9 Min
Startup financial projections fail on their inputs, not their arithmetic. To build them, label every assumption by where it came from. Then build revenue from whatever limits sales, add costs in the steps they really arrive, and track cash on the day it moves. A five-tab template keeps those pieces in order.
This walkthrough gives you the template, three build steps with a worked example, and two mistakes to avoid.
29 October 2026
)
Label every input Observed, Borrowed or Chosen. A projection with no Observed inputs is a plan, not a forecast.
A 2010 study found that preparing projected financial statements led to more overly optimistic sales forecasts, not fewer.
Build revenue from the constraint that runs out first, not from a share of a large market.
Model staff and rent as steps with start dates, never as a percentage of revenue.
Your funding need is set by the lowest point of the cash line, not by the loss on the P&L.
In a 2016 study, the median small business held cash for only 27 days of typical outflows.
Startup financial projections usually go wrong because of hopeful inputs, not broken formulas. Any founder can build a spreadsheet that adds up. The hard part is making it right about the future. Research suggests the act of building a detailed model can make that harder, because detail makes a guess look like a calculation.
Gavin Cassar's 2010 study in the Strategic Management Journal found that nascent entrepreneurs whose ventures reached operation overestimated future sales and employment. The more striking result was about method. Preparing projected financial statements led to more overly optimistic sales forecasts, not fewer.
The spreadsheet is not the problem. The inputs are.
Set it up as five tabs, with data flowing one way. The first tab is an assumption register that lists every input, what it is and where it came from. Revenue drivers and headcount feed a monthly P&L, and the P&L feeds a cash tab that shows your funding need. No number is typed twice.
Each input in the register falls into one of three classes:
Observed = measured in your own business, such as a conversion rate from a pilot.
Borrowed = taken from outside, such as an industry benchmark or a competitor's price.
Chosen = a decision you control, such as your price or your hiring date.
Tab | What it holds | Feeds |
1. Assumptions | Every input, with its class and source | All other tabs |
2. Revenue drivers | Capacity, volume, price, ramp | Monthly P&L |
3. Headcount and costs | Roles, start dates, fixed costs | Monthly P&L |
4. Monthly P&L | Revenue, direct costs, overheads | Cash |
5. Cash | Timing of receipts and payments | The funding need |
Quick test: count the Observed inputs in your model. If the answer is zero, you do not have a forecast yet. You have choices and borrowings dressed as one, and a reader will ask which numbers you have actually seen.
Need a linked balance sheet too? See the 3-statement financial model guide.
Build in three steps once the register exists. Set revenue from the constraint that limits sales. Add costs as steps with start dates. Then move every figure to the month the cash actually changes hands. The example below follows one small business through all three, so you can check each number.
Revenue is capped by whatever runs out first: sales capacity, delivery hours, seats, stock or marketing cash. Identify that constraint and build volume from it, not from a share of a large market.
Monthly revenue ceiling = Capacity x Price
Illustrative example: a bookkeeping firm has two bookkeepers, and each can serve 25 clients. At $300 a month per client, revenue cannot pass $15,000 a month until a third bookkeeper starts. No market size changes that.
What to model: the constraint, its monthly limit, and how it grows.
Why it works: the forecast cannot outrun the business's ability to deliver.
Bottom-up market sizing then answers a narrower question: is there enough demand to fill the capacity you planned?
Many costs arrive in whole units. You hire a person, sign a lease or buy a machine, and the cost jumps, then stays flat until the next jump. Modeling these as a percentage of revenue makes them shrink when sales disappoint, which is exactly when they do not.
In the example, two bookkeepers at $4,500 a month cost $9,000, which is 60% of $15,000 in revenue. If revenue drops to $10,000, a percentage model shows payroll of $6,000. The real payroll is still $9,000, so the model hides a $3,000 monthly loss.
What to model: each role and fixed cost with a start month.
Why it works: it exposes the months where cost runs ahead of revenue.
Price each step line by line with a startup cost calculation before it enters the headcount tab.
Profit is recorded when a sale is earned. Cash flow reflects when the customer pays. Customers who pay in 30 or 60 days open a gap between the two lines. So do suppliers who want payment upfront and annual licenses paid in January.
In the example, clients pay 30 days after each invoice. January's $15,000 of revenue arrives in February, while January's $9,000 payroll leaves on time.
What to model: payment terms in, payment terms out, one-off annual bills.
Why it works: the funding need is set by the cash low point, not by the loss.
The margin for error is thin. The JPMorgan Chase Institute's 2016 study of 597,000 small businesses found the median held enough cash for 27 days of typical outflows. A quarter held fewer than 13 days. One late invoice cycle can use that up.
Most projections that look finished fail in one of two ways. Costs are entered as a percentage of revenue, so losses vanish in bad months. Or the model has a single future, the one the founder hopes for. Both are quick to build and easy to believe, which is why a careful reader checks for them first.
Staff, rent and software are entered as shares of revenue because it is quick. The model then stays profitable at any size. The real business cannot fire half an employee. The fix: give every whole-unit cost a start date in the headcount tab. Keep percentages for costs that scale with each sale, such as card fees.
The model has one case, and it is the case the founder hopes for. The fix: run a downside case that moves only the Borrowed inputs to less favorable values. Run a second that starts revenue three months later while costs keep their dates. If the business survives both, the plan has earned its valuation. If not, the downside belongs in the risk section.
Problem | Impact |
No source column | Nobody can tell evidence from hope |
Costs as a percentage of sales | Losses hidden in the bad months |
Revenue from market share | Volumes the business cannot deliver |
Profit modeled, cash ignored | Funding need understated |
One scenario | Optimism mistaken for precision |
Startup financial projections are a register of beliefs about the future, arranged so a reader can check them. The spreadsheet is the easy part. The discipline is labeling every input honestly and building revenue from what limits it. Then cost people and premises in steps, and follow cash on the day it moves.
Done in that order, the model shows which three or four inputs decide whether the business works. Test those first in the market. For what the finished numbers must show a lender, see financial projections for a business plan. The mechanics of linking the tabs are in building a startup financial model.
PrometAI's business plan generator builds the financial projections and the plan together. You start from a structured model and spend your time on the assumptions.
)
)
)
)
)
)
)
)
)
)