Calibrating Appetite With Scenario Anchors

Why Use Scenario Anchors?

A qualitative appetite statement can express a clear preference—such as protecting policyholder service during severe disruptions—without specifying the point at which a particular risk is no longer acceptable. A scenario anchor makes that preference more concrete. It describes a plausible adverse situation and its consequences, then records whether the organization considers that situation within appetite, at its boundary, or outside appetite.

Anchors help the board and management discuss what a stated preference means in practice. They can expose vague language, inconsistent judgments, or a threshold that conflicts with an organizational constraint. The purpose is calibration: to connect a risk preference to recognizable conditions and indicators. A scenario is not a prediction, and its use does not by itself estimate the probability or distribution of losses.

Definition: A scenario anchor is a defined, plausible adverse situation used as a reference point when judging risk appetite. It states the relevant circumstances, time horizon, organizational exposure, and consequences. An appetite threshold is a stated boundary on one or more indicators that distinguishes conditions accepted within appetite from conditions that are not.

A useful scenario says enough to support a consistent judgment. For example, “a technology incident” is too vague to anchor a threshold. A more useful description states which service is affected, the duration, the period considered, the number or type of customers affected, and any important consequences. The scenario should be severe enough to test appetite but plausible enough to inform decisions. “Worst imaginable” is not a useful substitute for “plausible adverse.”

Scenario anchors should be compared on a consistent basis. If one describes a one-day event and another describes cumulative annual effects, their judgments cannot be compared without explicitly accounting for that difference. Similarly, an indicator needs a clear definition: “service disruption” should specify, for example, whether it means elapsed time until restoration, the number of affected transactions, or both.

From Qualitative Preference to a Boundary

Calibration starts with the organization’s stated preference and relevant constraints, not with a convenient number. The people responsible for appetite should identify a scenario that is clearly tolerable and a more severe scenario that is clearly outside appetite. They can then state which consequences separate the two judgments and decide where the boundary belongs.

For a single ordered indicator, the logic can be stated simply. Suppose larger values of an indicator represent more severe outcomes. If a value of $x_A$ is accepted and a value of $x_O$ is outside appetite, with $x_A < x_O$, the boundary must lie above the accepted anchor and no higher than the outside-appetite anchor. The anchors constrain the boundary; they do not determine an exact number by themselves. A specific threshold requires an explicit organizational judgment or constraint.

x_A \leq T < x_O

Here $T$ is a chosen threshold on the indicator. This relation follows directly from the two judgments: setting $T < x_A$ would classify the accepted anchor as outside appetite, while setting $T \geq x_O$ would fail to keep the outside-appetite anchor beyond the boundary. If there is a hard constraint, such as a minimum level of resources that must remain, that constraint can determine $T$. Otherwise, the organization must make and document a judgment within the range allowed by the anchors.

When the scenario has several important dimensions, the boundary may be a set of conditions rather than a single number. One dimension may have a numeric limit, while another may have a non-negotiable condition. A scenario that breaches any one of those conditions may be outside appetite. Do not combine distinct dimensions into a single score unless the organization has a clear, justified method for doing so; an average can conceal an unacceptable consequence.

State the preference.
Identify the objective or protection the appetite statement is meant to express, such as maintaining timely service to policyholders.

Describe comparable scenarios.
Specify the event, scope, time horizon, exposure, and consequences. Use the same indicator definitions across scenarios.

Record the appetite judgment.
Label each anchor as within appetite, at the boundary, or outside appetite, and state why the consequences warrant that judgment.

Set the threshold or conditions.
Use an explicit organizational constraint where one applies. Otherwise, choose and document a boundary consistent with the anchor judgments.

Test the result.
Check whether the thresholds classify the anchor scenarios as intended and whether any separate dimension or constraint changes the judgment.

Worked Scenario Calibrations

Worked Example: Setting a Service-Continuity Boundary

A hypothetical insurer wants its claims platform to remain usable during technology disruptions. Its stated appetite permits a disruption lasting no more than 8 hours, no more than 3,000 delayed claims, and no missed legally due payments. In a common one-day claims peak scenario, a 4-hour interruption delays 1,200 claims and misses no due payments; management judges this within appetite. A 12-hour interruption delays 4,100 claims and causes 18 legally due payments to be missed; management judges this outside appetite. A proposed scenario lasts 7 hours, delays 3,200 claims, and misses no legally due payments. Determine its classification and explain the role of the anchors.

Method. Treat the stated appetite conditions as separate constraints. A scenario is within appetite only if it meets all three: interruption duration is at most 8 hours, delayed claims are at most 3,000, and missed legally due payments equal zero.

Step 1: Check the anchors against the stated conditions. The 4-hour anchor satisfies the duration condition because $4 \leq 8$, the claims condition because $1{,}200 \leq 3{,}000$, and the payment condition because it misses none. Its within-appetite label is consistent with the proposed thresholds. The 12-hour anchor fails the duration condition because $12 > 8$, fails the claims condition because $4{,}100 > 3{,}000$, and fails the payment condition because 18 payments are missed. Its outside-appetite label is also consistent.

Step 2: Classify the proposed scenario. Its 7-hour duration meets the limit because $7 \leq 8$. Its 3,200 delayed claims exceed the limit by $3{,}200 - 3{,}000 = 200$. It meets the no-missed-payment condition, but it fails the claims condition.

Conclusion and check. The proposed scenario is outside appetite because it breaches one of the three stated conditions. The anchor comparison confirms that the threshold set is consistent with the two reference judgments; it does not establish how likely either scenario is. A common mistake is to classify the proposed case as acceptable because its 7-hour duration is below 8 hours. That ignores the separate claims threshold. The correction is to test each condition individually and require all of them to be met.

Worked Example: Using a Resource Constraint as an Anchor

A hypothetical insurer has $240 million of resources available for a defined adverse event scenario. Its governance constraint is that at least $175 million must remain afterward. A scenario with a $50 million impact is judged within appetite; one with a $70 million impact is outside appetite. A proposed scenario has a $63 million impact. Determine the maximum impact consistent with the resource constraint, classify the proposed scenario, and explain what the scenario judgments do and do not establish.

Method. Use the stated minimum-resource constraint to calculate the boundary on the scenario impact. The impact is already specified on the same basis as the available resources; no loss distribution or frequency estimate is needed.

Step 1: Calculate the boundary. If $L$ is the scenario impact in millions of dollars, resources remaining are $240 - L$. The constraint requires $240 - L \geq 175$. Rearranging gives $L \leq 240 - 175 = 65$. Thus the maximum impact consistent with the constraint is $65 million.

Step 2: Check the anchors. For the $50 million anchor, resources remaining are $240 - 50 = 190$ million, which exceeds the $175 million minimum. For the $70 million anchor, resources remaining are $240 - 70 = 170$ million, which is $5 million below the minimum. The given within- and outside-appetite judgments agree with the constraint.

Step 3: Classify the proposed scenario. Resources remaining are $240 - 63 = 177$ million. Since $177 \geq 175$, the scenario meets the stated constraint and is within appetite on this dimension. It has a $2 million margin above the minimum.

Conclusion and check. The boundary of $65 million lies between the $50 million accepted anchor and the $70 million outside-appetite anchor, as expected. The common mistake is to treat the $70 million outside-appetite anchor as the threshold. That would permit an impact that leaves only $170 million, below the required minimum. The constraint, not the midpoint between the anchors, sets the boundary. The anchors confirm its interpretation; they do not provide a probability of the event.

Worked Example: Keeping Multiple Consequences Visible

A hypothetical insurer depends on an external service provider for policy records. Its appetite statement says that an adverse provider incident must not cause a service interruption longer than 36 hours, affect more than 1,500 policies, or cause any statutory filing to be late. A 24-hour interruption affecting 900 policies with no late filings is judged within appetite. A 72-hour interruption affecting 2,200 policies and causing three late filings is judged outside appetite. A proposed scenario lasts 30 hours, affects 1,400 policies, and causes one late filing. Classify it and explain why the indicators should not be averaged.

Method. Translate the appetite statement into three separate tests. The maximum duration is 36 hours; the maximum affected-policy count is 1,500; the permitted number of late filings is zero.

Step 1: Check both anchors. The first anchor meets all conditions: $24 \leq 36$, $900 \leq 1{,}500$, and zero is no more than zero late filings. The second anchor breaches all three: $72 > 36$, $2{,}200 > 1{,}500$, and three is greater than zero.

Step 2: Test the proposed scenario dimension by dimension. Duration meets the limit because $30 \leq 36$. The affected-policy count meets its limit because $1{,}400 \leq 1{,}500$. The filing condition fails because one late filing is greater than the permitted zero.

Conclusion and check. The proposed scenario is outside appetite because it breaches the filing condition, even though it meets the other two limits. The common mistake is to make a combined score in which the acceptable duration and policy count offset the late filing. That correction is not justified: the appetite statement makes timely statutory filing an independent condition. The example also shows why a scenario description should include each consequence that matters to the stated preference.

Common Calibration Traps

Scenario anchors clarify judgment, but they can create false confidence if they are used carelessly. Keep the distinction between a reference scenario, an appetite threshold, and an estimate of event likelihood clear. A scenario can be useful even when no probability has been assigned to it.

Choosing the midpoint by default. Two anchors constrain a boundary, but their arithmetic midpoint is not automatically meaningful. Use a midpoint only if the organization has a reason for that choice.
Calling an extreme imaginable event a plausible anchor. A boundary based on an event with no credible description or scope may not guide appetite decisions. State why the scenario is plausible and what assumptions define it.
Changing the measurement basis. Comparing a gross impact in one scenario with an after-adjustment impact in another can make the anchors inconsistent. Define the indicator and basis before comparing outcomes.
Ignoring a separate condition. A scenario may meet a financial threshold while breaching a service or legal constraint. Test every stated appetite condition; do not let success on one dimension cancel a failure on another.
Confusing threshold calibration with forecasting. A scenario anchor says how the organization judges a defined situation. It does not, by itself, say how often that situation will occur.

A useful calibration record includes the scenario narrative, the indicator definitions, the within- or outside-appetite judgment, the rationale for that judgment, the resulting threshold, and any important constraint. This makes the boundary understandable and allows the organization to check that its appetite statement and scenario judgments remain consistent. The approach is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Key takeaway: Use plausible scenarios to make a qualitative risk preference concrete. Let the scenarios constrain and test appetite thresholds, but set each boundary using an explicit organizational judgment or constraint. Preserve separate conditions when one unacceptable consequence cannot be offset by another acceptable one.

Check Your Understanding

For each problem, apply the stated appetite conditions to the scenario anchors and proposed case. Treat all organizations and figures as hypothetical.

A hypothetical insurer accepts a 6-hour processing interruption affecting 800 claims and rejects a 14-hour interruption affecting 2,400 claims. Its appetite conditions are a maximum 10-hour interruption and no more than 1,500 affected claims. A new scenario lasts 9 hours and affects 1,600 claims. Is it within appetite? (Answer: Outside appetite; it exceeds the 1,500-claim limit.)
A hypothetical company has $180 million of resources and must retain at least $140 million after a defined scenario. A $25 million impact is accepted and a $50 million impact is outside appetite. What is the maximum impact allowed by the resource constraint, and how much remains after a $37 million impact? (Answer: Maximum impact $40 million; $143 million remains after the $37 million impact, so it meets the constraint.)
A hypothetical insurer’s appetite permits a vendor outage of no more than 24 hours, at most 700 affected policies, and zero missed regulatory deadlines. A scenario lasts 20 hours, affects 650 policies, and misses one deadline. Classify it and identify the controlling condition. (Answer: Outside appetite; the zero-missed-deadline condition is breached.)
Two scenario anchors have impact values of $30 million (within appetite) and $54 million (outside appetite). No other constraint or boundary judgment is supplied. Is $42 million automatically the appetite threshold? (Answer: No; $42 million is the arithmetic midpoint, but the anchors alone do not justify choosing it.)
A hypothetical organization accepts a scenario with a 12-hour outage and 400 affected customers, and rejects one with a 36-hour outage and 1,000 affected customers. Its proposed appetite limits are 24 hours and 800 customers. A new scenario has a 20-hour outage and 850 affected customers. Classify it and explain why averaging the two measurements is not appropriate. (Answer: Outside appetite because 850 exceeds 800; the separate limits must be tested individually, so a favorable outage duration cannot offset the customer-count breach.)

What You'll Learn