Model Audit, Governance, Ethics, and Communication

Introduction

A financial model is not merely a spreadsheet. It is a structured representation of economic assumptions, accounting relationships, financing arrangements, valuation techniques, and managerial judgments. Because models influence investment, lending, acquisition, budgeting, restructuring, and strategic decisions, errors in a model can create losses far exceeding the cost of preparing it. A model that is mathematically correct but poorly governed may still be unreliable; a model that is technically sophisticated but inadequately communicated may lead decision makers to misuse it.

Model audit and governance therefore form an integrated discipline. Audit tests whether the model behaves as intended. Governance establishes who may build, change, approve, use, and retire it. Ethics determines how the professional should act when assumptions are uncertain, incentives conflict, or results are commercially inconvenient. Communication converts model outputs into decision-relevant evidence while making uncertainty and limitations visible.

This chapter develops a practical framework for model assurance in a postgraduate finance environment. It covers error checking, reconciliation, validation, version control, peer review, documentation, data lineage, valuation standards, confidentiality, ethical judgment, and communication through dashboards and recommendations.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Distinguish model review, audit, validation, and governance.
2. Design structural, accounting, logical, and sensitivity checks.
3. Reconcile model outputs to source data and independently calculated benchmarks.
4. Apply version control, documentation, peer review, and change-approval procedures.
5. Assess model risk, data lineage, assumptions, limitations, and valuation standards.
6. Identify ethical risks involving bias, confidentiality, conflicts of interest, and misleading presentation.
7. Communicate model results to technical and non-technical decision makers.

1. The Model Assurance Lifecycle

Model assurance is the collection of activities that provide reasonable confidence that a model is fit for its intended purpose. “Reasonable confidence” does not mean that a model is free from all uncertainty. Forecasts necessarily contain judgment and estimation error. Assurance instead asks whether the model is conceptually appropriate, correctly implemented, adequately controlled, and transparently reported.

A useful lifecycle is:

1. Purpose definition: identify the decision, users, time horizon, material outputs, and required level of precision.
2. Conceptual design: specify relationships among operating drivers, financial statements, financing, tax, and valuation.
3. Data and assumption sourcing: record the origin, date, owner, units, and transformation of every material input.
4. Build and implementation: construct the model using controlled structures, transparent formulas, and appropriate safeguards.
5. Testing and review: perform self-checks, independent review, reconciliations, scenario tests, and validation.
6. Approval and release: document findings, resolve issues, assign an approval status, and freeze the released version.
7. Use and monitoring: track changes in assumptions, actual-versus-forecast performance, and continuing suitability.
8. Retirement: archive the model, data, approvals, and final outputs when it is no longer appropriate.

Model risk arises when model outputs are wrong, misunderstood, misapplied, or used outside their design conditions. It is affected by model complexity, data quality, estimation uncertainty, implementation errors, user behavior, and governance weakness.

2. Types of Model Review

These terms are related but should not be treated as synonyms.

An audit may identify a broken formula without concluding that the underlying valuation methodology is appropriate. Validation examines the wider question of suitability, including theory, assumptions, data, limitations, and performance. In practice, a major valuation model may require all five activities.

3. Error Taxonomy and Control Design

Effective review begins by classifying errors rather than merely searching for visibly unusual numbers.

3.1 Structural errors

Structural errors concern the architecture of the workbook or model. Examples include broken links, hidden worksheets containing active calculations, inconsistent time periods, accidental circular references, hard-coded values in formula regions, and formulas copied one column too far. A model can produce plausible numbers despite these defects, making structural review particularly important.

3.2 Formula and implementation errors

These include incorrect signs, wrong cell references, omitted terms, incorrect tax treatment, misplaced parentheses, and inconsistent use of absolute and relative references. A classic error is applying a growth rate to an accounting balance when the rate should apply to a flow or operational driver.

3.3 Accounting and reconciliation errors

Financial statements should connect through explicit relationships. For a basic balance sheet:

The identity is a condition of balance, not proof that the statements are economically correct. For a cash-flow statement using the indirect method:

The signs and classifications must be defined consistently. A model may balance because an unexplained “plug” has been inserted. Therefore, a balance check should be supplemented by a review of the bridge from opening to closing balances.

3.4 Logical and economic errors

A logical error occurs when the model does not reflect the intended business rules. For example, debt may continue to grow after a contractual maturity date, a tax loss may be used after expiry, or a capacity constraint may be ignored. Economic errors arise when relationships are implausible: working capital may fall when sales grow rapidly, or margins may expand indefinitely without an operational explanation.

3.5 Data and unit errors

Data errors include stale, duplicated, incomplete, or incorrectly mapped data. Unit errors are especially dangerous: revenue reported in millions may be combined with costs reported in thousands; percentages may be entered as 15 rather than 0.15; monthly rates may be applied as annual rates.

3.6 Presentation errors

A negative cash balance may be displayed in a way that resembles positive cash; a chart may truncate its axis and exaggerate a difference; or a dashboard may show a point estimate without its scenario range. Presentation is a control issue because users often rely on summaries rather than inspecting calculations.

4. A Layered Error-Checking Framework

No single check is sufficient. A strong model uses several layers of controls.

4.1 Mechanical checks

Mechanical checks test whether the model is structurally intact. Examples include:

• formula cells versus hard-coded cells;
• missing or broken links;
• duplicate or inconsistent formulas across periods;
• circular references;
• unexpected blanks and error values;
• hidden rows, columns, sheets, and named ranges;
• dates, units, and sign conventions;
• formulas overwritten by values.

A control cell can return “OK” only when all component tests pass. For example, a consolidated status may be represented conceptually as:

Here, Balance breach is the difference between total assets and total liabilities plus equity, and Tolerance is a preapproved materiality threshold expressed in the same currency units. A zero tolerance is appropriate where exact balancing is expected; a non-zero tolerance may be justified by rounding, but it must not conceal material errors.

4.2 Accounting checks

Reconcile each statement and major schedule independently. Examples include:

• total revenue to product or segment schedules;
• EBITDA to operating profit plus depreciation and amortisation;
• debt closing balance to opening balance plus drawdowns minus repayments;
• fixed assets to opening assets plus capital expenditure minus depreciation and disposals;
• retained earnings to opening retained earnings plus profit minus dividends;
• cash to the bank or treasury schedule;
• enterprise value to the stated bridge from equity value.

A reconciliation difference is:

The difference should be investigated, not automatically written off. Its materiality depends on the decision, currency scale, and risk of the item.

4.3 Reasonableness checks

Reasonableness checks compare outputs with historical patterns, contractual terms, industry evidence, and business capacity. Useful questions include:

• Does growth exceed the addressable market or physical capacity?
• Do margins depart from historical or peer levels without explanation?
• Does working capital behave consistently with payment terms?
• Does debt repayment comply with the financing agreement?
• Does the implied terminal value dominate enterprise value?
• Are forecast cash flows consistent with the narrative presented to management?

Reasonableness is not a demand that the future resemble the past. It is a demand that deviations be explained.

4.4 Sensitivity and scenario checks

A sensitivity test changes one or more assumptions to observe output behavior. A scenario combines internally coherent assumptions, such as a recession affecting volume, price, margin, receivables, and financing costs together.

For a discounted cash-flow valuation, enterprise value is commonly expressed as:

where EV is enterprise value, FCF_t is free cash flow in period t, n is the explicit forecast period, WACC is the weighted average cost of capital expressed as a decimal per period, and TV is terminal value at the end of period n. The formula assumes that cash flows occur at the end of each period and that the discount rate is consistent with the risk, currency, inflation basis, and timing of the cash flows.

Under a perpetual-growth terminal-value method:

where FCF_(n+1) = FCF_n × (1 + g) and g is the perpetual growth rate. The applicability condition is WACC > g; otherwise the expression is economically invalid or undefined. The growth rate must be sustainable and consistent with the long-run economy and business characteristics.

Sensitivity tables should test such variables as revenue growth, operating margin, WACC, terminal growth, capital expenditure, working capital, and exit multiples. They should not be presented as probability distributions unless probabilities have been justified.

5. Reconciliation and Independent Reperformance

Reperformance means calculating a result independently rather than merely checking that the original formula exists. It is one of the strongest review techniques.

Suppose a model forecasts revenue of 120 million, an operating margin of 25%, depreciation of 8 million, capital expenditure of 12 million, an increase in working capital of 5 million, and cash tax of 4 million. A simplified unlevered free-cash-flow calculation is:

Assuming EBIT equals operating profit of 30 million and a tax rate of 30%:

The units are millions of currency units. This calculation assumes that the listed items are unlevered, that depreciation is non-cash, that capital expenditure is a cash outflow, and that the increase in net working capital is a use of cash. A reviewer should independently reproduce 12 million and then verify whether each input is sourced and classified correctly. The arithmetic can be correct even if the forecast margin or working-capital assumption is unreasonable.

Reconciliation should operate at three levels:

1. Input reconciliation: source data agrees with the imported or entered data.
2. Schedule reconciliation: supporting schedules agree with statement line items.
3. Output reconciliation: valuation, returns, leverage, and dashboard measures agree with independent calculations.

6. Documentation and Data Lineage

Documentation is part of the model, not an administrative afterthought. A competent user should be able to understand the model without relying on the memory of its builder.

A model file should identify its purpose, scope, owner, users, reporting date, currency, units, time convention, scenario definitions, key assumptions, limitations, and approval status. Important formulas should be explainable in plain language. Complex calculations should have a separate methodology note or calculation inventory.

Data lineage records the journey of an input from origin to output. A useful lineage record includes:

Lineage is essential when data is extracted from enterprise systems, vendor databases, APIs, or generative-AI tools. The reviewer must know not only what value was used, but also when, by whom, under what definition, and through which transformations it passed.

7. Version Control, Access, and Change Management

A file named “final_v7_revised_latest.xlsx” is not a control system. Version control should provide a unique identifier, date and time, author, change description, approval status, and relationship to preceding versions. The released model should be read-only for ordinary users, with a controlled process for amendments.

A change log should distinguish:

• correction of an error;
• update of actual data;
• change in business assumption;
• methodological change;
• formatting or presentation change;
• emergency change made after approval.

Material changes require impact assessment and re-review. A useful change-control record identifies the affected outputs, whether prior conclusions remain valid, who reviewed the change, and whether the model must be reapproved.

Access should follow least privilege. Builders may edit; reviewers may comment or approve; decision makers may use outputs; administrators manage permissions. Sensitive acquisition, financing, customer, employee, and personal data should be encrypted where appropriate, shared only with authorized persons, and retained according to legal and organizational requirements.

8. Peer Review and Model Validation

Peer review should be genuinely independent. Independence does not always require an external consultant, but the reviewer should not be the sole author and should have authority to challenge assumptions.

An effective review plan includes:

1. understanding the decision and intended use;
2. inspecting architecture and documentation;
3. tracing material inputs to source evidence;
4. testing representative and high-risk formulas;
5. reperformance of key outputs;
6. challenging assumptions and boundary conditions;
7. testing scenarios, sensitivities, and error handling;
8. assessing limitations and communication;
9. recording findings by severity and remediation status.

Findings can be classified as critical, high, moderate, or low. A critical finding may invalidate a valuation or create a material risk of loss. A high finding may affect important outputs but be remediable before release. Low findings may concern clarity or maintainability. Severity should reflect impact, likelihood, detectability, and the importance of the decision—not simply the number of defective cells.

Validation should answer four questions:

• Conceptual soundness: Is the method appropriate for the purpose?
• Data quality: Are data definitions, completeness, timeliness, and transformations adequate?
• Implementation accuracy: Does the code or spreadsheet implement the method correctly?
• Performance and use: Does the model behave acceptably in normal, stressed, and boundary conditions, and are users applying it within its limits?

9. Valuation Standards and Professional Judgment

Valuation is not a single mechanical answer. The selected basis—market value, investment value, fair value, liquidation value, or another basis—determines relevant assumptions and evidence. A report should clearly state the purpose, valuation date, standard or basis, premise, currency, scope, methodology, and restrictions on use.

Common methods include:

• income approaches, such as discounted cash flow;
• market approaches, such as trading or transaction multiples;
• cost or asset approaches, where replacement or net asset evidence is relevant.

Triangulation can improve judgment, but it does not justify averaging incompatible results without analysis. Differences may arise from growth, risk, control, liquidity, accounting definitions, or forecast quality. The professional should explain why one method is primary and how other methods corroborate or challenge it.

Key valuation controls include:

• matching nominal cash flows with nominal discount rates and real cash flows with real discount rates;
• matching debt-inclusive cash flows with an enterprise-value discount rate;
• ensuring terminal assumptions are sustainable;
• avoiding double counting of cash, debt, synergies, or tax benefits;
• reconciling enterprise value to equity value;
• documenting market-data dates and selection criteria;
• separating observed facts from management forecasts and analyst judgment.

Standards and regulatory requirements vary by jurisdiction and purpose. Professionals should consult applicable accounting, valuation, securities, audit, privacy, and financial-modeling guidance rather than assuming that one framework governs every assignment.

10. Ethics, Confidentiality, and Professional Judgment

Ethical model practice requires more than avoiding deliberate fraud. It requires resisting misleading omission, selective disclosure, unjustified precision, and pressure to produce a desired answer.

Important ethical risks include:

• manipulating assumptions to achieve a target valuation;
• hiding downside scenarios;
• presenting a range as a precise point estimate;
• using confidential information without authorization;
• failing to disclose conflicts of interest;
• copying a model without understanding its limitations;
• allowing algorithmic or historical data bias to influence decisions invisibly;
• using unsupported AI-generated sources or formulas;
• suppressing a material review finding.

When incentives conflict with professional judgment, the analyst should document the issue, distinguish fact from assumption, escalate through an appropriate reporting line, and refuse to certify a conclusion that cannot be supported. Confidentiality applies to source data, model logic, client identity, transaction terms, personal information, and unpublished results. Confidentiality does not justify concealing information from an authorized auditor, regulator, or decision maker who needs it to assess material risk.

A practical ethical test asks: Is the statement accurate? Is it complete enough not to mislead? Can it be supported by evidence? Would the presentation remain defensible if disclosed to an independent reviewer, client, regulator, or affected stakeholder?

11. Technology, Automation, and Artificial Intelligence

Automation can improve control by standardizing imports, running repeatable tests, comparing versions, and generating exception reports. It can also increase risk when a flawed transformation is applied repeatedly or when users trust a system they do not understand.

AI tools may assist with documentation, code review, anomaly detection, and scenario generation. They should not be treated as authoritative sources. AI-produced formulas, explanations, and data must be independently verified; confidential information should not be entered into an unapproved service; and the organization should retain a record of material AI use. Human accountability remains essential because an AI system may produce plausible but unsupported assumptions, fabricated citations, or inconsistent calculations.

A controlled technology environment should include access management, audit logs, tested interfaces, data-quality rules, reproducible scripts, exception handling, and a fallback process when automated systems fail.

12. Communicating Model Results

Communication is successful when the decision maker understands what action the model supports, why the result changes, how uncertain it is, and what could invalidate it. A technical appendix cannot compensate for a misleading executive summary.

A useful communication structure is:

1. Decision: What decision is being requested?
2. Conclusion: What does the model indicate?
3. Drivers: Which assumptions explain the result?
4. Range: How does the conclusion change under alternative scenarios?
5. Risks: What limitations, data gaps, or external conditions matter?
6. Recommendation: What should management do, and what should be monitored?

Dashboards should show a limited set of decision-relevant indicators, such as revenue, margin, cash conversion, leverage, liquidity headroom, valuation range, and scenario status. Each indicator should have a definition, unit, period, source, owner, and threshold where relevant. Traffic-light colors require explicit rules; “green” should not mean merely “better than last month.”

Charts should preserve scale, label units, identify actual versus forecast values, and avoid implying precision beyond the data. A valuation range should be accompanied by its methodology and key sensitivities. Recommendations should distinguish model evidence from managerial judgment.

13. Illustrative Case: Acquisition Valuation Under Review

Background

Northstar Foods, a private company, is considering acquiring Riverbend Beverages. Management presents a spreadsheet indicating an enterprise value of 250 million and claims that the acquisition will create substantial value. The model includes five years of forecasts and a terminal value.

Problem

An independent reviewer is asked to determine whether the model is suitable for the investment committee. The transaction is confidential, and the investment team has a strong incentive to support the acquisition.

Analysis

The reviewer identifies the following issues:

• revenue in the source statements is in thousands, while one imported schedule is in millions;
• the debt schedule excludes a mandatory repayment in year three;
• the terminal value uses a perpetual growth rate of 4.0% and a WACC of 4.2%, making the valuation extremely sensitive because the denominator is only 0.2 percentage points;
• projected margins rise materially above historical levels without identified capacity or procurement initiatives;
• management synergies are included in cash flows but the transaction narrative does not identify implementation costs;
• the dashboard reports only the base case;
• the file has no reliable change log.

The reviewer independently recalculates revenue, debt, and free cash flow; corrects the unit mismatch; adds the contractual debt repayment; and separates standalone cash flows from synergies. The valuation is then presented under downside, base, and upside cases. Rather than reporting a single figure, the report identifies a range and explains that the terminal value is a major driver.

Approach and solution

The review team establishes a controlled release version, records all corrections, obtains evidence for each synergy, and requires management to approve explicit assumptions. The investment committee receives a dashboard showing valuation range, leverage, liquidity, sensitivity to WACC and terminal growth, and key due-diligence actions.

The recommendation is conditional: proceed only if the purchase price remains below a defined decision threshold after validated synergies and if debt-service headroom remains adequate under the downside case.

Outcome

The committee does not reject the transaction solely because the original model contained errors. Instead, it defers approval until the data and assumptions are corrected. The revised analysis produces a lower and wider valuation range, leading to a renegotiated offer and stronger protection against downside risk.

Learning points

The case demonstrates that model assurance is not an attempt to eliminate judgment. It makes judgment visible, tests its consequences, protects confidentiality, and improves the quality of the decision. A balanced model is not necessarily a valid model; a precise output is not necessarily an accurate conclusion; and a favorable result is not evidence of a sound model.

Challenges and Limitations

Model audit cannot eliminate uncertainty in forecasts, market prices, management behavior, or external events. Review is also constrained by time, access to data, reviewer expertise, and the possibility that errors remain undiscovered. Excessive controls can slow legitimate decision making, while weak controls can create unrecognized exposure.

Materiality should be applied intelligently. A small numerical error in a highly sensitive terminal-value calculation may matter more than a larger error in an immaterial line item. Similarly, a model may be appropriate for strategic scenario analysis but inappropriate for regulatory capital or transaction pricing. Fitness for purpose is therefore central.

Finally, governance can fail through culture. If employees fear escalation, if reviewers are rewarded for speed rather than challenge, or if senior executives demand predetermined outcomes, formal checklists will not provide effective assurance.

Best-Practice Checklist

Before release, the responsible team should be able to answer “yes” to the following:

• Is the purpose and intended use explicit?
• Are assumptions, units, dates, and sources documented?
• Are statements and schedules reconciled?
• Have key outputs been independently reperformed?
• Have downside, upside, and boundary conditions been tested?
• Has an appropriately independent person reviewed the model?
• Are findings resolved, accepted, or escalated formally?
• Is the released version identifiable and protected from unauthorized edits?
• Are valuation basis, methodology, and limitations disclosed?
• Are confidential data and model access controlled?
• Does the communication distinguish facts, assumptions, uncertainty, and recommendations?
• Is there a monitoring and update plan?

Chapter Summary

Model audit, governance, ethics, and communication are complementary elements of financial-model assurance. Mechanical checks identify structural defects; accounting reconciliations test financial integrity; reasonableness and sensitivity analysis challenge economic logic; and independent reperformance tests whether important outputs can be reproduced. Documentation and data lineage make inputs, transformations, assumptions, and limitations traceable. Version control, access management, peer review, validation, and change approval provide accountability throughout the model lifecycle.

Valuation requires a clearly stated purpose, basis, methodology, evidence, and treatment of uncertainty. Discounted cash flow formulas are useful only when cash flows, discount rates, timing, currency, inflation, and terminal assumptions are consistent. Ethical practice requires confidentiality, independence, disclosure of conflicts, resistance to target-driven manipulation, and responsible use of technology and AI. Finally, effective communication translates model evidence into decisions without concealing uncertainty or overstating precision.

Key Terms

Model assurance: Activities that provide reasonable confidence that a model is fit for purpose.

Model audit: Independent examination of model structure, controls, calculations, and evidence.

Model validation: Assessment of conceptual suitability, data, implementation, performance, and use.

Model governance: Framework of responsibilities, approvals, controls, and oversight.

Reconciliation: Comparison of model amounts with an independent benchmark or related schedule.

Reperformance: Independent recalculation of a model result.

Data lineage: Traceable record of an input’s origin, transformations, and destination.

Version control: System for identifying, preserving, and managing model revisions.

Change log: Record of amendments, authors, dates, impacts, and approvals.

Materiality: Significance of an error or uncertainty to a decision or user.

Reasonableness check: Assessment of whether an output is economically and operationally plausible.

Sensitivity analysis: Examination of output changes resulting from changes in selected assumptions.

Scenario analysis: Evaluation of internally coherent alternative futures.

Enterprise value: Value attributable to operating assets before allocation between debt and equity claims.

Terminal value: Value of cash flows beyond the explicit forecast period.

Data quality: Fitness of data regarding accuracy, completeness, timeliness, consistency, and relevance.

Data confidentiality: Protection of information from unauthorized access or disclosure.

Professional skepticism: Critical assessment of evidence and willingness to challenge assumptions.

Audit trail: Evidence showing what changed, when, by whom, and under what authority.

Fitness for purpose: Suitability of a model for its defined decision and conditions of use.

Review Questions

Short Answer Questions

1. Distinguish a model audit from model validation.
2. Why is a balance-sheet check insufficient to establish model reliability?
3. What information should a data-lineage record contain?
4. Explain the difference between sensitivity analysis and scenario analysis.
5. Identify three ethical risks in financial modeling.

Long Answer Questions

1. Explain how a layered error-checking framework can detect structural, accounting, logical, data, and presentation errors.
2. Discuss the role of version control, access management, and change approval in model governance.
3. Explain the assumptions and applicability conditions of the discounted cash-flow terminal-value formula.
4. Develop a model-validation plan for an acquisition valuation.
5. Discuss how a financial analyst should communicate uncertainty and limitations to an investment committee.

Analytical / Application Questions

1. A model balances, but revenue is imported in millions while costs are imported in thousands. Design tests that would detect the error and explain why the balance check may not do so.
2. A valuation uses a WACC of 6.1% and perpetual growth of 6.0%. Evaluate the governance and communication issues raised by this assumption spread.
3. Management asks an analyst to remove the downside case from a board dashboard because it may “confuse” decision makers. Apply ethical and communication principles to determine an appropriate response.

Multiple Choice Questions with Answers

Question 1

Which activity primarily assesses whether a model is conceptually suitable for its intended purpose?

A. Model validation

B. File formatting

C. Data entry

D. Dashboard styling

Correct Answer: A. Model validation

Question 2

Which relationship is a fundamental balance-sheet control?

A. Assets = Liabilities + Equity

B. Revenue = Cash + Debt

C. Profit = Assets − Revenue

D. Equity = Revenue + Inventory

Correct Answer: A. Assets = Liabilities + Equity

Question 3

A model forecasts EBIT of 30 million, a tax rate of 30%, depreciation of 8 million, capital expenditure of 12 million, and an increase in net working capital of 5 million. What is simplified unlevered free cash flow?

A. 8 million

B. 12 million

C. 21 million

D. 30 million

Correct Answer: B. 12 million

Question 4

Under the perpetual-growth terminal-value method, which condition is required for the standard formula to be economically valid?

A. WACC = 0

B. WACC < g

C. WACC > g

D. g must exceed inflation by 10%

Correct Answer: C. WACC > g

Question 5

What is the main purpose of data lineage?

A. To make a model visually attractive

B. To trace data origin, transformations, and destination

C. To eliminate all forecast uncertainty

D. To replace independent review

Correct Answer: B. To trace data origin, transformations, and destination

Question 6

Which practice provides the strongest evidence that a key model output can be reproduced?

A. Changing the font color

B. Reperformance using an independent calculation

C. Adding more worksheets

D. Hiding formula cells

Correct Answer: B. Reperformance using an independent calculation

Question 7

Which is an appropriate version-control practice?

A. Saving every file as “final”

B. Allowing unrestricted editing of the released model

C. Maintaining unique identifiers, change descriptions, and approvals

D. Deleting previous versions

Correct Answer: C. Maintaining unique identifiers, change descriptions, and approvals

Question 8

Which statement best describes a scenario analysis?

A. It changes one cell without considering related assumptions

B. It combines internally coherent alternative assumptions

C. It proves the base case is correct

D. It removes the need for professional judgment

Correct Answer: B. It combines internally coherent alternative assumptions

Question 9

An analyst is asked to hide a material downside case from decision makers. What is the most appropriate response?

A. Hide it because management owns the model

B. Present the downside and explain its assumptions and limitations

C. Delete the entire model

D. Replace it with an unsupported probability

Correct Answer: B. Present the downside and explain its assumptions and limitations

Question 10

Which statement about AI-generated financial-model content is most appropriate?

A. It is authoritative if it appears plausible

B. It requires independent verification and approved handling of confidential data

C. It eliminates the need for a reviewer

D. It can be used without recording its role

Correct Answer: B. It requires independent verification and approved handling of confidential data