A 3-year financial model is not a prediction of the future. It is a structured argument that says, "If these assumptions hold, this is the business we build." Investors know your numbers will be wrong. What they are really testing is whether you understand the levers of your own business and whether your logic hangs together. A clean model, built the right way, does that for you. Here is how to structure one in a spreadsheet, tab by tab.
Start with a separate assumptions tab
The single most important rule: every number a human types goes on one dedicated "Assumptions" tab, and every other cell in the model is a formula that references it. Never hard-code a price or a growth rate inside the revenue tab. When an investor asks "what if you charge 20 percent less," you change one cell and the whole model updates.
Group your assumptions clearly:
- Revenue drivers: price per unit or plan, number of customers added per month, monthly churn, average order value.
- Cost drivers: salaries by role, marketing spend, cost of goods, rent, software.
- Timing and structure: which month you launch, headcount hiring plan, and tax treatment.
Build revenue bottoms-up, not tops-down
Avoid the classic trap of "the market is worth 50,000 crore and we will capture 1 percent." That is not a model, it is a wish. Build revenue from your own activity instead. For a SaaS or subscription business: new customers per month, multiplied by price, minus churned customers, gives you a running total of paying users each month. For a marketplace or D2C business: orders per month multiplied by average order value multiplied by your take rate or margin.
Model this monthly for at least year one, then you can summarise years two and three quarterly or annually. Show the numbers before GST. GST that you collect from customers is not your revenue: you are collecting it on the government's behalf and remitting it. Note that GST registration becomes mandatory once turnover crosses roughly 40 lakh for goods or 20 lakh for services in most states, with lower limits in special category states.
Build costs the same disciplined way
Split costs into two buckets. Cost of goods sold (COGS) scales directly with revenue: payment gateway fees, cloud hosting per user, shipping, packaging. Operating expenses are your fixed running costs: salaries (usually your largest line), rent, marketing, legal, and accounting. Tie your salary line to a real hiring plan, one row per role with a start month, so a January hire only costs you from January. Remember that when you pay salaries, contractors, or rent, you are responsible for deducting TDS and depositing it.
Assemble the profit and loss
The P&L pulls it all together in a fixed order:
- Revenue (from your revenue build)
- less COGS = Gross Profit, and gross margin as a percentage
- less Operating Expenses = EBITDA (operating profit)
- less depreciation, interest, and tax = Net Profit
On tax, be realistic: most early startups run losses for the first few years, so the tax line is often nil while losses carry forward. When you do turn profitable, a DPIIT-recognised startup incorporated on or after 1 April 2016 (the window now extends to before 1 April 2030) can claim a 100 percent profit deduction under Section 80-IAC for any three consecutive years within its first ten. Outside that, the common rate for a domestic company opting into Section 115BAA is 22 percent plus surcharge and cess, an effective 25.17 percent.
The outputs investors actually look at
Investors rarely read every row. They jump to a summary of five things:
- Revenue growth: the trajectory across the three years, and whether it is believable.
- Gross margin: does the unit economics work as you scale.
- Monthly burn: how much cash you lose each month.
- Runway: cash in bank divided by burn, which tells them how long the raise lasts.
- Path to profitability: the month the business stops losing money.
Build a clean summary tab showing exactly these. A model that a founder can defend line by line, tracing every output back to one assumption, earns far more trust than a hockey-stick chart with no logic underneath.