Introduction: Assessing a Claims Process as a Connected System
An operational risk assessment is most useful when it connects the work being done to the ways it can fail, the controls intended to prevent or detect failure, and the evidence that those controls operated. This tutorial applies those steps to a hypothetical insurer’s claims process. The case follows a claim from intake through coverage review and payment, then uses the same process to examine a control assessment, an incident, and key risk indicators (KRIs).
The ratings and thresholds below are an illustrative assessment convention, not universal industry standards. A sound assessment records the criteria being used, the evidence supporting a judgment, and the person responsible for follow-up. This integrated approach is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Define the Process and Its Exposures
Suppose a hypothetical insurer handles property claims through three stages. Staff log incoming claims, confirm coverage and approval authority, and then release payments. Exposure descriptions should identify a plausible failure in this work, not just repeat a broad label such as “processing risk.” A useful description makes it possible to identify both a consequence and a control objective.
| Process stage | Exposure: what could go wrong? | Possible consequence | Example control objective |
|---|---|---|---|
| Intake | A claim is entered twice, or required documents are not attached to the correct claim. | Duplicate handling, delay, or an incorrect payment. | Identify duplicates and confirm that required documents are linked to the claim. |
| Coverage and authority review | A claim is approved without adequate coverage support or by a person without the required authority. | Payment outside policy terms or approval limits. | Require documented coverage checks and approval within delegated authority. |
| Payment | A payment instruction is duplicated or sent for an incorrect amount. | Financial loss, recovery work, and customer or reporting consequences. | Match approved amounts to payment instructions and detect duplicate transactions. |
These descriptions distinguish the event (for example, a duplicate payment) from a cause (such as a repeated file import) and from its consequence (such as money paid twice). Keeping those elements distinct helps an assessor select controls that address the actual failure path.
Build an RCSA With Evidence and Explicit Rating Rules
For the worked case, use a five-point likelihood scale and a five-point impact scale. Likelihood ratings run from 1 (rare) to 5 (frequent); impact ratings run from 1 (minor) to 5 (severe). The illustrative inherent and residual score is the product of the two ratings. For this case, scores of 1–4 are low, 5–9 moderate, 10–14 high, and 15–25 critical.
These numerical scores are a structured prioritization aid, not precise measurements of loss or probability. The scales are ordinal: a likelihood rating of 4 is not necessarily twice a likelihood rating of 2. Therefore, assessors should record the rationale and evidence alongside each score and should not interpret the score as a dollar estimate.
The reason for the product is to combine two dimensions in a simple prioritization grid. If impact is held constant, increasing likelihood increases the score; if likelihood is held constant, increasing impact increases the score. For example, a likelihood rating of 4 and an impact rating of 3 give a score of 12, which is higher than a score of 6 when likelihood is 2 and impact remains 3. This supports ranking within the chosen rubric; it does not prove that the underlying risk has doubled.
Control assessment should consider both design and operating effectiveness. A control may be well designed on paper but fail to operate consistently. Conversely, evidence that a control ran does not establish that it was designed to detect the relevant failure. In this case, assessors review the written procedure, system configuration, and a defined sample of completed claims, documenting what was tested and what exceptions were found.
Worked Example: Rate a Duplicate-Payment Exposure
A hypothetical insurer imports claims from a service platform into its payment queue. The process owner identifies the exposure that an imported claim could be duplicated and paid twice. Before considering controls, the assessor selects likelihood 4 and impact 4. A daily reconciliation is designed to compare approved claims with payment instructions. For the assessment period, the assessor reviews 30 reconciliation records: 29 are complete and timely, while one was completed late. The control therefore has evidence of operation, but not perfect operation. The assessor judges that it lowers likelihood to 2; the impact of a duplicate payment remains 4.
Method. Rate the exposure first without controls, then reassess the likelihood and impact after considering the control evidence.
Inherent score. Likelihood \(=4\), impact \(=4\), so \(4 \times 4=16\). A score of 16 is critical under the stated bands.
Evidence review. There are 30 records, of which 29 were timely. The timely proportion is \(29/30=0.9667\), or 96.67% when expressed as a percentage. One late completion is an exception; the control should not be described as operating without exceptions.
Residual score. Using the assessor’s documented residual ratings, likelihood \(=2\) and impact \(=4\), so \(2 \times 4=8\). A score of 8 is moderate. The decrease reflects the judgment that the reconciliation reduces the chance of an undetected duplicate; it does not remove the possible payment impact.
Check. The inherent score is greater than the residual score, and both products use the same impact scale. A common mistake is to reduce impact merely because a reconciliation exists. Here the control is judged to reduce likelihood only, so residual impact stays at 4.
Describe the process failure in terms that can be connected to causes, consequences, and control objectives.
Apply the selected likelihood and impact criteria before giving credit for controls.
Consider design and operating evidence; record the scope, period, sample, and exceptions.
Use only supported control effects, explain remaining gaps, and identify an owner and due date for any action.
From Incident Record to Corrective Action
An RCSA describes assessed exposures and controls; an incident record captures a specific event or near miss. A useful record distinguishes confirmed facts from preliminary explanations. It identifies what happened, when and where it happened, the actual consequence, any potential consequence, immediate containment, cause analysis, and corrective-action ownership. A near miss is a failure or hazardous condition that did not produce the possible consequence because it was detected or interrupted in time; it should not be recorded as an actual loss of that avoided amount.
Worked Example: Record a Claims-Queue Incident
In a hypothetical month, a claims platform sends a batch to the payment queue twice after a file is reloaded. Three duplicate payments of $1,250 each are released. The monthly reconciliation identifies the duplicates. The insurer recovers one $1,250 payment before the incident is closed. The same review finds two further duplicate instructions, each for $1,250, that were stopped before payment.
Method. Separate the event facts, actual financial impact, avoided potential impact, likely cause, and actions. Do not add stopped instructions to actual paid loss.
Event and actual impact. Three duplicate payments were released, so gross duplicate payments were \(3 \times \$1{,}250=\$3{,}750\). One payment of $1,250 was recovered. The remaining unrecovered amount is \(\$3{,}750-\$1{,}250=\$2{,}500\). Thus, the actual unrecovered amount at this point is $2,500, while gross payments and recovery are separately recorded.
Near miss. Two additional instructions were stopped before payment. The prevented amount is \(2 \times \$1{,}250=\$2{,}500\). Record this as potential impact avoided, not as an additional actual loss.
Cause and response. The known trigger is that a file was reloaded. A plausible contributing control gap is that the reconciliation occurred monthly rather than before release. The owner should confirm the detailed cause, prevent the same file from being reprocessed without review, and assess whether the reconciliation should run before payment release. Record an accountable owner, a target date, and evidence required to close each action.
Check. The gross duplicate payments, recovery, unrecovered amount, and stopped instructions reconcile: $3,750 gross less $1,250 recovered equals $2,500 unrecovered; the separate $2,500 stopped amount was never paid. A common mistake is to report $5,000 as actual loss by adding the prevented amount to the unrecovered payments.
Use Indicators to Detect Deterioration and Escalate
A KRI is a defined measure that can signal increasing exposure or weakening controls. A usable indicator specifies its numerator, denominator, reporting period, data source, owner, and thresholds. Thresholds should be linked to an action; a color-coded status without a response rule does not complete the monitoring process. KRI definitions must also be stable enough that a change in the reported value reflects a change in the process, rather than a changed denominator or data method.
For example, let \(E\) be the number of confirmed claim-processing errors in a period and \(N\) the number of claims processed in that period. The observed error rate is \(E/N\). This ratio is useful because it relates the error count to the volume of work. If each processed claim is classified as either an error or not an error, the total error count is the sum of the claim-level error indicators. By linearity of expectation, if each claim has the same error probability \(p\), then \(E[E]=Np\), so \(E/N\) estimates the per-claim error probability \(p\). In practice, claims may not have identical risk, and a monthly rate is a monitoring measure rather than proof that the process has a fixed error probability.
Worked Example: Calculate KRI Breaches and Escalate
A hypothetical claims unit processes 4,800 claims in a month and confirms 72 manual payment adjustments. Its adjustment-rate thresholds are amber above 1.0% and red at or above 1.5%. The unit also has 240 claims awaiting documentation at month-end; 18 have been waiting more than five business days. For this aged-queue indicator, amber is above 5.0% and red is above 8.0%.
Method. Calculate each indicator using its specified numerator and denominator, then apply the threshold wording exactly.
Manual-adjustment rate. \(72/4{,}800=0.015\), or 1.50%. This meets the red threshold because red is at or above 1.5%.
Aged-queue rate. \(18/240=0.075\), or 7.50%. This exceeds 5.0% but does not exceed 8.0%, so it is amber, not red.
Escalation. Report the red adjustment-rate breach to the designated process and risk owners under the stated escalation procedure; investigate the adjustment drivers and verify the underlying records. Report the amber aged-queue result to its owner, identify the cause of delay, and monitor whether it worsens or persists. Record actions and follow-up dates.
Check. The denominators are different because the measures track different populations: all claims processed for adjustments, and month-end pending claims for aged items. A common mistake is to divide the 18 aged claims by 4,800 processed claims. That would mix populations and produce an indicator that does not match its definition.
Common Assessment Traps
- Using a broad risk label instead of an exposure. “Payment risk” does not tell an assessor what can fail. Describe the specific process failure, such as a duplicate payment instruction.
- Giving control credit based on design alone. A documented procedure is not evidence that it operated. State what records were tested and what exceptions were found.
- Changing ratings without a rationale. Explain whether a control changes likelihood, impact, or both. Do not assume a lower score is justified merely because a control exists.
- Treating a rating product as a probability or dollar value. The illustrative product prioritizes within a rubric; it is not a calibrated loss model.
- Combining actual impact and a near miss. Report paid, recovered, and prevented amounts distinctly.
- Escalating by color alone. Apply the exact threshold rule, identify the designated recipient, and specify investigation and follow-up actions.
The assessment is complete only when the evidence, ratings, incident information, indicator definitions, and actions tell a consistent story. A critical inherent rating may be reduced by controls, but residual risk remains until the evidence supports that judgment. An incident may reveal that an assumed control did not operate as expected, which should prompt reassessment and action rather than a retrospective change to the record.
Check Your Understanding
For each hypothetical case, apply the stated definitions and show the calculation or reasoning needed to reach the answer.
- A claim may be approved by a staff member above their delegated limit. State one precise exposure, a relevant control objective, and whether “staff training exists” alone establishes operating effectiveness. (Answer: Exposure: approval outside delegated authority; objective: prevent or detect approvals above an employee’s limit; training alone does not establish operating effectiveness.)
- Using the case’s score bands, an exposure has inherent likelihood 3 and impact 5. After a tested control, residual likelihood is 2 and impact remains 5. Calculate both scores and classifications. (Answer: Inherent \(3 \times 5=15\), critical; residual \(2 \times 5=10\), high.)
- A hypothetical unit pays four duplicate claims of $900 each and recovers $1,800. It also stops one additional $900 duplicate instruction before payment. State the gross duplicate payments, recovered amount, unrecovered amount, and avoided potential amount. (Answer: Gross $3,600; recovered $1,800; unrecovered $1,800; avoided potential $900, reported separately from actual payments.)
- A queue indicator is defined as claims pending over seven business days divided by all claims pending at month-end. There are 16 aged claims among 200 pending claims. Amber is above 6% and red is at or above 9%. Calculate the rate and status. (Answer: \(16/200=8.0\%\); amber, because it is above 6% but below 9%.)
- A monthly sample contains 50 reconciliation records, of which 47 were completed on time. The process owner proposes rating the control fully effective because most records were timely. State the timely percentage and one necessary assessment qualification. (Answer: \(47/50=94\%\); disclose the three late exceptions and assess their causes and significance rather than describing operation as exception-free.)