing process?
• Who implements the recommendation after the actuarial analysis is complete?

The mix changes with reporting deadlines, product launches and team size.

Reading sources

• CIA · Standards of Practice
• CAS · Exam 5 readings and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.