How to Create a Financial Model for Your Startup (Step-by-Step)

Open your spreadsheet. Type "Revenue" in A1. Now freeze. What goes in the row beneath it?

That pause is where most startup financial models die — not at the board meeting, not in the investor deck, but at row two of a blank worksheet. I have built four startup financial models from scratch: one for a SaaS tool that raised a seed round, one for a services business that never needed outside money, and two that I rebuilt mid-flight because the first version was structurally wrong. The rebuilds cost me about three weeks each. The original builds took a weekend.

This is the walkthrough I wish I had the first time. What order to build in, which formulas link the three statements, where circularity will bite you, and the assumptions that actually move the needle. No filler.

How to create a financial model for your startup without drowning in tabs

The first model I ever built had eleven tabs. It was beautiful. It was also completely broken, because I had hardcoded a growth rate in three different places and they disagreed with each other. The investor found it in about four minutes.

What I learned from that embarrassment: a working startup financial model needs exactly five tabs, in this order.

• Assumptions — every number you are guessing at lives here, in one place
• Revenue build — driver-based, referencing the assumptions tab only
• Costs and headcount — payroll by role, with start months
• Three statements — P&L, balance sheet, cash flow, linked
• Dashboard — burn, runway, key ratios, pulling from the statements

That is it. If you find yourself adding a sixth tab for "misc," you have a modelling problem, not a spreadsheet problem.

The order matters more than the formulas

Here is the sequence I now follow every time. Getting this wrong cost me a full day of debugging on model number two.

1. Write the assumption list in plain English on paper first: "We sell to X, at Y price, with Z churn."
2. Enter those assumptions as named cells. Named ranges break less than A1 references.
3. Build revenue. Check it against your own gut: does month 12 make sense?
4. Build costs, starting with payroll — usually 60-75% of total spend for a software startup.
5. Link the statements. Start with the P&L, then cash flow, and build the balance sheet last.
6. Add the dashboard ratios only after the statements balance.

One more thing: colour your cells. Blue for hardcoded inputs, black for formulas. It takes ten minutes and saves you hours every time someone else opens the file.

What goes in each statement, concretely

Most guides say "link the three financial statements" and then move on. That sentence hides a full afternoon of work. Here is what it actually means.

The P&L comes first

Revenue minus cost of goods sold gives gross profit. Then operating expenses: payroll, software, rent, marketing. The bottom line is net income. Straightforward, but the trap is mixing capex into expenses. Buying a laptop is not an operating expense — it is an asset that depreciates. Get that wrong and your P&L looks better than reality.

Cash flow is where startups actually die

A profitable P&L means nothing if the cash is gone. Your cash flow statement starts from net income, adds back non-cash items like depreciation, then adjusts for working capital movements. If your customers pay in 60 days, that is two months of revenue you have earned but cannot spend. For a services business invoicing enterprise clients, I have seen this single line swing the runway number by 40%.

The balance sheet is your proof of honesty

Assets equal liabilities plus equity. If that does not hold, you have a broken model. This is the tab everyone skips — and it is the one that catches double-counting. When I rebuilt my second model, the balance sheet refused to balance for two days. The culprit was a payroll cost I had entered both as an expense and as an accrued liability. One duplicate line.

Statement	What it answers	Common build error
P&L	Are we profitable on paper?	Capex booked as expense
Cash flow	Will the bank account survive?	Ignoring payment terms
Balance sheet	Does any of this add up?	Duplicated accruals
Dashboard	How long until we run out?	Ratios computed on annual, not monthly data

Can ChatGPT do financial modelling?

It can write formulas, explain circularity, and draft the structure of a model faster than you can type it. What it cannot do is decide whether your churn assumption is realistic. That judgement is the model.

In practice, I use it for three things: converting a plain-English sentence into a spreadsheet formula, sanity-checking my logic ("does it make sense that working capital increases when revenue grows?"), and generating a first draft of the assumption list. I do not let it produce numbers I then trust. Every figure that ends up in front of an investor has to come from your own reasoning about your own business.

Honest assessment: useful as a sparring partner, unreliable as an author. The moment you outsource the assumptions, you have a model that describes a business nobody is running.

How to create a financial plan for a startup

A financial plan is the model plus the decisions around it. The model is the arithmetic; the plan is what you are going to do about the result.

Three pieces make up the plan, and they are not the same document:

• The operating model — the spreadsheet you just built, updated monthly with actuals against forecast
• The funding plan — how much you need, when, and what milestone it buys
• The scenario set — base case, a slower-growth case, and a case where your biggest customer leaves

The scenario set is the part founders skip. Do not skip it. In my experience, the base case is almost always wrong, and the value of the model is knowing how wrong, and in which direction. When I set up a base case with 8% monthly growth and a downside case at 3%, the downside case is what convinced me to delay a hire by two months. That decision saved roughly €45,000 in committed salary before the next round closed.

What are the 7 stages of startup?

The commonly cited sequence runs: ideation, validation, seed, early growth, expansion, late stage, and exit. Your financial model changes character at each one.

1. Ideation — no model needed, just a rough cost-of-living runway
2. Validation — a one-page unit economics sketch
3. Seed — the full three-statement model, monthly for 36 months
4. Early growth — add cohort retention and CAC payback
5. Expansion — segment revenue by channel and geography
6. Late stage — quarterly rolling forecast, board-level granularity
7. Exit — the model becomes a valuation tool, not an operating one

Most founders over-build at stage three and under-build at stage four. Cohort retention is where the real insight lives, and it is the piece I most often see missing.

Build from drivers, not from percentages

Typing "10% monthly growth" into a revenue cell is the single most common mistake I see. It hides everything that matters. If your growth comes from acquiring customers, then growth is a function of marketing spend divided by acquisition cost. Change the spend and the model should change with it.

So model it like this: customers at start of month, plus new customers acquired (budget ÷ CAC), minus churned customers (start count × churn rate), equals customers at end. Multiply by average revenue per account. That is your revenue line.

Now the model is alive. Cut the marketing budget and revenue falls in a way you can trace. Raise prices and you see the margin shift. That traceability is the whole point.

The churn assumption will make or break you

Here is a number that surprised me: in one of my models, moving monthly churn from 3% to 5% cut the 36-month revenue projection by more than half. Same growth spend, same pricing. Churn compounds in the opposite direction.

Founders consistently underestimate it, because churn is invisible in the early months when every customer feels like a win. Build the downside case first. It will recalibrate how you feel about the base case.

Three mistakes that cost me real time

Circular references. Interest expense depends on debt, debt depends on cash, cash depends on interest. Excel will scream. The fix is a manual toggle — calculate interest on the opening balance, not the closing one. Slightly less precise, infinitely more workable.

Mixing hardcoded values with formulas. My first model had the same growth rate typed in four places. When I changed one, the others stayed. Named ranges and a single assumptions tab solve this permanently.

Too much granularity. I once modelled 47 expense categories. Nobody read past the first twelve. Consolidate to eight or ten lines and put the detail in a supporting tab if you must.

And a fourth, more embarrassing: I built a model for a business I had not yet validated. The spreadsheet was immaculate. The market did not exist. Beautiful arithmetic does not rescue a bad premise.

Should you use a template or build from scratch?

Build from scratch. Once. Then use a template as a cross-check.

The reason is simple: building the model yourself is how you learn which assumptions actually move your business. A template hands you someone else's assumptions dressed in clean formatting, and you will not notice the mismatches until an investor does. Free templates are fine for structure reference — I keep one folder of them and open it roughly once a quarter. But the numbers inside must come from you.

If you want a starting point, most accelerator programmes publish their own model templates openly. Go look at two of them side by side. The differences between them will teach you more than either one alone.

Which brings me to the last thing. A financial model is a set of written-down guesses about a future nobody can see. The spreadsheet is just the container. The discipline of writing those guesses down, in order, with the arithmetic linked so they cannot lie to each other — that is the actual work. And the day your model predicts something you did not expect, and then it happens, is the day the whole exercise stops feeling like homework.