Most startup financial models are fiction dressed up as spreadsheets, and investors can spot the difference in about thirty seconds.
If you want to know how to build a startup financial model that survives real scrutiny, start by throwing out the one you probably already have open. The typical first draft works top-down: pick a market size, assume you'll capture 1% of it, back into a revenue number, then build a hockey stick around it. Every VC associate has seen this exact model hundreds of times, usually with the same "$50 billion TAM, we only need 1%" logic sitting on slide four. It tells an investor nothing except that you haven't done the harder work yet.
The harder work is bottom-up. Instead of starting with the market and working down to your revenue, you start with the actual mechanics of how a dollar enters your business and work up. How many salespeople can you hire this year, and how many deals can each one realistically close a month, based on sales cycles you've already observed. How many visitors hit your signup page, what fraction convert on your current funnel, not an aspirational one. What you charge today, not what you hope to charge once you've added enterprise features nobody has asked for yet. A bottom-up model is built from units you can defend individually: seats, transactions, subscriptions, average order value. Add those units up and the total revenue number falls out the other end. It's slower to build and far less impressive on a first pass, and that's exactly why it works.
An investor doing diligence isn't reading your revenue line. They're clicking into your assumptions tab and pulling on the threads one at a time. If your model says you'll close 40 enterprise deals next quarter, they want to know your current pipeline, your close rate on the deals you've run so far, and how many reps you have to run them. A top-down model has no threads to pull on, because there's nothing underneath the top-line number except a percentage of somebody else's market research. A bottom-up model gives the investor something to argue with, and paradoxically, that's what makes it credible. You're not asking them to trust a conclusion. You're showing them the inputs and letting them check your arithmetic.
Y Combinator's own startup library, which has walked thousands of founders through fundraising since 2005, tells companies applying for funding to build models around unit economics rather than market share, precisely because YC partners have sat across the table from too many founders who couldn't answer a single follow-up question about their own spreadsheet. Bessemer Venture Partners, which tracks SaaS company benchmarks across its portfolio and publishes them annually as the Cloud 100 index, built its entire underwriting approach on the same logic: net revenue retention, CAC payback period, and magic number, not top-line growth in isolation. Those are bottom-up metrics. None of them can be reverse-engineered from a TAM slide.
The three numbers that actually carry a startup financial model template
Strip away the formatting and almost every credible model rests on three numbers. Customer acquisition cost, the fully loaded cost of turning a stranger into a paying customer, including the sales rep's salary and the ad spend, not just the ad spend alone. Lifetime value, what that customer actually pays you before they churn, calculated from your real churn rate rather than an assumed one. And payback period, how many months it takes to earn back what you spent acquiring them. Get these three right and the rest of the model, headcount, burn, runway, is mostly arithmetic. Get them wrong and no amount of formatting will save the pitch.
Founders love to inflate the middle number. LTV is where the fantasy creeps in, because it depends on a churn assumption nobody can immediately check, unlike CAC, which is sitting right there in the ad account. If your actual monthly churn is 4% and you model 1% because "we'll fix retention," an investor who has seen a hundred of these decks will ask you to defend that 1% before they ask about anything else. Use your real number. If it's ugly, say so and show the plan to fix it. That's a more fundable conversation than a clean number nobody believes.
Building startup revenue projections without inventing a market
Revenue projections built for investors need to be defensible twelve months out and directionally honest three years out, and those are different jobs. For year one, use your actual sales capacity: reps hired, ramp time, quota, current conversion rates. For years two and three, growth rates matter more than precision, because nobody, including the best-funded companies in the world, has ever hit a three-year forecast exactly. What investors are checking in the out years isn't accuracy. It's whether your growth rate compounds off a believable base or off a wish.
Airbnb's own early model, which Brian Chesky has described publicly in interviews about the company's fundraising history, reportedly started from a simple bottoms-up calculation: how many listings they could realistically add per city per month, multiplied by average nightly rate and occupancy, city by city, rather than a single global TAM number. That's the pattern worth copying. Build the projection market by market, or channel by channel, or cohort by cohort, and let the total emerge from pieces small enough that you can defend each one on its own.
What belongs in a financial model for a pitch deck versus the model itself
The pitch deck gets a summary. The model stays in the data room. Don't try to cram five tabs of assumptions onto a slide. What belongs in the deck is the headline trajectory, the unit economics, and maybe one chart showing how CAC payback has trended over the last few quarters. What belongs in the model, the one investors ask to see after the meeting, is every assumption laid bare in its own labeled cell, not buried inside a formula three columns over. A model where an investor has to hunt for where an assumption lives is a model that looks like it's hiding something, whether it is or not.
Build it in a plain spreadsheet, Google Sheets or Excel, not a slick tool that abstracts the logic away behind a nice interface. Investors want to open the file, trace a formula back to its source, and change one input to see how the whole model responds. If your model breaks when someone changes a growth assumption, or if half the numbers are hardcoded instead of calculated, that's the first thing a sharp analyst finds, and it costs you more credibility than a modest forecast ever would.
None of this guarantees a term sheet. A bottom-up model doesn't fix a weak business, and no amount of spreadsheet discipline invents customers who don't exist yet. What it does is remove the easiest reason an investor has to walk away: the sense that the founder either doesn't understand their own numbers or is hoping nobody checks. Show your work, use your real numbers even when they're unflattering, and build the model so someone else can poke at it without it falling apart. That's the whole test.
Also read: How to Price a SaaS Product With No Competitors to Copy • How to Value a Pre-Revenue Startup When There Is No Revenue • How to Build a Personal Finance Dashboard With AI Before You Talk to a Wealth Advisor
