Model documentation for review handover

Nothing here is financial advice, and none of it replaces your own checks.

Create one document that lets a reviewer trace the financial model from source data to reported outputs. Use this workflow when you need reviewer sign-off, a transaction handover, or a record of what changed. Nothing here is financial advice, and none of it replaces your own checks.

1. Assemble the handover folder

Create a dated working folder and place copies of the materials in it. Do not document from memory or from an uncontrolled workbook.

Collect:

• The current model workbook, including all visible and hidden sheets where your review policy permits access.
• The prior approved version, if there is one.
• The assumptions register, investment case, budget, forecast, or other document that authorised key inputs.
• Source data extracts, with file names, reporting periods and extraction dates.
• Change notes, reviewer comments and issue logs.
• The intended output, such as a board pack table, valuation schedule, covenant calculation or management report.
• The document template and sign-off requirements used by your team.

Record the workbook file name, file date, owner and reporting period at the top of the documentation. Give the document a version number that matches the workbook version.

2. Map the workbook before drafting

Open the workbook and make a sheet inventory. For each tab, record its name, purpose, owner where known, whether it contains inputs, calculations or outputs, and whether it is hidden.

Then identify the calculation flow. Start at the main output sheet and work backwards through linked schedules to input sheets. Note external links, named ranges, data queries, macros, manual paste areas and circular calculations. A reviewer needs to know where a number came from, not only where it appears.

Use a simple table in the document:

Sheet or range	Role	Feeds or depends on	Review point
Assumptions	User inputs	Operating forecast	Approved source and date
Revenue build	Calculation schedule	P&L output	Volume, price and timing logic
Summary	Reported output	Board pack tables	Agrees to supporting schedules

3. Extract inputs and assumptions

Create an inputs and assumptions section. Group assumptions by topic, not by the order in which you found them. Typical groups include revenue, costs, working capital, financing, tax, capital expenditure and scenario settings.

For every material assumption, record:

• The worksheet and cell or named range.
• The input value and unit, such as percentage, currency, days or headcount.
• The applicable period.
• The source document, source owner and source date.
• Whether it is hardcoded, linked, calculated or manually updated.
• The rationale or approval reference, if supplied.
• Any sensitivity or scenario in which it changes.

Do not convert an unexplained hardcode into a plausible rationale. Mark it as source not evidenced and add it to the open issues list.

If you use the model you are using to organise notes or draft prose, provide only the required extracts and label each item as source, assumption, calculation or comment. Its file and data handling behaviour can be version-dependent, so check the relevant guidance in the xAI documentation overview before sharing confidential material.

4. Describe formulas and dependencies

Document material formulas, not every copied cell. Focus on formulas that drive outputs, allocate values, bridge periods, select scenarios, apply accounting or tax treatments, or use lookup logic.

For each material calculation, write four short parts:

1. Purpose: what the calculation is intended to produce.
2. Formula logic: describe the calculation in plain language, including signs, timing and units.
3. Inputs: list the ranges, schedules or external files it uses.
4. Output and dependency: state where the result flows next.

For example, write: “Monthly revenue equals active customers multiplied by average revenue per customer. Active customers are derived from opening customers, additions and churn. The result feeds the operating profit and cash flow schedules.” Do not paste a long formula without explaining the business logic.

Call out formulas that contain manual overrides, nested lookups, error suppression, volatile functions, circularity switches or links to another workbook. These are not automatically wrong, but they need explicit review.

5. Record controls and reconciliation results

List every control in the workbook and state whether it passed at the time of documentation. Include the sheet, cell, test purpose, expected result, actual result and reviewer status.

Useful controls include:

• Balance sheet balancing checks.
• Cash flow movement reconciliations.
• Opening-to-closing balance roll-forwards.
• Scenario selection checks.
• Source-data totals reconciled to imported values.
• Output totals reconciled to the board pack or management report.
• Error-cell scans for #REF!, #VALUE!, #DIV/0! and similar results.

Where a check fails, record the amount, the affected output, the likely cause, the owner and the required action. Do not describe an unresolved variance as immaterial unless an authorised reviewer has made that decision.

6. Create the limitations and change log

Add a known limitations section before the sign-off page. This section protects the next user from treating the workbook as more certain or complete than it is.

Include items such as incomplete source data, manual assumptions pending confirmation, external links unavailable to the recipient, calculations excluded from scope, unsupported historical periods and known control failures. State the consequence for the output, not just the technical defect.

Then create a change log covering the period since the prior reviewed version. For each change, record the date, person, workbook area, description, reason, affected outputs and whether controls were rerun.

7. Run the review and hand over

Read the document while following the workbook yourself. Select several material outputs and trace each one back to its source data and assumptions. Then select several assumptions and trace them forward to the reported outputs. This two-way test exposes missing dependencies and undocumented overrides.

The output is wrong if a reviewer cannot locate a cited range, an assumption has no source or date, a formula description does not match the workbook, a control result cannot be reproduced, or an output does not reconcile to its stated report. Treat inconsistent units, signs, periods and scenario labels as defects until resolved.

Prepare the handover pack with the model, documentation, source-data index, open issues list and change log. Name the reviewer, preparer, review date, scope and sign-off status. Keep sign-off blank until the designated reviewer has completed their work.

When the workflow does not work

Stop and narrow the scope if the workbook is too large to document reliably in one pass. Start with the output sheets and their direct supporting schedules, then state the excluded areas in the limitations section. If source data, formula logic or change history is missing, raise an open issue with the model owner rather than filling the gap with an assumption. Return the document for correction when controls fail, links cannot be traced, or the workbook and documentation describe different logic.