- 1Anchor in actuals. At least 12 months of clean historical data. The model is only as credible as the base.
- 2Project revenue bottom-up by driver. Customers × ARPU, units × price, leads × conversion. Top-down "we'll capture 1% of TAM" never holds up.
- 3Build headcount as the engine of operating expenses. Most costs flow from the headcount plan. Show hiring by role and quarter.
- 4Connect P&L, balance sheet, and cash flow. Three statements that tie. Investors will check this.
- 5Document every assumption. Color-code or tag inputs. The reader should be able to find every assumption in under 30 seconds.
- 6Build base, conservative, and aggressive scenarios. Show how outputs change with key assumptions. This is where models earn trust.
- 7Show unit economics over time. CAC, payback, LTV, gross margin by cohort. Trend matters more than point estimate.
- 8Test against your own numbers. Does the model recreate last year's actuals when you input last year's assumptions? If not, the structure is wrong.
What investors actually look for
Investors evaluate three things about a model: are the assumptions defensible, does the structure link outputs back to assumptions cleanly, and does the result tell a coherent story about the business. Any one of these failing undermines trust in the entire model.
Assumptions are the first thing they test. A revenue forecast that assumes 10x growth without a clear driver is not credible. One that starts from current pipeline conversion rates, typical deal sizes, and planned sales hiring is credible. Specific, connected assumptions win; general optimism loses.
Structure matters because investors want to see how pulling one lever affects everything else. If you change the sales hiring pace, does revenue change appropriately? Does expense? Does cash? A well-linked model flexes consistently. One that requires manual updates in multiple places when assumptions change does not.
The key components
A good financial model has clear sections: assumptions page (all the levers in one place), revenue build (starting from units, not top-down), expense build (by function, tied to headcount and other drivers), summary P&L and cash flow (outputs, not inputs), and scenario toggle (base, upside, downside).
The assumptions page is the foundation. Every input that matters should live here, labeled clearly, with reasonable defaults. "AE productivity: $800K annual quota, 12-month ramp" is a good assumption. "Sales will grow 35%" is not - it is an output pretending to be an input.
The revenue build is where most models succeed or fail. Revenue should be built from unit drivers (customers, transactions, usage) not applied as a growth rate. This lets investors see the mechanics of growth: more customers, higher ACV, better retention. Each driver can be challenged and updated individually.
Common mistakes to avoid
Hardcoded numbers are the biggest trust-killer. If your Q3 revenue is a number typed directly into the cell rather than calculated from assumptions, investors cannot see how it was derived. Every number in the output should be a formula that ties back to assumptions.
Over-optimism in assumptions is the second trust-killer. A model that assumes 0% churn, 150% net retention, and 100% conversion rates is not realistic. Investors have pattern recognition on what benchmarks are defensible. A model outside those benchmarks without strong justification reads as either naive or dishonest.
Scenario analysis that breaks the model is a third problem. When you toggle to downside, some line items should change (lower revenue, slower hiring), others should not (rent, committed contracts). If your downside toggle produces nonsense, investors see that you have not thought through the business mechanics carefully.
How to present it
Walk investors through the model, do not just send it. A 20-minute model walkthrough is often more valuable than a 45-minute pitch deck. Explain the structure, show the key assumptions, demonstrate how it flexes under different scenarios. This builds confidence in the analysis behind the numbers.
Be ready to answer specific questions about the drivers. "What happens to EBITDA if churn is 2 points higher?" is a common question. You should be able to flex the model in real time and show the answer. Fumbling this question is worse than not being asked.
Leave the detailed model in the data room but do not overwhelm with it during initial meetings. A summary version with key outputs and one-page assumption sheet is often enough for first meetings. Investors who want deeper detail will ask for it; those who do not will appreciate the brevity.