Claims Governance, Ethics and Strategic Transformation

Introduction

Claims management is the point at which an insurer’s promises become observable to customers, regulators and society. A technically correct policy underwriting decision can still produce reputational, legal and financial damage if claims are handled unfairly, slowly or inconsistently. Conversely, disciplined claims governance can protect solvency, strengthen trust, identify emerging risks and provide strategic intelligence to the enterprise.

Claims governance is the system through which the board, senior management and operational leaders direct, control and monitor claims decisions. It includes authority structures, policies, controls, ethical standards, performance measurement, data governance, third-party oversight and accountability. Fraud diagnostics is not an isolated investigation function; it is part of enterprise risk management (ERM), which integrates risk identification, assessment, response and monitoring across the organisation.

The strategic challenge is to balance four objectives: fair treatment of legitimate claimants, effective control of fraud and leakage, compliance with law and regulation, and economically sustainable operations. Excessive suspicion can harm honest customers, while weak controls can increase losses and encourage organised fraud. This chapter examines how claims leaders can govern this balance and transform capabilities in response to climate change, social inflation, autonomous systems and evolving fraud threats.

Learning objectives

After studying this chapter, learners should be able to:

1. Explain claims governance and align claims and anti-fraud strategies with ERM, compliance and organisational objectives.
2. Design controls for outsourcing, conflicts of interest, whistleblowing, data use and conduct risk.
3. Evaluate ethical dilemmas involving fraud detection, vulnerable customers, privacy and automated decision-making.
4. Lead strategic change, workforce capability development and scenario planning.
5. Assess the implications of climate risk, social inflation, autonomous systems and emerging fraud techniques.
6. Recommend a governance architecture and implementation roadmap for a claims function.

1. Claims governance within enterprise risk management

1.1 Governance, management and assurance

Governance answers the questions: Who has authority? What outcomes are expected? How is performance challenged? Who is accountable when controls fail? Management converts those expectations into processes, people and technology. Assurance independently evaluates whether governance and controls operate effectively.

A practical model uses three lines of responsibility:

The board remains responsible for oversight, even when claims activity is delegated. A claims executive should therefore maintain a documented accountability map covering reserving, settlement authority, complaints, fraud referrals, vendors, data access and regulatory reporting.

1.2 Alignment with ERM

ERM views claims risk as a portfolio of interconnected exposures rather than merely an expense. Relevant risks include:

• Underpayment or unfair denial risk: legitimate claims are not paid correctly or promptly.
• Claims leakage: avoidable payments arise from poor controls, duplicate payments, weak negotiation or process errors.
• Fraud risk: false, exaggerated, staged or collusive claims create financial and social costs.
• Operational risk: system outage, inadequate staffing or poor documentation disrupts service.
• Conduct risk: incentives or practices produce harmful customer outcomes.
• Legal and regulatory risk: actions breach policy wording, consumer law, privacy rules or reporting obligations.
• Reputational risk: public confidence declines following perceived unfairness or misconduct.
• Model and technology risk: automated triage or fraud scores produce discriminatory or unexplained outcomes.

Risk appetite translates strategy into boundaries. For example, an insurer may accept moderate process cost to maintain a high standard of customer accessibility, but have zero tolerance for retaliation against whistleblowers, deliberate misrepresentation or unauthorised disclosure of personal data. Risk appetite statements should be measurable and linked to escalation thresholds.

A useful loss decomposition is:

The terms are management categories rather than universally separable accounting quantities; organisations must define them consistently to avoid double counting. Indemnity payments compensate insured losses, handling expense covers internal and external administration, fraud loss concerns payments caused by dishonest activity, leakage concerns preventable cost, and reinsurance impact reflects recoveries and associated costs. The identity is a management framework, not a statutory accounting formula.

1.3 Governance metrics and balanced measurement

A mature dashboard combines customer, financial, operational, risk and people indicators. Examples include settlement cycle time, complaint upheld rate, quality assurance score, fraud savings, false-positive rate, reopened claims, reserve adequacy, vendor incidents, vulnerable-customer outcomes and employee speak-up activity.

A fraud referral rate alone is unsafe: a team can increase referrals without improving fraud detection. More informative measures include precision and recall. If a model identifies 100 claims as suspicious and 40 are subsequently confirmed as fraud, its observed precision is:

If 80 fraudulent claims existed in the reviewed population and 40 were detected, recall is:

These measures assume reliable confirmation criteria and a known or sampled population. They do not by themselves establish fairness, financial value or causation. A balanced dashboard should also measure false positives, customer harm, investigation cost and time to resolution.

2. Ethical foundations and conduct risk

2.1 Ethical principles in claims

Rules are necessary but insufficient. Ethical claims decisions apply principles such as:

• Fairness: similar cases should receive consistent treatment, while relevant differences are recognised.
• Proportionality: investigation intensity should correspond to risk and potential harm.
• Transparency: customers should understand material reasons for delay, information requests or rejection, subject to legitimate confidentiality constraints.
• Respect and dignity: claimants may be distressed, injured, bereaved or financially vulnerable.
• Confidentiality: personal, health, financial and investigative information must be protected.
• Accountability: a person or committee must be able to explain and review important decisions.
• Integrity: employees, experts and vendors must not manipulate evidence, targets or outcomes.

Ethical tension is common. For example, extensive surveillance might improve fraud detection but intrude on privacy. A strict document requirement may reduce fraud but disproportionately burden a claimant with disability or limited digital access. The appropriate response is not to abandon controls, but to apply necessity, proportionality, alternative channels and human review.

2.2 Conduct risk

Conduct risk is the possibility that an organisation’s behaviour causes poor customer, market or societal outcomes. In claims, it can arise from unreasonable delay, opaque denials, aggressive investigation, inaccessible communication, conflicts in expert selection, target-driven settlement or inappropriate use of vulnerable-customer information.

Conduct controls should cover the full claim lifecycle:

1. Notification: provide accessible channels and explain immediate requirements.
2. Assessment: gather relevant evidence without fishing expeditions.
3. Decision: apply policy wording consistently and document reasoning.
4. Settlement: pay promptly and explain deductions or compromises.
5. Dispute: offer meaningful internal review and external escalation.
6. Learning: analyse complaints, appeals and near misses for systemic causes.

A useful control question is: “Would the organisation be prepared to explain this action to the claimant, regulator and an independent reviewer?” This does not replace legal analysis, but it exposes practices that may be technically defensible yet ethically weak.

2.3 Conflicts of interest

A conflict exists when a personal, financial, professional or organisational interest could improperly influence judgement, or reasonably appear to do so. Examples include an investigator handling a claim involving a relative, a repair network owned by an executive, an expert paid according to claim denial, or a vendor rewarded solely for reducing settlement values.

Controls include declarations, pre-assignment screening, segregation of duties, independent review, rotation where appropriate, remuneration design and a documented recusal process. Conflicts are not always misconduct; undisclosed or unmanaged conflicts create the governance failure. A conflict register should identify the interest, affected decisions, mitigation, reviewer and closure date.

3. Outsourcing and third-party governance

Insurers outsource adjusting, medical assessment, repair, investigation, legal work, call-centre operations, technology and data analytics. Outsourcing transfers activities, not ultimate accountability. The insurer remains responsible for selecting competent providers, defining service standards and monitoring customer outcomes.

3.1 Due diligence and contracting

Before appointment, due diligence should assess ownership, financial resilience, licences, competence, information security, privacy controls, subcontractors, business continuity, complaints history, conflicts and cultural compatibility. The contract should specify:

• scope, authority limits and prohibited activities;
• service levels and customer accessibility;
• evidence and record-retention requirements;
• data ownership, permitted use, security and breach notification;
• audit and regulatory access rights;
• model validation and explainability where analytics are supplied;
• subcontracting approval;
• incident escalation and remediation;
• exit assistance and secure data return or destruction.

Service-level agreements should not reward undesirable behaviour. A target based only on average settlement cost may encourage underpayment. Balanced incentives should include accuracy, timeliness, complaints, quality assurance, fair treatment and confirmed fraud outcomes.

3.2 Ongoing oversight

Vendor governance is a lifecycle, not a procurement event. Risk-tiered monitoring may include file reviews, outcome testing, customer interviews, access-log analysis, thematic audits and review of subcontractor changes. Critical providers require contingency plans and tested exit arrangements. The board or risk committee should receive material outsourcing incidents, not merely operational statistics.

4. Whistleblowing and speak-up systems

Whistleblowing is the reporting of suspected wrongdoing through protected internal or external channels. It differs from an ordinary service complaint: whistleblowing generally concerns misconduct, fraud, regulatory breach, safety, retaliation or concealment.

An effective programme provides confidential or anonymous reporting options, multiple channels, independent case intake, triage, investigation protocols, evidence preservation, non-retaliation protection and feedback to the reporter where lawful. Reports should be classified by urgency and risk, with conflicts screened before an investigator is appointed.

The organisation must distinguish confidentiality from absolute anonymity. A reporter’s identity may need to be disclosed under law or fair investigation procedures, but access should be restricted and retaliation actively monitored. Retaliation includes dismissal, demotion, exclusion, threats, adverse scheduling or subtle career disadvantage.

Metrics should not treat a high or low number of reports as automatically good or bad. A high number may indicate trust, widespread misconduct or both; a low number may indicate a healthy culture or fear. More useful indicators include time to triage, substantiation rate interpreted cautiously, retaliation allegations, remediation completion and employee survey evidence of psychological safety.

5. Data governance, analytics and responsible AI

Claims and fraud functions use sensitive data: identity, health, location, financial transactions, vehicle telemetry, images, communications and external databases. Data governance establishes rules for collection, quality, access, use, retention, sharing and deletion.

5.1 Data principles

A responsible data framework applies:

• Purpose limitation: collect and use data for specified legitimate purposes.
• Data minimisation: obtain what is relevant and necessary, not everything available.
• Accuracy: correct unreliable or outdated information.
• Security: protect data through access controls, encryption, monitoring and resilience.
• Retention discipline: retain records for legal, evidential and business needs, then dispose of them securely.
• Accountability: document decisions, owners, impact assessments and approvals.
• Rights and redress: provide appropriate notice, access, correction and challenge mechanisms under applicable law.

Data sharing with fraud bureaux, public bodies, reinsurers or vendors requires a lawful basis, defined purpose, secure transfer and controls against secondary use. Cross-border processing may create additional legal and operational requirements.

5.2 Automated decision-making

Artificial intelligence can support triage, image assessment, anomaly detection, document extraction and network analysis. It should normally assist rather than silently replace accountable decision-makers in high-impact claims decisions. Governance should test data provenance, representativeness, performance drift, explainability, disparate outcomes, cybersecurity, human override and appeal rights.

Suppose an automated model flags claims for investigation. A threshold is a policy choice, not merely a technical output. Lowering the threshold may increase recall but also increase false positives and investigative burden. A human reviewer should consider contextual evidence and record why a decision was accepted, modified or rejected.

Model governance should include pre-deployment validation, independent challenge, monitoring, version control, incident reporting and retirement criteria. Protected characteristics should not be used improperly, and proxy variables require scrutiny because apparently neutral variables may reproduce discrimination.

6. Strategic transformation of the claims function

Transformation is a sustained change in operating model, capability, technology and culture that improves outcomes. It is not synonymous with digitisation. Replacing paper forms with an online form does not transform a fragmented decision process.

6.1 A transformation framework

A practical sequence is:

1. Diagnose: map customer journeys, controls, costs, data flows and failure points.
2. Define outcomes: specify strategic objectives such as faster fair settlement, resilience or reduced organised fraud.
3. Design the target operating model: clarify roles, authority, platforms, vendors, skills and assurance.
4. Prioritise initiatives: assess value, risk, feasibility, dependencies and customer impact.
5. Pilot safely: test in a limited population with control groups or quality sampling where feasible.
6. Scale with governance: train users, monitor outcomes and preserve human escalation.
7. Institutionalise learning: update policies, incentives, models and capability plans.

A simple prioritisation score can be used when assumptions are made explicit:

Benefit and effort must be measured on defined scales, and confidence should reflect evidence quality. The score is a decision aid, not an objective truth; legal obligations, safety and fairness can override a high commercial score.

6.2 Capability development

Future claims professionals require more than policy knowledge. Capability domains include coverage interpretation, investigation, negotiation, customer vulnerability, ethics, data literacy, cyber awareness, vendor management, climate science, communication and change leadership.

A capability strategy should define proficiency levels, assess current gaps, provide role-specific learning and verify competence through observed work. Training is stronger when linked to case simulations, calibration sessions and feedback than when measured only by course completion. Leaders should protect time for learning and recognise ethical escalation rather than rewarding silent throughput.

7. Scenario planning for emerging risks

Scenario planning explores plausible futures to test whether strategy remains robust. It is not a prediction exercise. A useful process identifies driving forces, selects critical uncertainties, constructs internally consistent scenarios, tests current capabilities, defines trigger indicators and prepares options.

7.1 Climate risk

Climate risk affects claims through physical hazards, secondary perils, changing frequency and severity, infrastructure vulnerability, supply-chain disruption and insurance affordability. Historical loss data may become less representative as hazard patterns change. Claims operations may face simultaneous losses, inaccessible locations, contractor shortages, power outages and heightened vulnerability.

Strategic responses include catastrophe staffing plans, remote assessment with safeguards, pre-agreed supplier capacity, event-based triage, resilient communications, vulnerable-customer protocols, exposure mapping and post-event learning. Climate scenario analysis should distinguish acute hazards, such as flood or wildfire, from chronic changes, such as sea-level rise or heat. Insurers should avoid unfairly shifting adaptation failures onto claimants where coverage and law do not support it.

7.2 Social inflation

Social inflation describes claims-cost growth driven by litigation trends, expanding liability interpretations, larger awards, changing social expectations, third-party litigation funding, legal advertising and distrust of institutions. It is not simply general price inflation. It can increase settlement values, defence costs, cycle times and uncertainty in liability claims.

Responses include early case assessment, stronger legal trend monitoring, disciplined reserving, transparent authority levels, mediation strategy, expert quality controls and analysis of jurisdictional differences. Aggressive tactics may reduce a single claim cost while increasing reputational and conduct risk; strategic claims management therefore evaluates total cost and fairness.

7.3 Autonomous systems

Autonomous vehicles, drones, industrial robots and AI agents alter causation, evidence and responsibility. A claim may depend on software version, sensor performance, human supervision, manufacturer design, maintenance or cyber intrusion. Traditional assumptions that a human driver or operator is the primary decision-maker may no longer hold.

Claims readiness requires preserving machine logs, establishing chain of custody, developing technical expertise, clarifying policy response to software and cyber events, and coordinating with manufacturers and regulators. Autonomous systems may reduce some accidents but create concentrated systemic failures when a common software defect affects many insured assets.

7.4 Evolving fraud threats

Fraudsters increasingly exploit synthetic identities, generative media, deepfakes, account takeover, staged accidents, organised repair networks, cyber-enabled invoice manipulation and collusion across jurisdictions. Generative AI can make false documents and voice communications more convincing, while also helping insurers compare images, detect anomalies and authenticate evidence.

A resilient strategy combines prevention, authentication, network analysis, intelligence sharing, specialist investigation and customer education. Controls should not assume that digital evidence is automatically genuine. Provenance, metadata, independent corroboration and secure chain of custody are increasingly important. Fraud responses must remain proportionate: suspicion is not proof, and an algorithmic flag is not a finding of dishonesty.

8. Illustrative case: Meridian Mutual’s claims transformation

Background

Meridian Mutual, a fictional regional insurer, experienced severe flooding, rising repair costs and a 22% increase in suspected motor fraud referrals over two years. It outsourced field adjusting and vehicle repair management. Its dashboard emphasised average settlement cost and claim closure speed.

Problem

An internal review found inconsistent referral thresholds, limited vendor oversight, inadequate explanation of automated flags and a conflict involving a repair-network manager. Complaints from older customers revealed that digital-only evidence submission caused delays. At the same time, climate events created surge volumes and social media criticism of alleged unfair denials.

Analysis

The board’s risk committee mapped the issue as a combined conduct, operational, fraud, outsourcing, technology and reputation risk. The high referral rate did not demonstrate success because false-positive rates and customer impact were not measured. Vendor incentives encouraged low settlement values, while surge staffing plans assumed stable weather patterns. The organisation also lacked a protected channel for reporting vendor misconduct.

Approach

Meridian adopted the following programme:

• established a claims governance committee with risk, compliance, customer, actuarial and technology representation;
• replaced cost-only vendor incentives with balanced measures for accuracy, timeliness, complaints and confirmed fraud;
• created a conflict register and reassigned affected files;
• introduced telephone, postal and assisted-digital channels for customers unable to upload evidence;
• required human review before adverse action based materially on an automated score;
• validated the fraud model by product, geography and relevant customer segments;
• implemented confidential whistleblowing channels and anti-retaliation monitoring;
• developed climate surge scenarios with pre-contracted capacity and mobile assessment teams;
• trained staff in vulnerability, data ethics, deepfake indicators and autonomous-vehicle evidence;
• reported false positives, appeal outcomes and vendor incidents to the risk committee.

Outcome

After twelve months, Meridian reported shorter average cycle time for straightforward claims, fewer unsupported fraud referrals and improved complaint resolution. Confirmed fraud savings initially fell because the organisation stopped counting unsubstantiated suspicions as savings, but measurement credibility improved. The case demonstrates that ethical governance may reduce apparent short-term performance while improving sustainable value.

Learning points

1. Metrics shape behaviour; poorly designed targets create conduct risk.
2. Outsourcing requires active oversight and does not transfer accountability.
3. Automation must be governed as a high-impact decision support system.
4. Resilience planning must include climate-driven volume surges and vulnerable customers.
5. Fraud control is stronger when legitimacy, fairness and evidence quality are measured together.

9. Implementation roadmap and limitations

A realistic roadmap begins with regulatory obligations and material risks, then sequences foundational controls before advanced analytics. First establish ownership, policies, data inventories, access controls, complaint learning and vendor registers. Next improve workflow, quality assurance, investigation standards and workforce capability. Only then should the organisation scale sophisticated AI or network analytics.

Transformation has limitations. Better data can remain biased; scenario planning cannot eliminate uncertainty; independent review consumes resources; and excessive controls can slow legitimate settlement. Cost-benefit analysis should include customer harm, legal exposure, resilience and option value, not just immediate savings. Governance should also permit controlled experimentation, because refusing all innovation can leave an insurer less safe than a carefully tested alternative.

A claims leader should ask five final questions:

• Does the operating model produce fair outcomes, not merely efficient processing?
• Can every material decision be traced to accountable people, evidence and authority?
• Are third parties and models governed throughout their lifecycle?
• Can the organisation absorb a simultaneous catastrophe, technology failure or fraud campaign?
• Does culture reward truth, challenge and learning?

Chapter Summary

Claims governance connects board oversight, management controls and independent assurance to the insurer’s promises and risks. Effective governance aligns claims and anti-fraud strategies with ERM, compliance and organisational objectives through clear accountability, risk appetite and balanced metrics. Ethical practice requires fairness, proportionality, transparency, confidentiality and respect, especially for vulnerable customers.

Outsourcing does not transfer ultimate accountability. Contracts, due diligence, monitoring, audit rights and exit planning are essential. Conflicts of interest require disclosure, mitigation and independent review. Whistleblowing systems must protect confidentiality, prevent retaliation and support credible investigation. Data governance and responsible AI require lawful purpose, minimisation, accuracy, security, explainability, human oversight and redress.

Strategic transformation combines process redesign, technology, capability development and culture. Scenario planning strengthens readiness for climate risk, social inflation, autonomous systems and evolving fraud. Sustainable performance is achieved when fraud savings, customer outcomes, resilience, conduct and operational efficiency are assessed together rather than through narrow cost targets.

Key Terms

Claims governance: The structures and processes used to direct, control and monitor claims activity.

Enterprise risk management: An integrated approach to identifying, assessing, responding to and monitoring organisational risks.

Conduct risk: The risk that organisational behaviour causes poor customer, market or societal outcomes.

Claims leakage: Preventable claims cost caused by errors, weak controls or inefficient processes.

Fraud diagnostics: The use of evidence, investigation, analytics and controls to identify and manage dishonest claims activity.

Three lines model: A framework distinguishing operational ownership, risk challenge and independent assurance.

Risk appetite: The type and level of risk an organisation is willing to accept in pursuing objectives.

Conflict of interest: An actual, potential or perceived interest that may improperly influence judgement.

Outsourcing governance: Oversight of third parties performing activities on behalf of an insurer.

Whistleblowing: Reporting suspected wrongdoing through protected reporting channels.

Conduct control: A control designed to prevent or detect harmful treatment or outcomes.

Data minimisation: Collecting and using only data necessary for a specified legitimate purpose.

Model risk: The possibility that an analytical model is unsuitable, inaccurate, biased or misused.

False positive: A case incorrectly classified as suspicious or fraudulent.

Precision: The proportion of flagged cases that are confirmed as positive cases.

Recall: The proportion of all positive cases that are successfully identified.

Social inflation: Liability-cost growth associated with legal, social and behavioural developments beyond ordinary price inflation.

Scenario planning: Testing strategies against multiple plausible future conditions.

Autonomous system: A system capable of performing tasks or making decisions with limited direct human control.

Vulnerable customer: A customer whose circumstances may increase the risk of harm or difficulty in accessing fair service.

Review Questions

Short Answer Questions

1. What is the relationship between claims governance and enterprise risk management?
2. Why does outsourcing not transfer ultimate accountability from an insurer?
3. Identify four principles of ethical claims handling.
4. What is the difference between precision and recall in fraud analytics?
5. Why are whistleblowing systems important to claims governance?

Long Answer Questions

1. Explain how a claims function can align fraud control with customer fairness and conduct risk management.
2. Discuss the governance controls required across the lifecycle of an outsourced claims service provider.
3. Evaluate the ethical and operational risks of using artificial intelligence to triage claims.
4. Design a capability-development programme for a claims workforce facing climate risk and autonomous systems.
5. Explain how scenario planning can improve strategic preparedness for social inflation and evolving fraud.

Analytical / Application Questions

1. An insurer rewards an external adjuster solely for reducing average settlement cost. Analyse the governance and conduct risks and propose redesigned measures.
2. A fraud model flags twice as many claims from one geographic area as from others, but confirmation data are incomplete. What validation, fairness and operational steps should management take?
3. Following a major flood, claims volumes exceed capacity while deepfake documents and organised repair fraud increase. Develop a 90-day response plan covering customers, vendors, data, investigations and board reporting.

Multiple Choice Questions with Answers

Question 1

Which statement best describes claims governance?

A. The process of paying every claim without investigation

B. The system for directing, controlling and monitoring claims decisions

C. A method used only by external fraud investigators

D. The accounting treatment of insurance premiums

Correct Answer: B. The system for directing, controlling and monitoring claims decisions

Question 2

Which is an example of conduct risk in claims?

A. Using a documented authority limit

B. Providing accessible communication to a vulnerable claimant

C. Delaying a legitimate claim to improve monthly closure statistics

D. Independently reviewing a disputed decision

Correct Answer: C. Delaying a legitimate claim to improve monthly closure statistics

Question 3

An insurer flags 100 claims and confirms 40 as fraudulent. What is the observed precision?

A. 20%

B. 40%

C. 50%

D. 60%

Correct Answer: B. 40%

Question 4

Which principle requires collecting only data necessary for a specified legitimate purpose?

A. Data minimisation

B. Unlimited retention

C. Purpose avoidance

D. Result manipulation

Correct Answer: A. Data minimisation

Question 5

What does outsourcing generally transfer?

A. Ultimate board accountability

B. The insurer’s legal identity

C. Performance of specified activities

D. All regulatory obligations

Correct Answer: C. Performance of specified activities

Question 6

Which control is most directly relevant to a conflict of interest?

A. Automatic deletion of all claim files

B. Disclosure, recusal and independent review

C. Increasing settlement targets

D. Removing all customer communication

Correct Answer: B. Disclosure, recusal and independent review

Question 7

Why should a material adverse automated claims decision normally include human oversight?

A. To eliminate the need for data security

B. To ensure accountability, contextual review and challenge

C. To guarantee that every claim is paid

D. To prevent any use of analytics

Correct Answer: B. To ensure accountability, contextual review and challenge

Question 8

Social inflation refers primarily to:

A. Only increases in consumer prices

B. Liability-cost growth influenced by legal and social developments

C. A reduction in claims frequency after a catastrophe

D. Currency conversion losses

Correct Answer: B. Liability-cost growth influenced by legal and social developments

Question 9

Which feature is essential in a credible whistleblowing programme?

A. Retaliation against unsubstantiated reporters

B. A single channel controlled by the accused manager

C. Protected reporting and independent case triage

D. Automatic publication of every reporter’s identity

Correct Answer: C. Protected reporting and independent case triage

Question 10

Scenario planning is best understood as:

A. Predicting one certain future

B. Testing strategies against multiple plausible futures

C. Replacing all risk controls with forecasts

D. Calculating only historical claim averages

Correct Answer: B. Testing strategies against multiple plausible futures