Safety Norms: Corporate Finance and Financial Modeling

1Purpose, Scope, and Safety Principles

These norms protect the organization, its shareholders, employees, lenders, and counterparties from harm caused by flawed financial analysis, unauthorized changes to financial records, misuse of confidential information, and fraud. In this context, safety means reliable numbers, controlled access, and decisions made by authorized people using accurate information.

The norms apply to anyone who builds, reviews, approves, or uses corporate finance analysis, including FP and A staff, treasury, controllers, tax and accounting teams, investor relations, and external advisors working under contract. They cover spreadsheets, modeling software, enterprise planning systems, data extracts, and presentations drawn from them.

• Mandatory: no model output supports a binding decision until an independent reviewer has checked it against documented assumptions.
• Mandatory: financial information is handled only for authorized business purposes.
• Recommended: treat every material model as a controlled document with an owner, version history, and stated purpose.

2Roles, Competence, and Accountability

Accountability rests with named individuals, not with a general team. The model owner is responsible for design, assumptions, and documentation. The reviewer checks logic and reasonableness. The approver authorizes use of results in budgets, pricing, financing, or acquisition decisions. The controller or designated control owner sets access rules and reviews exceptions. Where a role is vacant, the most senior finance leader present assigns it in writing.

Competence is a prerequisite for authorization. Modelers should understand the accounting principles relevant to the entity, the valuation methods they use, and the tools themselves. Authority to post journal entries, change system rates, or approve payments should be granted through a documented process and reviewed regularly. Interns and advisors work under supervision and hold no approval rights unless the organization grants them formally.

• Mandatory: each model has a named owner, reviewer, and approver recorded before use.
• Mandatory: no individual approves their own model, payment, or journal entry.
• Recommended: maintain a competence matrix listing skills, completed training, and access granted for each role.

3Realistic Hazards and Risk Assessment

The main hazards in this subject are not physical. They include formula errors, broken links, hard-coded plugs, stale ledger extracts, currency or unit mistakes, aggressive assumptions presented as base cases, and deliberate manipulation of inputs to support a predetermined result. Each can lead to misallocated capital, misstated forecasts, covenant breaches, tax exposure, or financial loss.

Owners should rate risk by likelihood and impact using a simple scale such as low, medium, and high, and reviewers should confirm the rating. Models that feed external reporting, debt covenants, transaction pricing, or incentive compensation are normally rated high and need the strongest controls. Exact rating criteria should be set by local policy.

• Errors: broken links, circular references, hard-coded numbers inside formulas, and mismatched time periods.
• Data: unreconciled balances, outdated extracts, and inconsistent chart of accounts mapping.
• Judgment: unsupported growth rates, terminal value assumptions outside documented ranges, and downside cases removed before presentation.
• Fraud and misconduct: altered inputs, backdated versions, concealed adjustments, and pressure to deliver a target result.
• Confidentiality: sharing unreleased results, deal terms, or valuation outputs through personal email or unapproved cloud tools.

4Preventive Controls and Segregation of Duties

Preventive controls reduce the chance of error or fraud before it occurs. Segregation of duties is the central control. The person who initiates a transaction, builds a model, changes a master data field, or records a payment should not also approve, release, or reconcile it. Where staffing is small, compensating controls such as dual sign-off, independent review of system logs, or external review must be documented and approved by the controller.

Disciplined model design is also a control. Inputs should be separated from calculations, assumptions should be labeled with sources, and outputs should flow through built-in checks. Changes to live models should follow a documented change process. Each material change should record the reason, author, approver, and affected outputs.

• Mandatory: maintain an assumptions register with the source, owner, and last update reference for every key input.
• Mandatory: include integrity checks such as balance sheet balancing, cash tie-out to the cash flow statement, and totals agreeing to source reports.
• Mandatory: protect formula sheets in production models and restrict edit rights to the model owner.
• Recommended: run sensitivity and scenario analysis for any decision above a threshold set by local policy.
• Recommended: use unique version identifiers and retire superseded files from shared folders.

5Safe Modeling and Decision Practices

Before a model supports a decision, the analyst should confirm its purpose, the decision maker, and the precision the decision requires. Base cases should reflect management-approved or evidence-supported assumptions, not targets. When management selects an assumption, the model should record who selected it, why, and what evidence supports it. Presentations should show the range of outcomes, not only the most favorable case.

Valuation, capital budgeting, and financing models rely on methods such as discounted cash flow, net present value, internal rate of return, and leverage analysis. Each method has limits. Terminal value often drives a large share of total value, so its assumptions need explicit review. Results should never be described as certain. Transaction, impairment, and tax-related analyses should be reviewed by qualified professionals.

• Mandatory: disclose every manual override, plug, or adjustment in the model along with its approver.
• Mandatory: obtain the designated approver's sign-off before any model output is shared outside the authorized group.
• Recommended: compare key outputs with an independent benchmark, such as a prior approved version or a simplified cross-check.
• Recommended: avoid circular references unless a documented control and iteration limit are in place.

6Tools, Environment, and Data Confidentiality

Financial models often contain nonpublic earnings, pricing, customer, employee, and transaction information. Confidentiality norms apply to spreadsheets, planning software, shared drives, email, collaboration tools, printouts, and screens viewed in open offices or during travel. Access follows least privilege: people receive only the data and edit rights their role requires, and access is removed promptly when roles change.

Personal devices, unapproved AI tools, and public file sharing services must not be used for company financial data unless the organization has approved them in writing and a privacy or security function has reviewed the arrangement. Privacy and data protection obligations depend on jurisdiction and should be confirmed with counsel or the privacy office.

• Mandatory: store models only in approved repositories with access logging and backups.
• Mandatory: never paste nonpublic financial data into external tools the organization has not approved.
• Mandatory: protect insider information about planned transactions, results, or guidance in line with applicable securities rules and company policy.
• Recommended: use screen privacy in shared workspaces and clear whiteboards after confidential sessions.
• Recommended: restrict macros and external links in models to approved items, since these can introduce malicious code or hidden data flows.

7Regulatory Compliance and Accurate Records

Corporate finance work supports statutory reporting, tax filings, lender reporting, and securities disclosures. Requirements come from accounting frameworks, securities rules, tax law, debt agreements, and internal policy, and they vary by jurisdiction, industry, and entity type. This document does not state any specific legal requirement. Local compliance, legal, and audit professionals must confirm which rules apply and which versions are current.

Accurate records are the foundation of safe finance practice. Every material model should trace to source data, approved assumptions, and the reasons for changes. Records should be retained under the organization's retention schedule and any legal hold. Models used in audits, financings, or disputes should be preserved in their approved form.

• Mandatory: reconcile model historical figures to the general ledger or other system of record before use.
• Mandatory: report suspected misstatements to the controller and never adjust records to force a balance.
• Recommended: keep a compliance map listing each model, its governing rules, and its owner.
• Recommended: have counsel review forecast disclosures and covenant calculations where thresholds are tight.

8Incident Reporting, Fraud Escalation, and Response

Anyone who finds an error, near miss, control failure, or suspicious activity must report it promptly through the channel set by local policy. Warning signs include requests to bypass review, unexplained input changes after approval, round-number adjustments without support, vendor bank detail changes received only by email, pressure to reach a predetermined result, access to files outside a job role, and missing or altered supporting documents. Good-faith reporting should be protected by policy, and no one should investigate alone.

Do not confront the person involved, alter the file, delete logs, or discuss the matter outside the response team. Preserve the current version and related evidence. The controller, compliance function, internal audit, or legal counsel decides investigation steps. If there is a risk to life, a safety threat, or an immediate threat of significant loss, contact emergency services or the appropriate authority as local procedure directs.

• Mandatory: report known or suspected fraud within the time frame required by local policy, and do not wait for certainty.
• Mandatory: freeze the current model and retain system logs before any correction is made.
• Mandatory: do not discuss suspected fraud with the individual involved until the response lead directs.
• Recommended: run a post-incident review that identifies the control gap, its owner, and the corrective action.

9Training, Monitoring, and Review

Training should cover model design standards, change control, data handling, fraud warning signs, and escalation routes. New staff complete introductory training before receiving approval rights, and everyone refreshes training when systems, policies, or regulatory requirements change. Effectiveness can be checked through review findings, error rates from control testing, and questions raised by staff.

Monitoring confirms that these norms work in practice. Control owners review access lists, model inventories, change logs, and exception reports on a schedule set by local policy. Findings lead to documented actions with named owners and completion status. These norms should also be reviewed whenever the organization changes systems, enters new markets, adopts new reporting rules, or experiences an incident.

• Mandatory: maintain a model inventory listing owner, risk rating, purpose, and storage location.
• Mandatory: review finance system access rights on a set schedule and remove rights no longer needed.
• Recommended: independently review a sample of models and report results to the controller.
• Recommended: record lessons learned in a shared log and update templates and checklists accordingly.

Implementation checklist

• Confirm each model has a named owner, reviewer, and approver before use.
• Check that the person approving a payment or journal entry did not prepare it.
• Reconcile model historical figures to the ledger and document any differences.
• Record every manual override, plug, and key assumption with its source and approver.
• Run balance and cash tie-out checks before any output leaves the team.
• Store models only in approved repositories and remove unauthorized copies.
• Do not paste nonpublic financial data into unapproved tools.
• Present base, downside, and sensitivity cases together when supporting a decision.
• Report suspicious changes or pressure to alter results to the controller or compliance lead.
• Preserve the current file and system logs before correcting any suspected error.