Resource

Financial modelling best practices

The structural conventions, driver design, integrity checks and review procedures that make a three-statement model usable by a transaction team under time pressure.

01

Start from the decision the model must support

A model built to support a financing discussion, a valuation and a board paper will do all three badly unless the primary use is decided first. The intended decision determines the level of granularity, the periodicity, the length of the forecast horizon and the outputs that must be presented without further manipulation.

Before building, agree the deliverable: the specific schedules that will be circulated, who reviews them, and which sensitivities will be requested. Retro-fitting scenario capability into a model built as a single-case forecast is the most common cause of rebuild.

02

Structure and layout

Structure is what makes a model reviewable by someone who did not build it. The separation between what is assumed, what is calculated and what is presented should be visible on opening the file.

  • Three zones

    Inputs, calculations and outputs are kept in separate sheets. No hardcoded assumption sits inside a calculation block.

  • Consistent time axis

    One date row drives every sheet, in the same columns, with the same periodicity. Historical and forecast periods are flagged by a single switch row.

  • One row, one calculation

    A formula written in the first forecast column copies across the row unchanged. Any break in a row is a defect, not a feature.

  • Formatting as documentation

    Consistent colour conventions for inputs, formulas and links let a reviewer see the model's logic before reading a single formula.

  • No hidden logic

    Avoid hidden rows and sheets, merged cells, circularity where an algebraic alternative exists, and macros unless the deliverable requires them.

03

Driver design

The credibility of a forecast rests on whether the drivers reflect how the business actually earns money. A revenue line grown at a single percentage is an assertion; a model built from volume, price and mix is a position that can be tested against diligence findings.

  • Operational drivers

    Build revenue from the units management actually manages — customers, contracts, capacity, utilisation, cohorts, price per unit.

  • Cost behaviour

    Split fixed, variable and step costs explicitly so operating leverage falls out of the model rather than being assumed.

  • Working capital

    Forecast on days ratios derived from the historical trading cycle, and confirm the ratios reconcile to the diligence position.

  • Capital expenditure

    Separate maintenance from growth capex, and link growth capex to the capacity the revenue build assumes.

  • Assumption provenance

    Every material input carries a source note: management pack, diligence report, contract, or explicitly labelled analyst estimate.

04

Three-statement integrity

A model that does not balance is not a model. Integration between the income statement, balance sheet and cash flow must be structural rather than plugged, and the checks that prove it must be visible.

  • Balance sheet check

    A single check row, live on every sheet, showing assets less liabilities and equity to the nearest currency unit.

  • Cash flow derivation

    Cash flow is derived from movements in the balance sheet and income statement, never entered independently.

  • Debt and interest

    Opening balance, draw, repayment, closing balance for every facility, with interest on average balances and an explicit treatment of circularity.

  • Tax

    Model current tax on taxable profit with losses carried forward, and keep deferred tax separate where it is material to the equity bridge.

  • Check panel

    One sheet aggregating every check in the model, returning a single pass or fail flag visible from the outputs.

05

Scenarios and sensitivity

Scenario capability should be designed in, not added later. A clean scenario architecture lets the model answer the questions asked in a meeting rather than in the week after it.

  • Single switch

    One scenario selector drives all scenario-dependent assumptions through a lookup block; no scenario logic embedded in calculations.

  • Named cases

    Management case, base case and downside case are defined by what changes and why, and documented alongside the numbers.

  • Break-even analysis

    Identify the level of each key driver at which covenants breach, funding runs out, or returns fall below hurdle.

  • Sensitivity tables

    Two-way tables on the outputs that matter — equity value, IRR, leverage at exit, minimum liquidity.

06

Review before circulation

Review is a defined step with its own procedure, not a read-through. The objective is to find the errors that change a decision, in the time available.

  • Formula consistency

    Scan each row for inconsistent formulas, hardcodes inside calculations and broken or external links.

  • Reasonableness

    Compare forecast margins, growth and working capital days to the historical period and to the diligence findings; explain every discontinuity.

  • Extreme value testing

    Set key drivers to zero and to implausible highs; the model should behave predictably and the checks should hold.

  • Reconciliation

    Tie historical periods to audited or diligence-adjusted figures and document any difference.

  • Version control

    Dated file naming, a change log of material revisions, and one owner of the live version at any time.

Discuss an Engagement

If you require support with financial modelling, business valuation or financial due diligence for a live transaction or strategic engagement, we'd be pleased to discuss your requirements.

Discuss an Engagement