Checking a financial model takes four tests. First, compare it with actuals: feed in the inputs for past months and check the results against reality.
Second, technical checks: the balance sheet balances and there are no broken links or numbers hard-coded into formulas. Third, sensitivity analysis.
Fourth, a stress test at extreme values. A model that passes all four can be agreed with management or an investor and then updated monthly against actuals.
This is the fourth and final article in our financial model series, steps 15–20 of 20.
A model becomes a full management tool only after the final stages: checking the calculations, agreeing the inputs and adapting to changes in the project.
Without this work it risks remaining a theoretical exercise that management neither accepts nor uses. The earlier parts of the series:
- Financial model: what it is, what it consists of and how to build one — steps 1–4;
- Sales forecast and resource planning — steps 5–9;
- Financial model calculations: formulas, scenarios and assumptions — steps 10–14.
Step 15. Agree the inputs with the project lead
Building a model is not just calculation — it is reconciling reality with expectations.
The project initiator almost always has an idea of what the model should show, and objective calculations may not match it.
That is normal, but if it is not handled in advance, the model will be rejected and left unused.
So the key inputs and assumptions behind the formulas are agreed before the calculations.
How to agree them
- Fill in and approve the inputs sheet. What prices and sales volumes are assumed? Which costs are the baseline? Which parameters change over time?
- Agree the relationships and assumptions. How are sales and marketing linked? How does unit cost change as output grows? How is revenue growth forecast?
- Compare assumptions with actuals. If the project has history, check how past results compared with forecasts. If not, use benchmarks from the industry.
Why it is critical
- It removes subjective expectations. Initiators often subconsciously bend the numbers towards what they want, and agreeing in advance allows an unbiased view.
- It builds trust in the results. If the manager understands what data the model rests on, they will not reject an unexpected result.
- It lets you choose the best scenario instead of arguing with the model’s results.
A model is a decision tool, and all its inputs must be agreed, transparent and logical.
Step 16. Fill in the template: structure, links, formulas
This is the technical build: structure, links and formulas.
It is not enough to insert formulas — the logic must be transparent, so the model is convenient and clear for every user.
Sheet structure
- Inputs — all assumptions, variables and parameters.
- Calculations — revenue, cost of sales, taxes, investment.
- Outputs — P&L, cash flow, balance sheet.
- Additional sheets — scenarios, sensitivity, charts, checkpoints.
Links and formulas
- All parameters come from a single source: change one variable and every dependent metric updates automatically.
- Numbers typed directly into formulas (hard-coding) are forbidden: they make the model unmanageable.
- Basic formulas: revenue = sales volume × price; cost of sales = production + logistics + storage; taxes — from profit or revenue depending on the tax regime.
How to avoid errors
- Keep formulas simple: unwieldy formulas are hard to check.
- Structure the links so the model does not break when key parameters change.
- Add checks: control totals, balance sheet balancing, logic tests.
If the structure is chaotic and the links unclear, the model loses its value.
Step 17. Testing and validation: checking the calculations are objective
Before using the model or presenting it, make sure it is correct: calculation errors lead to wrong conclusions and wrong decisions. A model is tested in five ways.
1. Comparison with historical data
If the project has been running for at least 3–6 months, check how well the model reflects actuals:
- enter the actual inputs for the past period into the model — prices, volumes, costs, tax rates;
- compare the model’s output with the actual results for the same period;
- assess the gaps. If they are significant, find the cause: errors in formulas or relationships, over- or under-stated coefficients, wrong links between metrics.
A significant deviation is acceptable only if something fundamental changed in the forecast period: a new business model, a new product, a new market.
If nothing changed and the model diverges sharply from actuals, the calculations need revisiting.
2. Errors and logical inconsistencies
- Assets and liabilities balance.
- The difference between cash flow and profit (P&L) is correctly reflected in the balance sheet.
- Control totals and formulas are intact, with no broken links or errors.
3. Sensitivity test
The model must respond sensibly to changes in key parameters:
- What happens if raw material prices rise 10–20%?
- How does profit change if sales come in below plan?
- What is the effect of exchange rate or tax changes?
If the response is unrealistic, the model is not flexible enough or contains wrong relationships.
4. Modelling real situations
Run scenarios: a sharp rise in raw material prices, supply delays, a sudden fall in demand. How does the model react? What corrective mechanisms are in place?
5. Testing “at the limit”
Push the metrics to extreme values to find the project’s critical breaking point and the factors that affect the result most.
Testing is the last checkpoint. A model that passes the historical data check and stress tests can be considered an objective decision tool.
Financial model checklist
| Check | Sign of a problem |
|---|---|
| The balance sheet balances in every period | Assets minus liabilities is not zero |
| Cash does not quietly go negative | Negative cash with no loan or investment in the model |
| No numbers inside formulas | A value does not change when the scenario changes |
| Sales growth is backed by resources | Revenue grows, marketing and staff do not |
| Taxes follow the regime and jurisdiction | VAT counted as revenue, no payroll contributions |
| The model matches past months’ actuals | Large gaps with no change in the business |
| The response to ±10–20% on key drivers is sensible | Profit does not move, or jumps |
Step 18. Agree the results with management or the client
After testing, the model is presented to whoever makes the decisions — a manager, an investor, a client. The model does not always show what people want to see.
If the forecast is less optimistic than expected, the analyst’s job is not to adjust the numbers but to explain the logic and the ways to improve.
How to present the model
- Key conclusions: the investment needed to launch; profit, payback, financial risks; cash cover at each stage.
- Scenarios: how the project looks in the optimistic, realistic and pessimistic cases; the assumptions behind each; the points of no return beyond which the project fails.
- Sensitive factors: which parameters affect the result most; what happens if prices, volumes or taxes change.
- Risks and mitigation: where the model is most vulnerable, what reduces the risks, how the strategy can be adjusted.
How to avoid awkward moments
- Argue your case. If the project does not pay back under the current strategy, explain why and offer solutions.
- Do not distort data for a prettier picture: reality will test the forecast anyway.
- Prepare alternative scenarios in case the client disagrees with the base case.
The model’s job is to show the real limits of what is possible, not to paint the desired picture.
Step 19. Refine the model for real decisions
After the presentation and discussion, the model is almost always adjusted. It must not stay static. What may need revising:
- Inputs and assumptions — market parameters (prices, volumes, purchasing terms), production, logistics, staff and marketing costs, macroeconomics (inflation, exchange rates, taxes).
- Relationship formulas — the link between sales and marketing, capacity and the headcount plan, the split of variable and fixed costs.
- Scenarios and strategy — new options (less investment at launch, a new market), revising the base case on actual data or investor requirements.
- Positioning — a new pricing strategy if profitability is too low, adjusting the audience or range.
- The schedule — moving the launch, funding stages or capacity expansion.
- The plans the model rests on — the strategic business plan, the approach to scaling.
The model as a strategy adaptation tool
After working with the model, a company often revises not just the numbers but the logic of the project:
- it postpones scaling by a year;
- it decides to raise money in stages rather than in one tranche;
- it cuts inefficient cost lines the model has revealed.
Revisions after the first version are almost inevitable, and they may touch the project concept itself.
Step 20. Keep the model updated as the project runs
The model’s work does not end at the presentation or approval.
It is a living tool that is updated regularly, because actuals diverge from the plan: purchases cost more, the launch slips, prices, exchange rates, supply terms and taxes change.
How to update it
- Regular updates: planned figures are replaced with actuals, and the forecast for the following periods is refreshed with the results to date.
- Variance analysis: plan and actual diverged — why? A forecasting error or a changed market? The model helps adjust strategy, not just record deviations.
- Operational and strategic management: a regularly updated model tells you when to scale, when to raise funding, when to cut costs. The project is managed in real time, not blindly.
From practice: with Darvino, the work did not end with the model and investor materials — after launch we set up management accounting, which is exactly step 20: regular plan-versus-actual. If an operating business needs this process, it is a budgeting job.
Frequently asked questions
How do you check a financial model for errors?
Check that the balance sheet balances in every period, cash does not quietly go negative, there are no numbers hard-coded into formulas and taxes follow the right regime. Then feed in past months’ actuals and compare the output with reality.
What is a financial model stress test?
A test at extreme values: a sharp rise in purchase prices, a fall in demand, supply delays. It shows the project’s critical breaking point and how much reserve is needed to get through it.
How often should a financial model be updated?
Compare plan with actuals every month and recalculate the forecast for the following periods. A model that is not updated describes a different business within six months.
What if the model shows the project does not pay back?
Do not adjust the numbers — look for a path: other financing, location, scale or pricing scenarios. Sometimes the right conclusion is to change the project concept or postpone it, and that too is a valuable result of the model.
Series summary: 20 steps of a financial model
We have covered all 20 steps — from the goal and inputs to testing, validation and continuous updating:
- Focus on the goal. A model is not an end in itself but a tool for management decisions.
- Transparency. The clearer the model’s logic, the more management and investors trust it.
- Flexibility and updating. A project is a living process, and the model must adapt to change, providing up-to-date analysis at any moment.
Want to review your own model or a real case? We regularly run financial modelling webinars.
All 20 steps in one file — the guide “Financial Model: 20 Practical Steps” (PDF).
We wish you success in building financial models and, above all, in delivering your projects!
