How to Build Your First Financial Forecast

Start with revenue

Build your revenue forecast first, from the bottom up. What are your revenue streams? For each stream, what are the key drivers, client count, average contract value, conversion rate, churn? Model revenue by building up from those drivers rather than picking a top-line growth rate and working backward. A bottom-up forecast forces you to articulate your assumptions explicitly, which is both more accurate and more defensible.

Revenue is the first and most important line. For a SaaS company, start with current MRR, project new MRR additions by month based on sales pipeline and conversion rates, apply expected churn. For a service company, project based on existing client contracts and expected new business. For a product company, project unit sales by channel and price point.

Be realistic about the timeline. A sales pipeline of $2M that historically converts 25% within 6 months produces $500K of expected revenue, not $2M. First-time forecasters often forecast the gross pipeline value without applying historical conversion and timing. The forecast ends up at 2-3x actual performance and gets dismissed within one quarter.

Then build expenses

Expenses fall into two categories: those tied to revenue (cost of goods sold or cost of revenue) and those that are more fixed (operating expenses like salaries, rent, and software). For COGS, model them as a percentage of revenue, if your delivery cost is 40% of revenue, that should hold as revenue scales unless you have specific reasons to expect it to change. For operating expenses, build from your current run rate and add planned hires and investments explicitly.

Expenses start with what is committed: current payroll, recurring software, rent, insurance, other known monthly costs. Then layer in planned additions, hires by start date, new tools, one-time projects. Then add variable costs tied to revenue, payment processing, shipping, commissions. Each layer produces a more complete expense forecast.

The specific expense categories that matter: payroll (broken out by function), software and SaaS, sales and marketing spend, professional fees, rent and utilities, travel and entertainment, cost of goods sold. A forecast that lumps everything into three categories is too coarse to be useful. A forecast with 40 line items is too detailed to maintain.

The output you need

A complete forecast produces three outputs: a projected P&L (revenue minus expenses, showing projected profit or loss), a projected cash flow statement (when cash actually moves, which differs from the P&L on accrual basis), and a projected balance sheet. The cash flow projection is most critical for operational decisions, it shows you when you might run short of cash even if the business is profitable on paper.

The output of a first forecast should be: monthly P&L projection for the next 12 months, monthly cash flow projection, ending cash balance by month, and key metrics (MRR, headcount, runway) tracked monthly. The cash flow projection is often the most important output because it tells you when (if ever) you run out of money under current plans.

Build sensitivity into the output. What happens to cash if revenue comes in 20% below plan? What happens if you hire on schedule but revenue is delayed by one quarter? These scenarios are the actual value of the forecast. A single-point forecast gives you one answer. A sensitivity analysis gives you the range of outcomes and where the business is fragile.

Making it useful

A forecast is only useful if you compare it to actuals regularly. Set it up so you can see, month by month, where you came in above or below your assumptions. The variance analysis is where the learning happens, understanding why you missed a forecast line is more valuable than the forecast itself.

A forecast is useful when it drives decisions, not when it sits in a file. Monthly actuals should be compared to forecast, variances should be explained, and the forecast should be updated based on what you are learning. A forecast that gets built once and never updated is just a budget, and a stale one at that.

The cadence that works: full forecast rebuild twice a year, monthly updates to reflect actuals, quarterly reviews with the board comparing actual-to-forecast. This keeps the forecast current enough to be useful without creating so much maintenance that it becomes a burden.