How to Build a Startup Financial Model, Step by Step

What is a startup financial model for?

A startup financial model is a spreadsheet that writes down how your business works, in a small number of drivers, and shows what those drivers mean for revenue, costs, cash and hiring over the next few years. It is not a prediction. Nobody can forecast a young company's revenue three years out with any precision. What a model can do is show the consequences of decisions you have to make now.

Most early-stage models do four jobs. They set the pace of hiring, which is usually most of a startup's spending. They show runway, the months of cash left at the current burn, and the lowest point ahead. They size and time a fundraise by tying an amount to milestones. And once you have investors, they become the shared language of the board.

Cash is why this matters. CB Insights analysed 431 venture-backed companies that shut down since 2023 in a report published in March 2026. Of the 385 with an identifiable reason, 70% ran out of capital, which the report treats as the final cause rather than the root one, with poor product-market fit at 43%. A model will not fix the root cause, but it tells you how long you have to find it.

What should you decide before you open a spreadsheet?

Write down the three decisions the model must support in the next six months, and who will read it. A model for the founders' own hiring decisions can be rough. A model for a new investor needs clean checks, a sources list for every assumption and a version you are happy to have forwarded.

Then separate what you know from what you are guessing. Known: customers, prices, payroll, supplier invoices. Guessed: conversion from a channel you have never tried, churn beyond the first year. The guesses become your list of things to test, and they tell you which drivers deserve a downside case.

• The decisions the model must answer
• The readers and how polished it must be
• Today's runway: cash divided by last month's net burn
• What you know from your own data
• What you are guessing and how you will test it

What is driver-based modelling?

Driver-based modelling means revenue and costs are never typed in directly. They are calculated from drivers: numbers you can observe, influence and explain. Conversion rate, price per unit, monthly churn, cost per lead and the start month of each hire are drivers. 'Growth' is not a driver; it is what happens when the drivers move.

Start by writing revenue as a formula in words, then break each part down until you reach things you can measure. For a usage-based software company that might be customers times a platform fee, plus usage units times a price per unit, where customers equal last month's customers plus new customers minus churned customers, and new customers equal leads times conversion for each channel.

Build bottom-up from your own activity. Top-down estimates, such as a share of a large market, are useful to check the market is big enough, but they cannot be tested month by month, and experienced investors discount them.

How do you model revenue for different business types?

The structure of the revenue build depends on how you make money. A seat-based subscription template hides the dynamics of a usage-based or marketplace business, so pick the shape that fits and build it month by month.

How do you model costs, hiring and cash?

Payroll is usually the largest cost, so the hiring plan is the heart of the model. List every current and planned person with a start month, a salary and a fully loaded cost that adds employer on-costs such as payroll taxes, pension contributions and benefits. These vary by country, so agree a placeholder rate with your accountant. Tie each hire to a milestone, and mark the hires you would pause if things go badly.

Split the rest of your costs into cost of revenue (what it costs to deliver what you sell, such as hosting, usage-based third-party fees, support and payment fees), operating expenses, and capital spending such as laptops. Build cost of revenue from unit costs rather than assuming a percentage, especially if your product has real costs per use.

Then model cash separately from profit. Customers pay after you invoice them, equipment is cash today but a cost spread over years, and a funding round is cash and capital, never revenue. A simple balance sheet that must balance every month is the best check that all three statements are linked correctly.

How do you calculate runway and know if you are default alive?

Runway is closing cash divided by net burn, the cash going out minus the cash coming in before new funding. Track today's runway and, more importantly, the lowest cash point in the plan and the month it falls. That is often the month before a round arrives, which is also when the round is least certain.

Paul Graham proposed a sharper test in his October 2015 essay 'Default Alive or Default Dead?'. Assume costs stay where they are and revenue keeps growing at its recent rate: do you reach profitability on the money you have left? If yes, you are default alive. If not, you depend on raising money or changing course. Run the test monthly. Be careful early on, when a tiny base can make growth rates look spectacular.

How should you build scenarios?

A scenario changes a handful of drivers that matter, not everything at once. Most early-stage models need three cases: an honest base, an upside and a downside. For the downside, cut new customers, raise churn, trim price or usage, make paid channels more expensive, and delay and shrink the next round. Switch between cases and record lowest cash, runway, ARR and headcount for each.

The most useful part of scenario work is what you agree in advance. Write a downside trigger as a number and a month, for example two consecutive months below a set number of new customers, or no term sheet by a given date. Then write the first actions in order: which hires pause, which spend stops, when you would seek an extension. Decisions made calmly in advance are better than decisions made in a panic.

Which unit economics and SaaS metrics do investors ask about?

Investors in subscription and usage-based businesses ask about a short list. Net revenue retention compares revenue today from customers you had a year ago with what they paid then; above 100% means expansion outweighs churn. CAC payback is the months of gross profit needed to earn back the cost of winning a customer. LTV to CAC compares a customer's lifetime gross profit with that cost, and a ratio around three is a common rule of thumb.

Burn multiple, proposed by David Sacks of Craft Ventures in April 2020, is net burn divided by net new ARR over the same period. He described around 2 as reasonable for an early-stage company and 3 or more as a warning. The Rule of 40, widely credited to investor Brad Feld, adds revenue growth rate and profit margin and looks for at least 40%, mainly for companies past the early stage.

Two habits keep these honest: use fully loaded CAC, including sales salaries and the founders' share of time spent selling, and read trends over quarters, because a single lost customer can swing a small company's monthly churn. If you want to check how well you estimate and forecast before you rely on your own numbers, AssessAll's Estimation & Forecast Calibration assessment is one way to test it.

How do you present the model and keep it alive?

Investors test whether the story is believable; boards test whether the plan is still right. In a ten-minute walkthrough, show the drivers and the three you are least sure of, the hiring plan, the annual summary, the cash and runway, and the downside with its trigger. Know five numbers by heart: ARR, net burn, runway, lowest cash and its month, and what the round buys. A SAFE or priced round adds cash and capital; Y Combinator introduced the SAFE in late 2013 and a post-money version in 2018, and the legal terms belong with your lawyer.

After the raise, compare actuals with plan within ten working days of every month end, decide whether each variance is timing or permanent, and change drivers only for permanent differences. Keep one owner, one master copy and a change log. AI assistants can explain and write formulas and draft scenario narratives, but a person checks every number before it leaves the company. Revenue in the model should follow the same recognition rules as your accounts; IFRS 15, for example, sets a five-step model for revenue from contracts with customers.

The Startup Financial Model Kit from Bodhih Training gives you the workbook, forms, walkthrough script and monthly routine for each step in this guide. If you want to turn better financial judgement into a personal goal, Jobulary explains how a development plan built around specific skills works.