What separates a model from a spreadsheet
Most businesses have a spreadsheet. Far fewer have a model.
A spreadsheet contains numbers, frequently hard-coded, sometimes right. A model has a structure: inputs separated from calculations, calculations separated from outputs, assumptions in one place and changeable, and a set of checks that fail loudly when something does not balance.
The practical difference shows up the first time somebody asks a question. With a spreadsheet, answering it means rebuilding. With a model, it means changing an input and reading the result. That is the entire value, and it is why the structural discipline matters more than the sophistication of the arithmetic.
Which businesses this applies to
Businesses raising investment or debt, where the counterparty will interrogate the model rather than read it. Businesses evaluating a significant project — a new location, a production line, an acquisition — where the decision is large enough to justify testing.
Businesses with a bank facility and covenants that need forecasting. Businesses considering a restructuring. And businesses whose plan currently exists only in the owner’s head, which works until somebody external needs to be persuaded by it.
The work, step by step
What this looks like in practice:
- Agree what the model is for. A fundraising model, a project appraisal and an operational planning tool are different builds, and building one when another was needed wastes the whole exercise.
- Design the structure first — input sheets, calculation sheets, output sheets, clearly separated and consistently formatted.
- Build the revenue engine from drivers, so that the model can be interrogated rather than only run.
- Build all three statements: profit and loss, balance sheet, cash flow, fully linked. A model without a balance sheet cannot be checked.
- Add the integrity checks — balance sheet balancing, cash tying to the cash flow statement, no circular references left unmanaged.
- Build the scenario layer, so cases are switched rather than rebuilt.
- Add sensitivity analysis on the variables that actually drive the outcome, which is frequently not the ones the owner expects.
- Document the assumptions with their source, because an assumption nobody can justify undermines everything built on it.
- Hand it over with training, so you can run it without us.
What makes a model credible to a third party
Investors, lenders and buyers read models constantly, and they form a view in the first few minutes. What they look for is consistent:
- Assumptions in one place, clearly labelled, with sources — nothing hard-coded into a formula
- A balance sheet that balances, with a visible check. Its absence is the fastest way to lose a reader
- Working capital modelled on days rather than as a percentage of revenue
- Sensible scenarios including a genuine downside, because a model with only an optimistic case tells the reader something about the author
- Consistent formulas across rows — an inconsistency is where errors hide, and experienced readers test for it
- Clear formatting conventions so inputs are visually distinguishable from calculations
A model that fails these tests undermines the plan it was built to support, however good the underlying business is. Fair or not, the model is taken as evidence of how the business is run.
The assumptions do the work
The arithmetic in a financial model is rarely difficult. The judgement is entirely in the assumptions, and that is where models are won and lost.
A revenue ramp that assumes a market share nobody has achieved. A cost base that grows more slowly than revenue with no explanation of how. Receivable days materially better than the business has ever managed. A downside case that is only mildly worse than base. Each of those makes the output arithmetically correct and commercially worthless.
We will challenge assumptions during the build, and where we think one is not defensible we will say so before a third party does. That is uncomfortable and it is the point — a model whose assumptions do not survive scrutiny is worse than no model, because it costs credibility as well as time.
Common mistakes
The expensive mistakes in this area are consistent:
- Hard-coded numbers inside formulas, which make the model impossible to test and impossible to trust.
- No balance sheet, so there is no way to check that the model is internally consistent.
- Working capital as a percentage of revenue rather than modelled on days.
- Only an upside case, which tells a reader more about the author than about the business.
- Inconsistent formulas across a row, which is where errors hide.
- Assumptions with no source, undermining everything built on them.
- A model only its builder can operate, abandoned as soon as the engagement ends.
Timing and deadlines
Before the process it supports, with enough time to test it. A fundraising model finished the week before a pitch has not been stress-tested, and it will be stress-tested by somebody else.
For project appraisal, before the commitment rather than as a justification afterwards. That distinction matters: a model built to test a decision is a different exercise from one built to support a decision already made, and readers can generally tell which they are looking at.
What you get
- A three-statement model, fully linked and with integrity checks
- Driver-based revenue and cost engines
- A scenario layer with base, downside and case-specific scenarios
- Sensitivity analysis on the variables that actually drive the outcome
- An assumptions book with sources
- Handover and training so you can run and update it
Documents we will ask for
What we ask for up front:
- Historic financial statements, three years where available
- Management accounts to the most recent month
- The commercial drivers and how they behave
- Pricing, cost and capacity data
- Headcount plan with dates and full costs
- Capital expenditure plans
- Existing facilities, covenants and repayment schedules
- Whatever the model is for — the investor deck, the project scope, the bank’s requirements
Fees
Fixed fee, scoped on complexity: the number of revenue streams, entities, scenarios and the level of third-party scrutiny expected. A model for internal planning is a lighter build than one that will be interrogated by an investor.
We quote after understanding what the model is for, because that determines the build more than the size of the business does. Updates and revisions during a live process are quoted separately or covered under a retainer.
Related
Frequently Asked Questions
What is a three-statement financial model?
One where profit and loss, balance sheet and cash flow are fully linked, so a change in any assumption flows through all three. Without the balance sheet there is no way to check internal consistency, which is why its absence is the fastest way to lose a professional reader.
What makes a model credible to investors?
Assumptions in one place with sources and nothing hard-coded, a balance sheet that balances with a visible check, working capital modelled on days, a genuine downside case, and consistent formulas across rows. Readers form a view in the first few minutes and they test for exactly these things.
Can you model a specific decision rather than the whole business?
Yes — a new location, a production line, an acquisition. Project models are usually simpler and more focused, and the important discipline is building it to test the decision rather than to justify one already made. Readers can generally tell the difference.
Will you challenge our assumptions?
Yes, during the build. A revenue ramp assuming a market share nobody has achieved, or receivable days better than the business has ever managed, makes the output arithmetically correct and commercially worthless. Better that we raise it than a counterparty does.
Can we update the model ourselves?
That is how we build it — documented, with inputs separated from calculations, and handed over with training. A model only its builder can operate gets abandoned as soon as the engagement ends.
How long does a model take to build?
A focused project model is one to two weeks. A full three-statement model with scenarios for a fundraising process is longer, and should be finished well before the process starts so it can be stress-tested internally first.
What is sensitivity analysis?
Testing how the outcome changes as each key variable moves, to identify which ones actually drive the result. It is frequently not the variables the owner expects, and that is usually the most useful output of the whole exercise.
A fundraising model, a project appraisal and a planning tool are different builds. Tell us which, and who will be reading it.
Check my compliance status 058 101 9570
Last reviewed 27 July 2026. Rates, thresholds and deadlines change — the e-invoicing provider deadline has already moved once. Confirm current requirements with the Federal Tax Authority before acting, or ask us to check your position.