Commercial underwriting connects operational facts, loss experience, account economics, and insurer appetite to a defensible decision. This guide develops those connections through twenty original practice items and sixty substantive concepts, including worked calculations and short written account cases. You will distinguish eligibility from authority, test submission evidence, map loss pathways, compare insurance structures, and explain recommendations clearly. All numerical rules and authority limits in the cases are hypothetical assumptions. The allocation is editorial rather than an official examination weighting. Current official syllabus alignment and examination logistics remain unverified as of October 16, 2026; the exercises provide a representative subject preview, not a full-length mock examination.
What this guide covers
Practice commercial underwriting with 20 original questions, worked calculations, and 60 focused concepts.
Underwriting objectives, authority, and judgment
1. Risk appetite as an eligibility filter
Risk appetite describes the classes, activities, locations, and characteristics an insurer intends to assume. It operates before detailed price negotiation because some exposures remain unsuitable even at an attractive premium. Guidelines translate appetite into usable eligibility rules and referral boundaries. Distinguish a firm exclusion from a preference or an exception requiring approval; the wording determines the available decision. Apply appetite to actual operations and exposure relationships rather than the applicant's business label. Revenue proportions can inform exposure analysis without overriding an explicit activity restriction.
Worked example: A business calls itself a wholesaler, but its employees also manufacture pressure-containing components. The underwriter compares that activity with the insurer's manufacturing restrictions before evaluating the producer's proposed premium or the account's otherwise favorable history.
Common mistake: Treating an attractive premium as permission to accept an operation that the insurer's stated appetite excludes from consideration.
2. Multidimensional delegated authority
Delegated authority establishes what an underwriter may commit the insurer to without additional approval. Its dimensions can include limits, premium credits, classifications, territories, coverage wording, and binding conditions. An account must satisfy every applicable dimension; spare authority in one dimension cannot offset an excess in another. Referral requests approval for a specific departure and should occur before commitment. The applicable authority schedule controls the decision, so illustrative dollar amounts or percentages should never be generalized into industry rules or assumed to apply across insurers.
Worked example: An underwriter may approve a $4 million property limit but only an 8% discretionary credit. A proposed $3 million limit with a 10% credit still requires referral because the credit exceeds the separate delegated boundary.
Common mistake: Assuming that a small account or a limit below the maximum automatically places every requested coverage and pricing term within authority.
3. Expected profitability versus observed profit
Underwriting selection concerns prospective profitability, while historical profit is one imperfect observation of the loss process. Expected losses, expenses, exposure changes, and uncertainty all affect whether proposed premium is adequate. A claim-free account can still present substantial severe-loss potential, and a sound risk can experience an unusual large loss. Separate the quality of the current proposal from the luck embedded in past results. Use historical outcomes to update assumptions, then evaluate whether the coming coverage period is supportable under the insurer's objectives.
Worked example: A factory paid $150,000 premium and had no claims last year. It now adds a combustible process. The underwriter estimates the new process's expected loss and severe-loss implications instead of treating last year's apparent profit as evidence of adequate renewal pricing.
Common mistake: Equating a favorable prior loss ratio with adequate future premium despite material changes in the operations or coverage being proposed.
4. Conditional acceptance and modified terms
Conditional acceptance requires a specified fact, action, or control to be satisfied, whereas modification changes the insurance arrangement itself. A repair verification requirement addresses evidence or risk quality; a higher deductible changes the insured's retained loss. The approaches can be combined, but they solve different problems. State whether a condition must be met before quoting, binding, or a later agreed date under the applicable arrangement. A condition should have an observable completion standard, while modified terms require evaluation of their actual coverage and economic effects.
Worked example: A warehouse proposal requires sprinkler restoration before binding and offers a $25,000 deductible. The restoration requirement addresses the physical protection assumption; the deductible changes loss sharing. Accepting the deductible does not demonstrate that the sprinkler repair has been completed.
Common mistake: Treating a deductible increase as fulfillment of a physical risk-control condition when the underlying hazard and verification requirement remain unresolved.
5. Referral, declination, and unresolved evidence
Referral seeks a decision from someone with the necessary authority; declination concludes that the proposal will not be accepted on the evaluated basis. A third state, unresolved evidence, means that material facts remain insufficient for a defensible disposition. These states should not be collapsed. An eligible account requesting a higher limit may warrant referral, while an expressly prohibited operation may warrant declination under the stated rules. Missing information may justify a targeted hold rather than either result. Identify what specific fact or authority decision would change the disposition.
Worked example: An acceptable contractor requests a limit beyond the underwriter's delegation. The underwriter prepares a higher-limit referral. A different contractor performs an activity explicitly excluded by appetite, so the same referral reasoning does not automatically justify considering that account.
Common mistake: Declining every authority exception, or treating every prohibited exposure as merely an information gap that another document can resolve.
6. Anchoring on expiring terms
Anchoring gives an initial number or judgment excessive influence over later analysis. In underwriting, expiring premium, a producer's target, or an earlier acceptance memorandum can become the default even after exposures change. A practical countermeasure is to build an independent current assessment before comparing it with the prior position. Estimate present exposures, controls, and terms from their own evidence, then explain differences. Prior terms remain useful comparison data, but they should not supply an unexamined conclusion or dictate the size of an adjustment.
Worked example: A producer requests only a 3% renewal increase after a substantial expansion. The underwriter first evaluates current payroll, operations, loss expectations, and coverage. The resulting premium estimate is then compared with expiring terms, making the expansion's effect visible.
Common mistake: Beginning with the producer's requested adjustment and searching only for reasons to move slightly above or below that initial figure.
7. Confirmation bias and disconfirming tests
Confirmation bias favors evidence consistent with an existing belief and discounts evidence that challenges it. An underwriter impressed by management may accept reassuring summaries while overlooking contradictory inspection findings. A disconfirming test asks what observable evidence would make the favorable interpretation wrong. Translate that question into a focused review, such as checking overdue maintenance or sampling actual training records. This does not mean treating every adverse allegation as true. It means evaluating supportive and contradictory evidence using comparable standards of relevance, reliability, and specificity.
Worked example: Management says its maintenance program is excellent. Instead of reviewing only the written policy, the underwriter samples completed work orders and overdue items. Several unresolved critical repairs challenge the initial favorable impression and prompt a more precise assessment.
Common mistake: Collecting additional favorable references while dismissing contrary inspection details solely because they conflict with an initially positive view of management.
8. Outcome bias in prior underwriting decisions
Outcome bias evaluates a decision mainly by its result rather than by the information and reasoning available when it was made. A severe loss does not automatically prove poor underwriting, and a claim-free period does not validate unsupported acceptance. Review whether material exposures were identified, assumptions were reasonable, authority was respected, and terms addressed the recognized risk. Then use the outcome as new evidence about those assumptions. This distinction supports learning without either excusing weak analysis or unfairly condemning a defensible decision after an adverse event.
Worked example: A protected building suffers an unusual external fire despite a well-supported acceptance. Reviewers examine the original exposure assessment and controls before judging the decision. They separately assess whether the loss reveals a previously underestimated neighboring exposure that should change future analysis.
Common mistake: Calling every claim-free acceptance a good decision and every large-loss account a bad decision without examining the original evidence.
9. Scenario analysis under material uncertainty
Scenario analysis tests how a proposed decision performs under several plausible sets of assumptions. It is useful when a single expected value hides uncertainty about severity, restoration time, or control effectiveness. Define a base case and adverse cases, change explicit inputs, and identify which changes alter acceptability or required terms. Scenarios are conditional illustrations rather than probability forecasts unless credible probabilities are supplied. Their value lies in locating fragile assumptions and showing whether more information, different terms, or a different disposition is warranted.
Worked example: A manufacturer estimates equipment replacement at four months, but the supplier cannot confirm availability. The underwriter also evaluates eight-month and twelve-month interruptions, examining income exposure and alternative production. The recommendation changes if the longer interruption cannot be supported by available coverage.
Common mistake: Presenting a single optimistic restoration estimate as certain, or attaching invented probabilities to scenarios without evidence supporting those numerical likelihoods.
Submission quality and evidence evaluation
10. Named entities and submission completeness
A complete submission identifies the legal entities, operations, locations, exposure measures, requested coverage, and relevant loss history needed for the decision. Completeness is functional: a thick packet can still omit a subsidiary or an important activity. Establish which entity performs each operation and owns, leases, or controls the relevant assets. Acquisitions and shared branding make this especially important. Entity inclusion under proposed coverage depends on the actual policy arrangement, so a business name appearing in marketing material does not establish insured status or resolve the underwriting exposure.
Worked example: A group submits one trading name and consolidated sales. A subsidiary performs installation while the parent distributes goods. The underwriter requests the legal entity schedule and entity-level activities so that the proposed insureds and installation exposure can be evaluated accurately.
Common mistake: Assuming that a consolidated financial statement or shared brand automatically identifies every legal entity requiring evaluation under the proposed insurance arrangement.
11. Reconciliation of conflicting exposure measures
Exposure figures can differ because they cover different periods, entities, accounting definitions, or operations. Reconciliation identifies these boundaries before choosing a number for analysis or rating. Ask what each figure measures, its effective period, whether it is historical or projected, and whether acquisitions or discontinued operations are included. Do not average inconsistent values unless a justified estimation method calls for it. The objective is a traceable bridge from source records to the exposure basis used in the proposed coverage, with remaining estimates labeled explicitly.
Worked example: An application lists $6 million sales while the financial statement reports $7 million. The reconciliation shows that the statement includes $1 million from a sold subsidiary. Projected sales for the continuing covered business are then evaluated separately for the coming policy period.
Common mistake: Selecting the larger number as automatically safer, or averaging both figures without determining whether they describe the same entities and period.
12. Loss-run coverage and valuation boundaries
A loss run is meaningful only when its covered entities, policy periods, lines of business, and valuation date are known. Missing years or acquired operations can make an apparently favorable history incomplete. The valuation date states when paid amounts and reserves were measured, not when all claims became final. Check whether the record includes open claims, closed claims, recoveries, and expense components according to its stated definitions. These checks establish the data boundary before calculations begin; they do not independently establish ultimate losses or the adequacy of case reserves.
Worked example: A five-year loss run excludes a subsidiary purchased six months ago and is valued before two recently reported claims. The underwriter requests the subsidiary's experience and an updated valuation before describing the combined account's loss history as complete.
Common mistake: Treating five displayed policy years as a complete account history without checking entity scope, missing operations, valuation dates, and claim definitions.
13. Evidence relevance to the actual decision
Relevant evidence bears directly on the proposition that must be established. A general safety award may say little about whether a specific protection impairment has been corrected. Define the decision proposition first, then identify evidence that tests it at the appropriate location, time, and operational scope. Reliability and relevance are separate: a trustworthy document can address the wrong question. Use indirect evidence cautiously when direct evidence is unavailable, and explain the inferential gap rather than treating general reassurance as confirmation of a specific condition.
Worked example: The decision requires confirmation that a damaged fire door now closes and latches. A current safety policy is authentic but indirect. A dated inspection identifying that door and its successful functional test addresses the required condition much more directly.
Common mistake: Giving a prestigious or professionally formatted document decisive weight when its contents do not address the specific underwriting proposition being evaluated.
14. Source provenance and evidentiary reliability
Provenance identifies who created information, how it was obtained, and whether it has been altered or summarized. Reliability depends on competence, access to the facts, methods, incentives, and traceability. Firsthand technical observations may carry more weight than an unattributed summary, but source status alone does not guarantee accuracy. Check dates, identifiers, scope, and supporting records. Distinguish an original report from a producer's paraphrase of it. This allows the underwriter to evaluate what was actually established and which conclusions still depend on another person's interpretation.
Worked example: A producer summarizes an engineer's visit as 'all satisfactory.' The original report identifies one untested alarm circuit and limits its scope to accessible areas. Reviewing provenance and original wording prevents the summary from becoming broader evidence than the inspection supports.
Common mistake: Assuming that information becomes independently verified merely because several intermediaries repeat it or attach their own professional-looking cover pages.
15. Timeliness and inspection scope decay
Evidence becomes less useful when conditions change after it was collected. Its age matters together with the stability of the underlying exposure: a recent report can be obsolete after a major process change, while older construction information may remain relevant if verified unchanged. Identify events occurring after the evidence date, including acquisitions, occupancy changes, equipment additions, and protection impairments. Update the affected portions instead of applying an arbitrary universal expiration period. The key test is whether the evidence still describes the condition on which the decision relies.
Worked example: A warehouse inspection is only two months old, but the tenant has since introduced flammable-liquid storage. The underwriter requests updated occupancy and protection information because the old inspection's scope no longer represents the proposed operation, despite its recent date.
Common mistake: Treating every recent inspection as current evidence without asking whether material operations, occupancy, or protection changed after the inspection occurred.
16. Decision-focused value of information
An information request is valuable when its answer could change disposition, terms, pricing, or confidence in a material assumption. Prioritize unresolved facts with both substantial decision impact and a realistic prospect of resolution. Asking for every available document can delay analysis without reducing important uncertainty. Formulate a specific question, identify acceptable evidence, and state the decision it informs. More information is not automatically better: repeated low-relevance documents can obscure the one missing fact that determines whether an operation fits appetite or needs specialist evaluation.
Worked example: A contractor's acceptability depends on whether it performs structural demolition. Rather than requesting another general company brochure, the underwriter asks for project descriptions, contracts, and the revenue share from demolition. Those answers directly determine the applicable appetite review.
Common mistake: Requesting a large standard document bundle while leaving the single unresolved activity that could change the account's eligibility unexamined.
17. Facts, inferences, and assumptions
A documented fact is an observation or statement established within the evidence's scope. An inference is a reasoned conclusion drawn from facts, while an assumption is a proposition provisionally used without full verification. Attribute statements accurately: a manager's assertion is a fact about what the manager said, not automatic proof of the underlying condition. Label material inferences and assumptions so another reviewer can test them. The distinction improves reasoning because it exposes where a recommendation depends on evidence, interpretation, or a condition still requiring confirmation.
Worked example: A photograph shows new sprinkler piping. The underwriter records visible installation as an observation, restoration as an unverified inference, and full operational protection as an assumption requiring testing evidence. These distinctions prevent the photograph from supporting an unjustifiably broad conclusion.
Common mistake: Writing an unverified management statement as an established operational fact, then building further conclusions on it without preserving attribution or uncertainty.
18. Independent corroboration and shared-source error
Corroboration strengthens a conclusion when genuinely independent evidence supports the same proposition. Several documents copied from one application do not provide several independent confirmations. Trace the information path and compare sources capable of observing different aspects of the issue. When sources conflict, examine scope, timing, and methods rather than counting documents or choosing the most favorable account. Independence is a matter of information origin, not merely different authors or formats. A single well-supported direct observation can outweigh numerous derivative summaries that repeat the same error.
Worked example: The application, broker presentation, and renewal spreadsheet all report no installation work because they copy last year's questionnaire. Current customer contracts and job records show installation obligations. The three matching summaries do not outweigh independent evidence of actual work.
Common mistake: Counting repeated copies of one original assertion as multiple independent confirmations and allowing document volume to substitute for evidence quality.
Commercial exposure and risk-quality assessment
19. Operation-to-loss pathway mapping
Exposure mapping translates what a business does into plausible pathways from activity to injury, damage, or financial loss. Follow inputs, processing, storage, transport, installation, customer use, and waste handling. At each stage identify who or what can be harmed, how an event could occur, and which relationships extend the exposure beyond the premises. This is more informative than relying on a classification label alone. The map identifies questions and relevant coverage areas; it does not establish that every mapped loss is covered under the proposed policy wording.
Worked example: A cabinet maker buys timber, sprays finishes, delivers products, and installs them at customers' premises. Mapping those stages reveals storage fire, finishing-process hazards, delivery exposures, and installation liability that a simple 'woodworking' label would not fully describe.
Common mistake: Using the applicant's primary industry code as a complete exposure description while overlooking ancillary activities and work performed away from its premises.
20. COPE and property loss mechanisms
COPE organizes property assessment around construction, occupancy, protection, and exposure. Construction affects combustibility and resistance; occupancy describes activities and contents; protection includes systems and response resources; exposure addresses external hazards such as neighboring operations and natural events. The categories interact rather than functioning as independent checklist scores. A protection system suitable for one occupancy may be inadequate for another. Use COPE to explain a plausible loss mechanism and identify information needs, while relying on qualified technical evaluation for design questions beyond the underwriter's competence.
Worked example: A masonry building contains combustible packaging beside an unprotected neighboring warehouse. Its walls alone do not establish low fire risk. The underwriter evaluates internal fuel load, sprinkler suitability, and external fire spread together before forming a property assessment.
Common mistake: Treating favorable construction as sufficient property protection without evaluating occupancy, system suitability, and hazards originating beyond the insured building.
21. Business interruption bottlenecks
Business interruption exposure depends on how a physical event affects the ability to generate income, not solely on the value of damaged property. A low-value component or specialized machine can be a production bottleneck with a long replacement period. Examine interdependencies, inventory buffers, alternate facilities, supplier constraints, and customer recovery. Restoration time includes more than delivery of replacement equipment when installation, testing, or approvals are necessary. Translate operational downtime into financial consequences using stated assumptions and the relevant coverage definitions rather than assuming all lost revenue equals an insured loss.
Worked example: A $40,000 control unit operates a production line generating substantial weekly contribution. Replacement takes six months and no spare exists. The underwriter evaluates the line's income dependency and restart requirements rather than using the unit's modest property value as the interruption measure.
Common mistake: Assuming that small replacement cost implies small interruption exposure, or that replacement delivery automatically restores the business to normal production.
22. Products and completed-operations pathways
Products and completed-operations analysis examines harm arising after a product leaves the business or work has been completed. Important factors include intended use, foreseeable failure modes, customer population, distribution, installation, quality controls, and traceability. A component's revenue share may be small while its role in a larger system creates substantial severity. Distinguish making a product from distributing or installing it because each role introduces different control opportunities and responsibilities. Coverage treatment depends on actual policy wording and facts; identifying a liability pathway does not determine a legal outcome.
Worked example: A supplier earns little revenue from a bracket used in overhead equipment. Failure could cause the larger assembly to fall. The underwriter asks about load specifications, testing, intended applications, and batch traceability rather than measuring exposure solely by bracket sales.
Common mistake: Equating low component revenue with low liability severity while ignoring the component's function and the consequences of failure in its final application.
23. Hired and nonowned vehicle exposure
Commercial vehicle exposure can arise even when a business owns no automobiles. Hired vehicles and employees' personal vehicles used for business can create operational and potential liability relationships requiring evaluation. Identify trip purposes, distances, drivers, vehicle types, contractual arrangements, and applicable coverage needs. Commuting and business travel should not be assumed equivalent without examining the facts and policy terms. The underwriting task is to identify the activity and controls, including driver qualification and use policies, while avoiding an unsupported conclusion that ownership alone determines responsibility or coverage.
Worked example: A consulting firm reports no owned vehicles, but employees regularly drive personal cars to client sites and rent vehicles during travel. The underwriter asks about business mileage, driver screening, rental arrangements, and the intended insurance structure for those activities.
Common mistake: Recording no automobile exposure because the company owns no vehicles, without asking how employees and rented vehicles support business operations.
24. Frequency and severity as separate dimensions
Frequency measures how often losses occur relative to an exposure base; severity measures the average or distribution of loss size conditional on occurrence. Expected loss cost can be expressed as frequency multiplied by average severity when definitions and units are compatible. The same expected cost can arise from frequent small losses or rare severe losses, but the volatility and control needs differ. Analyze both dimensions before recommending terms. A reduction in claim count does not establish improved total loss cost if the remaining claims become substantially more severe.
Worked example: One fleet has ten claims averaging $5,000; another has two claims averaging $25,000. Both total $50,000, yet the first suggests recurring incident controls while the second requires closer attention to severe-event characteristics and the stability of the small sample.
Common mistake: Calling an account improved solely because claim count fell, without checking exposure volume, average severity, and the possibility of larger remaining losses.
25. Loss prevention versus loss reduction
Loss prevention acts before an event to reduce its likelihood, while loss reduction limits consequences once an event occurs. Machine guarding may prevent contact; compartmentation may restrict fire spread. Some measures affect both frequency and severity, so identify the principal causal mechanism instead of forcing a rigid label. Evaluate whether the measure addresses the account's actual loss pathway and whether it introduces dependencies, such as maintenance or reliable activation. This distinction helps explain why one control cannot automatically substitute for another that acts at a different stage.
Worked example: A warehouse improves electrical maintenance to reduce ignition opportunities and installs fire doors to restrict spread. The underwriter evaluates both mechanisms separately because better maintenance does not provide compartmentation if a fire starts from another source within the building.
Common mistake: Treating any safety expenditure as interchangeable with every other control, even when the measures act on different stages of a loss event.
26. Control effectiveness and implementation fidelity
A control's written design differs from its effectiveness in daily operation. Assess whether responsible people understand the procedure, resources are available, exceptions are handled, and performance is checked. Training attendance alone may not establish competent task execution, while a maintenance policy without completed work orders may not establish maintenance. Seek evidence matching the control's mechanism, such as observed practice, test results, or reconciled completion records. Management commitment matters through these observable implementation processes rather than through broad assurances or the mere existence of a formal manual.
Worked example: A fleet has a driver-screening policy, but a sample of new hires lacks completed checks. The underwriter distinguishes a sound written design from incomplete execution and asks how hiring authorization will be linked to verified screening completion.
Common mistake: Assigning full credit for a documented control without checking whether it covers the relevant people, equipment, locations, and actual operating practices.
27. Material change and specialist escalation
A change is material when it can alter eligibility, expected losses, severe-loss potential, coverage needs, or the reliability of an earlier assessment. Novel processes, hazardous materials, taller storage, and complex interdependencies can exceed the scope of a prior inspection. Specialist review should answer a defined technical question rather than merely add a favorable signature. Supply the specialist with the actual configuration and intended operation, then integrate the findings into the underwriting decision. A larger deductible cannot establish that a technically uncertain hazard is adequately controlled.
Worked example: A metal-parts warehouse adds battery charging and storage. The underwriter requests quantities, arrangements, battery condition, and protection details, then obtains a configuration-specific technical review. The old noncombustible-stock inspection cannot establish suitability for the new activity.
Common mistake: Using unchanged premises or a recent inspection as proof that a newly introduced process is covered by the earlier technical assessment.
Loss experience and quantitative analysis
28. Paid, outstanding, and reported incurred losses
Paid losses represent amounts already disbursed under the dataset's definition. Outstanding case reserves estimate additional amounts for reported open claims. When the stated convention is paid plus outstanding, their sum is reported incurred loss. It remains sensitive to reserve changes and excludes unreported losses unless the definition expressly includes them. Check treatment of expenses and recoveries before combining figures. An increase in incurred losses can reflect new payments, revised reserves, or new claims; the components explain why the change occurred and prevent double-counting payments already included in incurred totals.
Worked example: A claim shows $40,000 paid and $60,000 outstanding, giving $100,000 reported incurred. After a $10,000 payment, outstanding falls to $50,000 with no estimate change. Reported incurred remains $100,000 because the payment simply moves value between components.
Common mistake: Adding paid losses to a figure already labeled total incurred, or assuming that every payment necessarily increases the total estimated claim cost.
29. Loss-ratio numerator and denominator matching
A loss ratio divides a defined loss amount by the corresponding premium amount. For a policy-period analysis, losses and earned premium should describe compatible coverage, entities, and periods. State whether losses are paid, reported incurred, or estimated ultimate and whether expenses are included. Written premium can differ from earned premium, so substituting it changes interpretation. The ratio measures loss cost relative to premium, not exposure frequency or total profitability by itself. Expenses and other costs require separate treatment before drawing conclusions about underwriting contribution.
Worked example: Reported incurred losses of $150,000 divided by $300,000 earned premium produce a 50% ratio. Dividing by $360,000 written premium instead produces 41.7%, but answers a different question when part of that premium relates to unearned coverage.
Common mistake: Comparing ratios built from different loss definitions or premium bases as though they measured the same period's underwriting performance directly.
30. Frequency, severity, and pure loss cost
Frequency equals claim count divided by exposure units. Average severity equals loss dollars divided by claim count. Multiplying them gives loss dollars per exposure unit, often called pure loss cost under the stated loss definition. The cancellation works only when counts, losses, and exposure cover the same population and period. If there are no claims, observed loss cost is zero but average severity cannot be estimated from that sample. These measures separate incident occurrence from claim size and support comparisons that raw totals cannot provide.
Worked example: A fleet records eight claims totaling $120,000 over 200 vehicle-years. Frequency is 0.04 claims per vehicle-year, severity is $15,000 per claim, and their product is $600 loss cost per vehicle-year. Each measure describes a different part of experience.
Common mistake: Combining claim counts from one coverage with loss dollars from another, then interpreting the resulting average severity as a meaningful account measure.
31. Exposure normalization during growth
Exposure normalization divides losses or claims by a relevant measure of opportunity for loss, such as vehicle-years, payroll units, or sales units. It helps distinguish growth from deterioration. Choose an exposure base connected to the loss mechanism and use consistent units across periods. Normalization does not eliminate changes in operational mix, reporting, inflation, or maturity; those require separate analysis. If the business doubles but claims rise only modestly, raw counts can look worse while frequency improves. Conversely, shrinking exposure can conceal worsening frequency behind lower total claims.
Worked example: Claims rise from ten to twelve while vehicle-years increase from 100 to 150. Frequency falls from 0.10 to 0.08 claims per vehicle-year, a 20% improvement. The underwriter still checks whether route types and claim reporting remained comparable.
Common mistake: Interpreting rising total claims as worsening risk quality without first considering whether the business had substantially more exposure during the later period.
32. Recurring losses and large-loss segmentation
Recurring smaller losses and isolated large losses can reveal different mechanisms. Repeated handling injuries may indicate a persistent process issue, while one major fire may highlight severe-event potential. Segment experience using a clearly stated analytical threshold, preserving both the recurring layer and excess amounts. Removing a large loss without retaining its implications understates risk; allowing it to obscure recurring patterns also weakens analysis. Any capped or excluded view should accompany the uncapped total and an explanation of why the segmentation helps answer a specific underwriting question.
Worked example: An account has twelve handling claims totaling $60,000 and one $500,000 fire. The underwriter separately evaluates the repeated handling process and the fire's protection implications, while retaining the full $560,000 total when describing overall reported experience.
Common mistake: Calling a large loss nonrecurring and deleting it from consideration without investigating whether its underlying severity mechanism remains present at the account.
33. Loss development and maturity adjustment
Loss development reflects changes as claims are reported, paid, reestimated, and closed. Recent periods often contain less complete information than older periods, especially for slower-settling claims. A supplied development factor converts reported losses into an estimated ultimate amount by multiplication, under that model's assumptions. The factor is not certainty or a universal multiplier. Compare periods at similar maturity or use justified adjustments before interpreting changes. Development estimates can be distorted when claim handling, coverage, exposure mix, or reporting patterns differ from the experience underlying the factors.
Worked example: A recent year reports $80,000 and uses an illustrative factor of 1.40, producing $112,000 estimated ultimate losses. An older year reports $100,000 with a 1.05 factor, producing $105,000. The lower recent reported amount does not imply improvement.
Common mistake: Treating a recent year's lower reported losses as final improvement, or presenting a development-adjusted estimate as a guaranteed ultimate claim outcome.
34. Loss trend and premium rate level
Loss trend and premium rate changes affect different sides of a loss ratio. Trend adjusts historical loss costs toward a target period for stated changes such as repair costs or claim severity. Premium on-level adjustment estimates what historical premium would have been at a specified rate level, holding the modeled exposure basis consistent. Neither adjustment substitutes for exposure growth or loss development. Use only supplied or justified assumptions and show each transformation separately. This prevents a ratio improvement caused by higher rates from being mistaken for better underlying risk quality.
Worked example: Historical losses of $100,000 are trended by an illustrative 10% to $110,000. Historical premium of $200,000 is adjusted to a 5% higher rate level, becoming $210,000. The adjusted ratio is 52.4%, with each assumption visible.
Common mistake: Attributing a lower loss ratio entirely to fewer or smaller losses when premium rate increases may explain much of the observed change.
35. Credibility and small-sample experience
Credibility describes how much weight an experience estimate deserves given its volume and stability. A small account's few observed claims can vary sharply by chance, so its raw average may be an unstable estimate of future cost. A stated credibility model can blend account experience with an appropriate broader estimate. The broader data must still be relevant to the account's operations and coverage. Do not invent a universal credibility cutoff or weight. Even a numerical blend requires judgment about comparability, changed exposures, and whether the observed period represents the future proposal.
Worked example: An exercise assigns 30% credibility to an account's $400 loss cost per unit and 70% to a comparable $700 benchmark. The blended estimate is $610 per unit. The supplied weights illustrate the method rather than establish an industry threshold.
Common mistake: Treating two claim-free years from a small account as conclusive evidence of negligible future losses, especially for infrequent severe events.
36. Outlier influence and sensitivity testing
Sensitivity testing shows how a result changes when an uncertain input or influential observation changes. In loss analysis, one large claim can dominate an average, so calculate the full result and a clearly labeled alternative to understand its influence. This is diagnostic analysis, not permission to discard inconvenient experience. Investigate whether the observation reflects an ongoing exposure, a corrected condition, or data error. A recommendation is fragile when modest plausible input changes reverse it; such fragility supports better evidence, explicit contingencies, or a more cautious structure.
Worked example: Nine claims total $45,000 and one claim adds $255,000. Average severity is $30,000 across ten claims, versus $5,000 for the nine smaller claims. The underwriter reports both views and investigates the large claim's continuing relevance.
Common mistake: Replacing the full experience result with a favorable outlier-excluded result without identifying the exclusion or evaluating the severe-loss mechanism it represents.
Account economics and underwriting alternatives
37. Permissible loss ratio and required premium
Under stated simplified assumptions, premium must fund expected covered losses, fixed expenses, variable expenses, and the target underwriting profit. If variable expenses and target profit are fractions of premium, required premium equals losses plus fixed expenses divided by one minus those fractions. The remaining fraction is the portion available for losses and fixed costs. This is different from adding the percentages as markups to costs. State excluded items such as taxes, capital costs, or investment income because the simplified result should not be mistaken for a complete pricing model.
Worked example: Expected losses are $140,000, fixed expenses are $20,000, variable expenses are 20% of premium, and target profit is 10%. Required premium is $160,000 divided by 0.70, approximately $228,571, under these hypothetical assumptions alone.
Common mistake: Adding a 30% markup to losses and fixed expenses when both the expense allowance and profit target are defined as percentages of premium.
38. Deductible layer allocation
A deductible allocates an initial layer of covered loss to the insured under the policy's stated mechanics. For a simple per-occurrence deductible with no other restrictions, insurer payment is the positive amount of covered loss minus the deductible. Apply any payment limit according to the specified wording. A deductible primarily changes retained frequency and loss sharing; it does not remove the physical possibility of a severe event. Evaluate whether the insured can fund retained losses and whether claims below the deductible remain visible enough for meaningful risk analysis.
Worked example: With a $15,000 per-occurrence deductible, covered losses of $8,000 and $40,000 produce insurer payments of zero and $25,000 before any other terms. The insured retains $8,000 on the first event and $15,000 on the second.
Common mistake: Subtracting one deductible from the annual loss total when the stated policy applies a separate deductible to each covered occurrence.
39. Limits, attachment, and tail exposure
A limit caps payment according to the coverage structure, while an attachment point identifies where a layer begins responding. For a simplified layer paying up to L above attachment A, payment is the lesser of L and the positive amount of covered loss minus A. Higher limits generally expose the insurer to more severe portions of the distribution, whose estimates may be uncertain. Distinguish per-occurrence limits from aggregates and preserve the supplied deductible interaction. A limit controls financial participation; it does not establish that the underlying operation fits appetite.
Worked example: A hypothetical layer pays $1 million above a $500,000 attachment. A covered $1.2 million loss produces $700,000 for that layer; a $2 million loss exhausts its $1 million limit. Other layers or uninsured portions require separate analysis.
Common mistake: Treating a layer's limit as the gross-loss level where payment begins, or assuming that a per-occurrence limit also supplies an annual aggregate.
40. Coverage restrictions and residual operating exposure
A coverage restriction changes which losses the insurer assumes. Evaluate its precise wording against plausible loss events, separating excluded losses from losses remaining within coverage. Removing one cause of loss does not eliminate other causes affecting the same operation or property. Estimate expected insurer payments under the restricted terms and assess whether the proposed premium supports the remaining losses and expenses. Also identify the insured's resulting coverage gap. A restriction can narrow the insurer's obligation while leaving the insured's underlying operating hazards unchanged, so appropriate controls may still matter.
Worked example: A courier's hypothetical cargo policy excludes damage caused by temperature variation but otherwise covers collision damage. Refrigeration failure spoils $30,000 of food: excluded. A separate collision breaks $20,000 of housewares: covered before applicable deductibles and limits. The restriction removes temperature-loss coverage while preserving collision exposure.
Common mistake: Assuming a temperature-variation exclusion removes all cargo exposure overlooks collision and other covered losses that still require pricing and underwriting review.
41. Risk-control investment economics
Risk-control economics compares the cost of a measure with its expected effects on loss frequency, severity, downtime, and retained costs. Use explicit assumptions about effectiveness and implementation rather than assuming a dollar-for-dollar premium reduction. Some benefits accrue to the insured through reduced interruption or uninsured losses, while insurer benefits depend on the covered layer. A measure can be necessary for acceptability even when a simple short-term payback calculation looks unfavorable. Separate the financial illustration from technical evidence that the control is suitable and will operate as intended.
Worked example: A hypothetical guard costs $12,000 and is estimated to reduce annual retained losses by $4,000. Simple payback is three years, excluding maintenance and discounting. The underwriter separately checks implementation and whether the measure affects insurer-paid losses above the deductible.
Common mistake: Promising a premium reduction equal to the insured's control expenditure without estimating covered-loss effects or applying the insurer's pricing authority.
42. Liquidity and deductible funding capacity
Liquidity concerns the ability to meet near-term obligations using accessible resources. Current assets divided by current liabilities gives a current ratio, but asset quality and timing matter: inventory and slow receivables may not fund an immediate deductible. Examine unrestricted cash, operating cash needs, credit availability, and possible simultaneous retained losses. No single ratio establishes universal acceptability. The underwriting relevance is whether financial pressure could undermine maintenance, continuity, or the proposed retention arrangement. A profitable income statement does not automatically demonstrate cash available when a loss occurs.
Worked example: A business has $300,000 current assets and $200,000 current liabilities, giving a 1.5 current ratio. Most assets are slow-moving inventory, and only $15,000 cash is available beyond operating needs. A $50,000 deductible therefore requires closer funding analysis.
Common mistake: Approving a large deductible solely because the current ratio exceeds an assumed cutoff, without examining accessible cash and the timing of obligations.
43. Leverage and interest-coverage pressure
Debt obligations can constrain a business's capacity to maintain equipment, fund controls, or absorb retained losses. One simplified interest-coverage measure divides earnings before interest and taxes by interest expense. It indicates earnings relative to interest burden, not cash availability or complete debt-service capacity. Principal payments, working-capital needs, and noncash earnings components require separate consideration. Use trends and stated financial definitions rather than a universal pass-fail ratio. Financial strain is a signal to investigate operational consequences, not proof that management will neglect safety or that a loss will occur.
Worked example: A manufacturer reports $240,000 earnings before interest and taxes and $120,000 interest expense, producing coverage of 2.0. The underwriter also asks about principal payments and deferred maintenance because the ratio alone does not show the cash available for controls.
Common mistake: Interpreting an interest-coverage ratio as a complete cash-flow forecast or applying an unsupported universal threshold to decide every account.
44. Expected contribution across account structures
Expected underwriting contribution compares premium with expected insurer-paid losses and the specified expenses for each structure. Reestimate losses when deductibles, limits, or coverage change; using the same loss estimate can misstate the effect of altered participation. Compare dollar contribution and, when useful, contribution as a percentage of premium. A higher percentage need not produce more dollars, and neither measure fully captures capital use or volatility unless included. Combine the economic comparison with eligibility, uncertainty, and the insured's ability to meet retained obligations before recommending a structure.
Worked example: Structure A produces $30,000 expected contribution on $200,000 premium, or 15%. Structure B produces $28,000 on $160,000, or 17.5%. B has the better percentage, but A contributes more dollars; neither comparison alone determines the preferred arrangement.
Common mistake: Choosing the highest contribution percentage without comparing contribution dollars, changed loss assumptions, severe-loss participation, and the practical feasibility of the proposed retention.
Portfolio context and ongoing account review
45. Geographic catastrophe accumulation
Geographic accumulation occurs when one event can damage multiple insured locations. Flood, wind, earthquake, and other hazards have different spatial footprints, so distance alone is an incomplete measure of independence. Map locations to the relevant peril and examine the insurer's combined event exposure, limits, and modeled assumptions. Individual protection and good loss history do not remove shared catastrophe potential. The underwriting question includes the incremental effect of the new account on an existing concentration, not simply whether that account would be acceptable if considered in isolation.
Worked example: Three well-maintained warehouses occupy the same floodplain. Adding a fourth can increase the insurer's loss from one flood even if its standalone expected loss is attractive. The underwriter checks the shared event footprint and existing commitments before approving additional capacity.
Common mistake: Assuming separate policyholders or excellent maintenance make locations independent when the same physical event can affect all of them together.
46. Industry and shared-dependency concentration
Portfolio concentration can arise from common operations or dependencies even when insured locations are geographically dispersed. Businesses may rely on the same supplier, technology, transport hub, or specialized process. A disruption can create correlated claims or financial stress across accounts, depending on coverage and circumstances. Identify shared drivers rather than grouping only by broad industry labels. Industry diversity does not guarantee dependency diversity. Conversely, several firms in one industry may have different loss drivers, so the analysis should explain the actual connection that could transmit an event across the portfolio.
Worked example: An insurer writes food manufacturers in several regions, all dependent on one refrigeration-service provider. A prolonged service disruption could affect multiple insureds. Reviewing only geographic spread would miss the shared operational dependency linking these otherwise separate businesses.
Common mistake: Calling a portfolio diversified solely because its insureds occupy different cities or industry classifications while overlooking a common critical supplier.
47. Marginal portfolio risk and correlation
An account's marginal portfolio effect depends on how its possible losses relate to losses already in the book. Expected loss totals add, but portfolio volatility also reflects dependence among accounts. A new risk sharing a catastrophe footprint can add more event concentration than an otherwise similar risk exposed to different drivers. This does not mean low correlation makes an ineligible account acceptable. Evaluate standalone quality and portfolio contribution together, using the insurer's stated capacity and modeling assumptions rather than treating premium volume as a sufficient measure of diversification.
Worked example: Two applicants have similar premium and expected losses. One lies inside the insurer's largest wind concentration; the other has different major loss drivers. Their standalone economics may match, but their incremental effects on the portfolio's event exposure can differ materially.
Common mistake: Assuming two accounts with equal premium and expected losses make equal contributions to portfolio volatility regardless of their shared event exposure.
48. Aggregate ratios and mix effects
An aggregate loss ratio is total losses divided by total premium, not the simple average of account ratios unless premiums are equal. Changes in business mix can improve the aggregate even when individual segments deteriorate. Analyze segment results and their premium weights before attributing a portfolio movement to better underwriting. Keep valuation and loss definitions comparable. This weighted-average logic prevents strong or large segments from concealing deterioration elsewhere and helps distinguish account-specific issues from broader patterns requiring changes in appetite, pricing, or control expectations.
Worked example: A $100,000-premium segment has a 90% loss ratio, while a $900,000 segment has a 40% ratio. Combined losses are $450,000 on $1 million premium, giving 45%. The simple average of 65% misrepresents the portfolio's actual premium weighting.
Common mistake: Averaging account loss ratios without premium weights, or treating aggregate improvement as proof that every underlying industry segment has improved.
49. Renewal as prospective re-underwriting
A renewal decision concerns the next period's exposures and terms. Review changes in operations, locations, management, losses, controls, coverage, and insurer appetite rather than automatically extending the original acceptance. Historical performance remains informative but may describe a different business. Identify which original assumptions still hold and which require replacement. Apply the relevant jurisdictional and contractual requirements when implementing any disposition, without assuming one universal renewal procedure. The technical decision is whether the proposed future arrangement remains supportable, including any necessary changes or authorized exception review.
Worked example: A profitable distributor begins installing products at customer sites. At renewal, the underwriter reassesses completed-work exposure, contracts, training, and loss-history relevance. An ordinary sales-based premium adjustment cannot by itself address the new operational pathway.
Common mistake: Treating renewal as an administrative continuation even when the insured's actual activities or the insurer's applicable appetite have materially changed.
50. Separating rate, exposure, and mix growth
Premium growth can result from rate changes, exposure growth, coverage changes, or shifts in business mix. These drivers have different implications for portfolio performance. More premium from more vehicles also brings more loss opportunities, while a rate increase on unchanged exposure can improve expected margin if other assumptions hold. Use a consistent cohort where possible and identify entrants, departures, and changed terms. A simple premium-growth percentage cannot establish improved pricing adequacy or diversification. Decompose the movement before interpreting what it says about underwriting results.
Worked example: A fleet's vehicle-years rise 20% and its rate per vehicle-year rises 10%, with unchanged coverage. Premium grows approximately 32% because 1.20 multiplied by 1.10 equals 1.32. Only part of that growth represents a higher rate on comparable exposure.
Common mistake: Describing all premium growth as improved pricing when increased exposure or a shift toward higher-hazard operations may explain much of the increase.
51. Monitoring triggers tied to assumptions
A monitoring trigger identifies an observable change that requires renewed analysis of a material underwriting assumption. Examples include a new process, a protection impairment, an acquisition, or losses exceeding a stated review condition. Define the information needed, review responsibility, and expected action under the applicable arrangement. A trigger prompts investigation rather than mechanically proving deterioration or dictating cancellation. Tie monitoring to specific causal concerns so that collected information can change a decision. Broad requests for periodic updates are less useful when they do not identify what would make the original analysis obsolete.
Worked example: Acceptance assumes no on-site installation. A monitoring condition requests notification if installation begins, along with contracts, revenue, and work descriptions. The trigger directly tests an assumption whose failure would change liability analysis and potentially the proposed terms.
Common mistake: Collecting routine status reports without specifying which changes matter or what analysis should follow when a material underwriting assumption fails.
52. Closing risk-control conditions with evidence
Closing a risk-control condition requires evidence that the stated action was completed to the relevant scope. Define the population, location, equipment, and completion criterion before reviewing proof. Reconcile records to the full population rather than relying on one sample certificate when complete coverage is required. Distinguish purchase, scheduling, installation, testing, and ongoing operation because they represent different stages. Record unresolved exceptions and the authorized response. This is a verification procedure for an agreed condition, separate from whether the control's original design was technically suitable for the exposure.
Worked example: A condition requires training for every active forklift operator. The underwriter matches the current roster to dated completion records and identifies two unmatched employees. A paid training invoice cannot close the condition until the coverage gaps are resolved appropriately.
Common mistake: Closing an implementation condition on evidence that work was purchased or scheduled when the requirement calls for completed and verified operation.
Decision documentation and underwriting communication
53. Evidence-to-decision underwriting rationale
A useful underwriting rationale connects evidence to material exposures, explains how those exposures affect the proposed arrangement, and states the resulting recommendation. Separate observations from interpretation and identify unresolved assumptions. Summarize credible alternatives and explain why the selected option fits appetite, authority, and economics. A list of documents or a statement that an account is acceptable does not show the reasoning. The note should be concise enough to use but complete enough for another reviewer to understand which facts drove the decision and what would cause it to change.
Worked example: A note identifies recurring water damage, verified pipe replacement, and remaining uncertainty about uninspected areas. It recommends terms reflecting that residual exposure and explains why unchanged acceptance was rejected. The reasoning links the evidence to the actual proposed decision.
Common mistake: Recording only 'good account' or copying the submission into the file without explaining the material exposure and resulting underwriting judgment.
54. Decision-ready exception referrals
An exception referral should identify the requested approval, the specific authority or guideline departure, and the underwriter's recommendation. Include decision-relevant facts, uncertainties, economic implications, and proposed safeguards. Separate verified improvements from promises and explain why the exception is supportable despite the normal boundary. Attachments provide detail but should not force the approver to reconstruct the argument. A referral is a request, so retain an accurate pending status until the authorized decision is received. Approval scope should be recorded precisely rather than assumed to cover unrelated departures.
Worked example: A $4 million limit request exceeds a $3 million delegation. The referral states the requested limit, exposure rationale, loss uncertainty, proposed deductible, and supporting inspection findings. It asks for that specific exception without implying that a separate pricing credit is also approved.
Common mistake: Sending a large document packet with no explicit requested decision, or treating approval of one exception as approval of every proposed deviation.
55. Actionable producer information requests
An actionable request states what is missing, why it matters, what evidence will satisfy the need, and when the information is needed for the transaction. Refer to the conflicting document or affected operation so the producer can obtain the right answer. Separate essential decision inputs from desirable background information. Use neutral wording that resolves facts without alleging misconduct. Where a deadline or requirement is hypothetical or account-specific, describe it as such. Clear requests reduce repeated exchanges because the recipient understands both the deliverable and the underwriting decision it enables.
Worked example: Instead of asking for 'updated financials,' the underwriter requests the latest balance sheet and unrestricted cash information to evaluate a proposed $50,000 deductible. The message identifies the relevant entity and reporting date so the producer can supply usable records.
Common mistake: Sending broad requests such as 'provide more detail' without specifying the missing fact, relevant period, acceptable evidence, or decision it affects.
56. Quote, subjectivity, and binding status
A quote communicates proposed terms, while binding creates the insurer's commitment through the applicable authorized process. A subjectivity is a specified requirement whose timing and effect must be clearly stated. Distinguish conditions required before binding from information permitted after inception under an authorized arrangement. Ambiguous status can create misunderstanding about whether coverage has been committed or a requirement waived. Apply the actual authority, wording, and jurisdictional setting rather than assuming all quotes or subjectivities operate identically. Record approvals and commitments accurately, including who made them and when.
Worked example: A quote requires verified sprinkler restoration before binding. The producer accepts the price but supplies only a repair schedule. The underwriter explains that the required evidence remains outstanding and does not describe acceptance of the price as completion of binding.
Common mistake: Assuming that agreement on premium automatically satisfies prebinding conditions or gives the underwriter authority to waive an outstanding evidence requirement.
57. Risk-based negotiation tradeoffs
Underwriting negotiation compares feasible combinations of premium, limits, deductibles, coverage, and controls. Explain which exposure or cost drives a proposed condition and how a requested change would affect participation. Concessions should remain within authority and use internally consistent loss assumptions. A lower premium paired with a higher deductible may be supportable, but only if expected insurer losses and insured retention capacity are evaluated. Avoid treating unlike concessions as equivalent merely because they help close the transaction. Any unresolved technical acceptability issue still requires a substantive solution.
Worked example: A producer requests a lower premium. The underwriter evaluates an authorized higher-deductible option using revised insurer-loss estimates and asks whether the insured can fund the retention. The discussion explains the tradeoff rather than offering an unsupported price reduction alone.
Common mistake: Trading away an essential physical protection requirement for additional premium without showing that the remaining risk is acceptable and the change is authorized.
58. Consistent criteria and reasoned exceptions
Consistency means applying the same relevant decision criteria to comparable facts, not producing identical outcomes for every applicant. Material differences in operations, controls, coverage, or authority can justify different terms. State those differences and document authorized exceptions so that discretion is explainable rather than arbitrary. Producer relationships and account size may provide business context without replacing the risk criteria. When a precedent is cited, compare its underlying facts and applicable rules instead of copying its result. This supports fair, reviewable decisions while preserving necessary distinctions between genuinely different risks.
Worked example: Two warehouses request the same terms. One has verified restoration of protection; the other has only scheduled repairs. Different dispositions are consistent when the decision rule requires operational protection and the file clearly records the factual difference driving each result.
Common mistake: Calling identical terms fair when applicants differ materially, or using an old precedent without checking whether its exposure facts and authority still apply.
59. Sensitive-information minimization
Underwriting files can contain financial records, personal information, claim details, and proprietary operational data. Collect and share information relevant to the decision, use approved access and transmission methods, and limit recipients to those with a legitimate role. Minimize unnecessary identifiers in summaries while preserving the facts needed for analysis. Retention, disclosure, and privacy duties depend on the applicable jurisdiction and organizational requirements, so avoid inventing a universal rule. Information protection should support effective analysis by controlling access and purpose rather than encouraging undocumented decisions or indiscriminate copying.
Worked example: A referral needs injury severity and claim status but not an employee's full identifying details. The underwriter summarizes the relevant claim facts, references the controlled source record, and sends the referral through the insurer's approved restricted-access channel.
Common mistake: Copying entire sensitive claim or financial files into broad email chains when a focused summary and controlled source access would meet the decision need.
60. Decision chronology and audit reconstruction
A reconstructable record preserves what information was available, what decision was made, who had authority, and when commitments or changes occurred. Date key evidence and distinguish initial proposals, pending referrals, approvals, and final terms. When new information changes the recommendation, retain the earlier basis and explain the revision instead of overwriting the history. This chronology allows reviewers to evaluate the decision using contemporaneous knowledge and avoids confusing later discoveries with facts known at acceptance. Record actual communications and approvals without implying that silence or an unreviewed attachment supplied authorization.
Worked example: An initial quote assumes no installation work. A later contract reveals installation obligations, prompting revised terms and a referral. The file preserves both stages, dates the new evidence, and records the approval so another reviewer can reconstruct why the recommendation changed.
Common mistake: Replacing the original note with a revised conclusion and leaving no record of when new facts emerged or who authorized the resulting change.