Financial Projections for Startups Before Product Launch

A pre-launch startup has zero revenue history, which means zero proof. A financial projection at this stage functions more as a model of the assumptions that would have to be true for the business to make money, written down so an investor (and the founder) can poke holes in it. Nobody expects a founder to know the future, but they do expect the founder to know their own math.

How to identify the assumptions that drive the whole model

Most of a projection's output comes down to a handful of numbers, and everything else is decoration.

Pin these down first:

• Conversion rate: the share of leads or site visitors who actually pay
• Average deal size or price per unit
• Customer lifetime value (LTV)
• Customer acquisition cost (CAC)

Before launch, none of these come from internal data, obviously, because there isn't any yet. They come from competitor pricing, public case studies, and industry benchmarks. B2B SaaS, for instance, tends to run notably high CAC, so a founder in that space citing a scrappy consumer-app benchmark is already off on the wrong foot. They also come from early customer conversations. Around 20–30 interviews is usually enough to surface real pricing signals and how sticky (or not) switching costs are, and they come from analogous companies at a similar stage, assuming those numbers are public.

Then run the sanity check: LTV:CAC. Investors commonly look for roughly three dollars of lifetime value for every dollar spent acquiring the customer, and if a model can't get close to that ratio, the unit economics conversation needs to happen before the pitch does, not during it.

Every assumption should carry a label and a source: "benchmarked against Company X's published CAC," or "based on 30 customer interviews conducted in March," rather than a vague nod to "industry standard." An investor should be able to point at any number on the page and get a straight answer about where it came from, and if the founder shrugs, that's the whole meeting.

Building the revenue model from the ground up

There are two ways to build the revenue line, and they should both show up in the model, doing different jobs.

Bottom-up starts from what the team will actually execute: sales calls per week, estimated close rate, price per deal. It's the more credible method at an early stage because it's grounded in real capacity, not market size, and it also exposes constraints fast. Can two founders actually close 40 deals a month while also building the product? Usually not, and the model should say so before the spreadsheet pretends otherwise.

Top-down starts from total addressable market and applies a share assumption to land on revenue. It's useful as a ceiling check, a way to ask whether the bottom-up number is even plausible given the size of the market. As a standalone method, though, it's close to fiction: "if we get a small slice of a massive market" is a wish with a percentage sign attached, not a projection.

Build bottom-up first, then check it against top-down. The gap between the two numbers tells the story: either the founder isn't being ambitious enough, or there's no real mechanism behind the growth they're claiming.

The revenue line also needs to match the actual business model:

• Subscription businesses should separate monthly recurring revenue, churn, and expansion revenue
• Transactional businesses should model volume, frequency, and average order value
• Marketplaces need both supply and demand sides modeled, plus the take rate

Year one should look boring: conservative, monthly, unglamorous. The model's job is to survive an argument, not to impress anyone at first glance.

Why CAC almost never stays flat as a startup scales

The single most common error in early financial models: CAC gets treated as one fixed number, forever, no matter how fast the company grows. That almost never holds up.

Here's the mechanism. Early customers come cheap, because they come from founder networks, existing communities, word of mouth, essentially free. But that pool is small and it dries up fast. Once it's gone, the startup has to buy attention in paid channels, where it's now competing against companies with bigger budgets and more patience. And the moment competitors notice a startup gaining traction, they start bidding up the same keywords and the same ad inventory, which pushes acquisition costs up further still.

A model that takes this seriously breaks CAC out by channel instead of blending it into one aggregate number, and assumes different cost curves for different growth phases. It shows, explicitly, when CAC is expected to climb and what would need to be true to keep it in check.

Retention deserves a mention here too, because it's the quiet lever most pre-launch models skip entirely. Every customer that doesn't churn is one less customer that needs replacing, which means one less dollar spent reacquiring someone the company already had. Retention functions as a cost control as much as a satisfaction metric.

One more thing worth saying plainly: pick one or two channels the team will actually run, not five. A slide with five acquisition channels and no prioritization reads as a founder who hasn't made a decision yet.

Mapping the cost structure before a dollar of revenue arrives

Costs split into two buckets, and the model needs to keep them separate or the whole thing gets muddy fast.

Fixed costs don't move with output: salaries, rent, software subscriptions, insurance. Variable costs scale with revenue or usage, things like payment processing fees, hosting costs per user, and cost of goods for anything physical.

Pre-launch, a few categories deserve their own line:

Product development usually eats the biggest chunk of early spend, mostly engineering time, design, and tooling. This is also where MVP scope discipline either saves a company or quietly bleeds it dry. Unnecessary feature development, building the version-two polish before anyone's confirmed version one works, is one of the largest avoidable costs a young startup racks up. Legal and formation costs matter too: entity setup, IP protection, the early contract work nobody enjoys but everybody needs. Early marketing and discovery costs (interviews, landing page tests, small paid experiments) belong in the model as well.

Founder salaries need honest treatment. Most investors understand a founder paying themselves modestly to stay solvent, while projecting a salary of zero tends to read as either unrealistic or a warning sign that the founder hasn't thought through their own survival.

Gross margin deserves its own line even this early. Software businesses tend to run high gross margins, while hardware and marketplace businesses vary a lot, and that variance shapes every number downstream of it. The model should also flag when fixed costs are set to jump, hiring engineer number two, signing a real lease, so an investor can see the founder has thought about scaling costs, not just scaling revenue.

Calculating cash runway and knowing what it tells you

Runway is the number of months a company can keep operating given its current cash and monthly burn. It's arguably the single most consequential number in the whole model, because it determines when the next fundraise has to happen and how much time exists to hit the milestones that justify it. A model showing 18 months of runway is telling an entirely different story than one showing six, even if every other line looks similar.

Building the cash flow projection starts with operating expenses by month, then layers in when revenue actually shows up as cash. Those two things, revenue and cash collection, are often not the same moment; invoicing lag and payment terms mean money earned on paper can sit uncollected for weeks. What's left after that is net cash position, month by month.

Context matters here too: the average seed round in 2025 was $2.2 million. That's a useful benchmark, but founders should model their own burn rate against their own raise, not assume a standard check buys a standard runway. Two companies raising the same amount can have wildly different survival windows depending on headcount and spend.

Good models run three scenarios. A base case where the assumptions hold roughly as expected, a conservative case where revenue comes in slower and CAC runs higher, showing what happens to runway under pressure, and an upside case, which matters less for investors and more for founders deciding when they can afford to hire.

On break-even: most businesses take years, not months, to get there, and the model should show the path and the conditions required to hit it. Claiming break-even by month nine, without a mechanism to back it up, reads as either naive or dishonest, and investors have seen enough decks to tell the difference.

How pre-launch projections connect to fundraising instruments

Most early-stage rounds today run on a SAFE (Simple Agreement for Future Equity), introduced by Y Combinator in 2013 and now the default instrument for pre-seed and seed deals. A SAFE carries no interest and no repayment schedule; it just converts into equity once a priced round happens.

The number that matters in that agreement is the valuation cap, and that cap should trace back to a coherent model of where the business is headed, not a number pulled out of thin air because it sounded competitive. Post-money SAFEs make dilution math visible: how much is being raised, and what slice of the company does that represent once it converts?

The projection tells an investor two things in this context. First, whether the amount being raised actually lines up with the milestones the model says need to happen before the next round. Second, whether the cap makes sense given the trajectory the model lays out.

Worth flagging directly: stacking multiple SAFEs at the same cap can create dilution that's much larger than founders expect once everything converts at once. That math should live inside the model well before Series A, not get discovered by surprise at the cap table meeting. Series A investors, for their part, want evidence of results, not just projections; the pre-launch model becomes the baseline they measure actual performance against later.

What makes a projection credible under investor scrutiny

An investor reading a projection is really looking for three things: named assumptions with real sources, numbers that agree with each other internally, and honest treatment of uncertainty instead of false precision dressed up as confidence.

"Based on benchmark data from three comparable companies" carries weight; a vague nod to "industry standard" does not. Revenue growth that assumes ten new enterprise deals a month better be matched by a hiring plan that includes enough salespeople to actually close ten deals a month. If the CAC number doesn't match the channel budget funding it, that's not a rounding error, that's a tell.

A few patterns reliably signal a weak model:

• A hockey-stick revenue curve with no explanation of what causes the bend
• CAC held flat no matter how aggressive the growth target gets
• A subscription model with no churn assumption anywhere
• Break-even landing suspiciously early, with no mechanism behind it
• Costs that stay flat while revenue is projected to multiply

Every projection should be able to answer one question without hesitation: what would have to be true for these numbers to actually happen? If a founder stumbles on that, the model isn't ready, no matter how clean the spreadsheet looks.

Unit economics carry more weight than the size of the revenue number itself. A defensible LTV:CAC ratio and a believable gross margin tell an early investor more than a big top-line figure ever will, and the model shouldn't sit still once it's built. The first customer interviews, the first pricing test, the first paid channel experiment, all of that is real data now, and it should flow back into the assumptions and change the output.

Founders who check their projections against actual results on some regular cadence, monthly, quarterly, whatever fits, build the habit of actually knowing their numbers cold. That habit is exactly what shows up in a diligence conversation, and it's the difference between a founder reciting a deck from memory and a founder who understands the business they're running.