A financial model is a set of assumptions wearing a spreadsheet. That is not a criticism — it is the thing to keep in mind every time you build or read one. The formulas are just arithmetic; the actual skill is choosing inputs you can defend, structuring them so someone else can follow your logic, and being honest about which numbers are known and which are a guess.
What changed in 2026
- AI tools draft a first-pass model fast, from a plain description of the business, which is a real time saver — but they default to generic assumptions that need replacing with your actual numbers before anyone should trust the output.
- Scenario ranges became the expected format, not an optional extra. A single-point forecast increasingly reads as unfinished rather than confident.
- More reviewers ask for the assumptions tab first. The output numbers matter less in a review than whether the inputs are labeled, sourced, and reasonable.
The structure of a model that holds up
- Assumptions tab, separated from everything else. Every input — growth rate, price, churn, cost per unit — lives here, labeled, with a note on where it came from.
- Revenue build. Start from a driver, such as users, units, or deals, not a top-down guess about market share.
- Cost structure. Fixed costs, variable costs, and the point at which each cost tier changes, such as a new hire or a new server tier.
- Cash flow, not just profit. A profitable model can still run out of cash if the timing of payments does not match the timing of costs.
- Scenario toggle. Base, upside, and downside cases driven by changing two or three key assumptions, not rebuilding the whole sheet.
Model formats compared
| Type |
What it answers |
Typical use |
| Back-of-envelope |
Is this roughly worth pursuing |
Early filtering of ideas, sanity checks |
| Driver-based operating model |
How does the business perform month to month |
Internal planning, board updates |
| Three-statement model |
Income statement, balance sheet, and cash flow linked together |
Fundraising, lending, detailed investor review |
| Scenario / sensitivity model |
How much do results move if key assumptions change |
Risk assessment, stress-testing a decision |
Common mistakes that undermine trust
- Hardcoded numbers inside formulas. If an assumption cannot be changed in a single cell and have that change flow through the whole model, the model cannot be tested or trusted.
- A single scenario presented as certain. Decision-makers assume you have not considered the downside if you do not show it.
- Precision that outruns the underlying data. A model with figures to the cent, built on assumptions that are honestly good to the nearest 10 percent, misrepresents its own confidence.
- No sanity check against a simpler estimate. If the detailed model and a rough gut-check disagree by an order of magnitude, trust the disagreement, not the spreadsheet.
Pairing a model with the decision it supports
A financial model earns its keep inside a larger document, not on its own. If you are proposing a new initiative, the model is the engine room for a business case: it supplies the cost-benefit numbers the case argues from. Build the model first, then write the narrative around what it shows, rather than the other way around.
FAQ
What software should I use to build a financial model?
A spreadsheet, such as Excel or Google Sheets, for almost every internal use case. Dedicated modeling software is worth it mainly for complex, multi-entity, or heavily regulated forecasting.
How far out should a financial model project?
Three to five years is standard for planning and fundraising. Beyond that, the assumptions carry so much uncertainty that the numbers are closer to narrative than forecast, so say so if you include them.
What is the difference between a budget and a financial model?
A budget is a plan for known, committed spend over a fixed period. A financial model is a tool for testing how the business behaves under different assumptions, including ones that have not happened yet.
How detailed should the assumptions be?
Detailed enough that someone unfamiliar with the business could see exactly where each number came from and challenge it. If you cannot explain an assumption in one sentence, it needs more work before it goes in the model.
Where to go next