Financial Forecasting for Startups

Most people project growth linearly, assuming that if they saved $5,000 this year they will save $5,000 next year. Successful financial trajectories, whether in careers, investments, or business revenue, rarely follow a straight line. They follow a geometric progression. To measure it, analysts use the Compound Annual Growth Rate (CAGR), which smooths the volatility of individual years to reveal the underlying rate required to move from a starting point to an ending point over a period.

Formula. CAGR = (Ending Value / Beginning Value) ^ (1 / n) - 1 where n is the number of years

Worked example: A career trajectory

An entry-level employee earning $40,000 aims to earn $200,000 within 10 years. They are not looking for linear raises. They are solving for a CAGR of about 17.5%: ($200,000 / $40,000) to the 1/10 power is 5 to the 0.1 power, which is about 1.175, minus 1. Consistent annual raises of roughly 17.5% are required to hit the target.

Understanding CAGR changes how you view the early years of any growth curve. On a compound curve, the absolute dollar growth in the early years is small even when the percentage growth is high. The vertical acceleration in later years is not a change in performance; it is the same rate applied to a larger base. This is true for salaries, for startup revenue, and for investment returns.

The measurement window matters. A standard TTM period works well for stable or steadily growing businesses but can mislead after an inflection. If the company lost its largest customer five months ago, TTM still includes seven months of that revenue, flattering the run rate. If it signed a major contract two months ago, TTM includes ten months before that contract existed, understating the trajectory. Diligence teams run multiple views side by side.

The best analysis does not ask which number is correct. It asks what each number tells us. TTM shows where the business has been, run-rate where it appears to be today, and NTM where management believes it is going. The spread between them often tells you more about forecast risk than any single metric. If NTM is three times TTM and run-rate depends on a recent spike, investors are being asked to underwrite an inflection that has not yet been proven.

Building a model is more art than science, and there is rarely one right answer. The following guidelines, adapted from Lishego (2020) as practitioner consensus rather than empirical findings, describe what investors and CFOs expect a well-built model to look like.

1. Build your model in Excel. Spreadsheets let anyone audit the logic, trace formulas, and modify assumptions. Excel is the professional standard for investor-facing models.
2. Build a cash flow model. Revenue does not equal cash. A company can be profitable on paper and run out of money. The income statement shows viability; the cash flow model shows survival.
3. Look ahead 3 to 5 years. Seed investors typically want three years, Series A and beyond up to five. The further out, the less precise, but the exercise forces you to think about the trajectory.
4. Build monthly, summarize annually. Monthly granularity is essential for managing cash, tracking variance, and spotting seasonality. Annual summaries communicate the big picture.
5. Build from the bottom up. Top-down 'we will capture X% of the market' assumptions are not operating plans. Build from the activities that generate revenue: reps, units per rep, price per unit.
6. Code and label assumptions dynamically. Every input belongs in a dedicated assumptions section, not hard-coded into formulas, so one change flows through the whole model.
7. Document assumptions with sources. Note where each input came from: customer conversations, market research, benchmarks, or prior operating data. This provides credibility.
8. Balance detail and simplicity. A model with 200 line items is not better than one with 40 if the detail does not change the conclusions. Model what is material; aggregate the rest.
9. Focus on expenses. Founders obsess over revenue, but expenses are more predictable and controllable. Precise expenses with conservative revenue is more credible than the reverse.
10. Build the base case from evidence. The base case is the most evidence-based operating plan, neither optimistic nor pessimistic. Then build a downside case and reserve the upside for favorable execution.
11. Make it user-friendly. Color-code inputs versus calculations, format consistently, and label every row. The model will be read by people who did not build it.

For subscription businesses, revenue depends not only on new subscribers each month but on the cumulative base retained. Lishego (2020) recommends modeling Monthly Recurring Revenue: ending MRR equals beginning MRR (prior month's ending), plus new MRR (new subscribers times price), minus churn MRR (monthly churn rate times beginning MRR).

Formula. Ending MRR = Beginning MRR + New MRR - Churn MRR

Worked example: a three-month MRR build

Assumptions: $2,250 price per month, 2 new units per month ($4,500 new MRR), and 2.0% monthly churn applied to beginning MRR.

The churn assumption is the most consequential number in a subscription model. A company with 2% monthly churn loses roughly 22% of its base annually; at 5% it loses over 45%. Small changes in churn compound dramatically over a multi-year forecast, which is why investors spend disproportionate time on retention.

Market size should not be the primary revenue driver in a startup forecast, but it serves as an external constraint and reasonableness check. A top-down 'we will achieve X% market share by year Y' is not an operating plan: it does not help you hire, does not tell you when to raise, and does not reveal which assumptions are fragile. TAM, SAM, and SOM still matter for framing venture scale and evaluating go-to-market credibility. Market size belongs in the pitch and the strategic framing, not in the cells of the revenue model. Use it as a sanity check: if projected revenue reaches an unrealistic share of the market, revisit the assumptions.

The driver-based approach applies equally to cost of goods sold, operating expenses, and headcount. Model COGS as a percentage of revenue (the gross margin assumption) or from component costs like hosting, support salaries, and payment fees; at an 80% gross margin, COGS equals revenue times 20%. Model headcount by function with start dates and fully loaded costs, the same logic as the sales team model. The key principle is consistency: if you model revenue from the bottom up, model expenses from the bottom up. A model that projects revenue from a detailed sales plan but plugs SG&A as a flat monthly number signals that the founder has not thought carefully about the cost structure.

The model should include a minimum cash balance, the floor beneath which operating becomes risky, commonly 3 to 4 months of gross burn. The financing need then tells the founder how much to raise. It is among the most important outputs for a company that is not yet cash-flow positive.

Formula. Financing Need = Cumulative Net Burn Through Target Milestone + Minimum Cash Balance - Cash on Hand

WORKED EXAMPLE: SIZING THE RAISE

Suppose cumulative net burn to the next milestone is $1.8 million, the minimum cash balance is $0.45 million (about 3 months of gross burn), and cash on hand is $0.8 million. The financing need is $1.8 million + $0.45 million - $0.8 million = $1.45 million. That is the amount to raise to reach the milestone with a safe cash floor intact.

A forecast that tracks cash without tracking profit, or profit without tracking cash, tends to mislead. The three financial statements articulate, meaning they lock together, and each one measures something the others do not.

• The income statement records accrual profit for the period. Revenue is booked when it is earned and expenses when they are incurred, regardless of when cash actually moves.
• Net income does not disappear at period end. It closes into retained earnings on the balance sheet, which is how a profitable period adds to equity and a loss erodes it.
• The cash flow statement reconciles accrual net income to the actual change in cash. It adds back non-cash charges such as depreciation, adjusts for changes in working capital (receivables, inventory, and payables), and layers in the investing and financing flows.
• The balance sheet must balance because every flow lands in two places. A dollar of depreciation, for instance, reduces net income and reduces the carrying value of the asset at the same time.

For a founder this is not bookkeeping trivia. Profit on the income statement is not cash in the bank. A company can report positive net income while its cash falls, because receivables are growing, inventory is building, or capital is being spent. Only the linked model, with all three statements wired together, shows when the company actually runs out of money.

A single-period example shows the wiring. Every number below flows into the next statement, and the balance sheet check confirms the model is internally consistent.

Income statement for the period. Revenue 1,000; COGS 400; gross profit 600; operating expenses 500 (including 100 of depreciation); EBIT 100. With no interest or tax, net income is 100.

Cash flow statement (indirect method). Start from net income 100. Add back depreciation 100. Subtract the increase in accounts receivable 50. Subtract the increase in inventory 30. Add the increase in accounts payable 20. Cash from operations is 140. Investing: capital expenditure (200), so cash from investing is (200). Financing: equity raised 300, so cash from financing is 300.

Formula. Net Change in Cash = 140 - 200 + 300 = 240

Balance sheet articulation (period change). On the asset side: cash +240; accounts receivable +50; inventory +30; property, plant and equipment +100 (that is 200 of capex minus 100 of depreciation). Total asset change is +420. On the other side: accounts payable +20; paid-in capital +300; retained earnings +100 (equal to net income). Total liabilities-plus-equity change is +420. The two sides agree, which is the built-in check that the model is internally consistent.

The takeaway: the company earned 100 of accrual profit yet its cash rose by 240 only because it raised 300 of equity. Operations plus investing alone drained cash, since 140 - 200 is negative 60.

A model is not an academic exercise. For a founder, it exists to answer specific questions. A forecast that leaves these five unanswered is generally incomplete.

• How much cash do we need? The forecast should produce a clear financing need from the cash flow model, not the income statement.
• How long is our runway? Given current cash and net burn, how many months before the company runs out? This is among the most important numbers in early-stage finance.
• What milestones can we reach before the next raise? Investors generally fund milestones, not time. Show what the company achieves within the current cash window.
• What assumptions need to hold for this business to work? Make conversion, churn, contract value, and sales cycle explicit.
• Which drivers create or destroy enterprise value? Some assumptions move valuation dramatically; the model should reveal which inputs have the highest leverage.

Customer acquisition cost measures the total cost of acquiring one new customer, including marketing, sales salaries and commissions, and tools. It is most useful segmented by channel and customer type, because blended CAC can hide one efficient channel and one that is hemorrhaging money. Lifetime value estimates the total value a customer generates over the relationship.

Formula. CAC = Total Sales and Marketing Spend / New Customers Acquired

Formula. LTV = ARPU per Month / Monthly Churn Rate

Formula. ARPU (Average Revenue Per User) is calculated by dividing your total revenue by the total number of active users over a specific period (such as a month or a year).

A conservative LTV uses gross profit rather than revenue: (ARPU times gross margin) divided by monthly churn. The LTV to CAC ratio is among the most-watched metrics in venture capital. A ratio of 3 to 1 or higher is the heuristic for healthy economics, meaning at least $3 of lifetime value for every $1 of acquisition cost. Below 1 to 1, the company pays more to acquire customers than it earns from them. Treat these thresholds as conventions, not measured constants.

LTV to CAC tells you whether the economics are profitable; CAC payback tells you how long it takes to recover the investment. A 12-month payback means funding a year of acquisition cost before the customer turns profitable, which directly affects cash and runway. For most SaaS businesses, a payback under 18 months is the benchmark.

Formula. CAC Payback (months) = CAC / (Monthly Revenue per Customer x Gross Margin %)

Churn is the most powerful lever in a subscription business. Logo churn is the percentage of customers who cancel; revenue churn is the percentage of MRR lost to cancellations and downgrades, which can differ when customers have different contract values. Net revenue retention is the percentage of revenue kept from existing customers after churn, downgrades, and expansion. Above 100% means the company grows even without acquiring new customers, the most efficient form of growth because it requires no incremental acquisition cost. Top SaaS companies reach 110 to 130%.

Formula. NRR = (Beginning MRR + Expansion MRR - Churn MRR - Downgrade MRR) / Beginning MRR

A forecast represents a single path, the base case. Since we are predicting the future, drawing conclusions from one scenario tends to invite surprises, and the longer the horizon, the more likely they become. Professional models use two stress-testing techniques before committing to a plan.

Sensitivity analysis isolates a single variable to measure its impact: if only monthly churn rises from 2% to 4%, what happens to Year 3 ending MRR? If ramp extends from 90 to 120 days, what happens to quarterly revenue? The purpose is to identify single points of failure and reveal which assumptions matter most. If a 1% change in churn breaks the model, churn is the critical variable and warrants the most rigorous forecasting.

Scenario analysis changes multiple variables at once to simulate a distinct narrative. The standard set is a bull case (high growth, favorable conditions, revenue above plan and expenses below), a base case (the most evidence-based operating plan), and a bear case (contraction: sales cycles lengthen, churn rises, a key hire fails, and costs run ahead of plan).

Strong analysts provide both. The most robust plans survive the bear case while positioning for the bull. For a founder, the bear case answers the question most investors ask silently: what happens if this does not work?

Traditional budgeting sets a static annual plan, then measures against a fixed baseline that reality rarely respects. A rolling forecast continuously extends the horizon: rather than forecasting the next calendar year, you forecast the next 12 months on a rolling basis, regardless of where you are in the fiscal year. Every month, drop the completed month and add a new one to the end.

Rolling forecasts are most valuable in high-volatility and growth-stage settings, during major transitions, and anywhere a twelve-month-old assumption would be dangerously stale. For startups, a monthly rolling forecast should be the default, so assumptions stay fresh and course corrections happen in real time rather than during an annual planning cycle.

A model is a hypothesis; reality is the test. Variance analysis, often called Forecast versus Actuals, compares the two. When a target is missed, the variance must be categorized to determine the response. A timing variance means the result happened in a different period than expected, such as a March payment arriving in April; no strategic change is required, only a revised timing assumption. A permanent variance means the result did not happen or the assumption was wrong, such as a cancelled contract; immediate adjustment is required.

Formula. Variance = Actual Result - Forecasted Result

The job of the finance expert is to understand the drivers of each variance, decide whether it is timing or permanent, and update the model accordingly. This feedback loop transforms a static plan into a living management tool.

Forecasting is a discipline for preparedness, not a claim of accuracy. The tools in this module structure the guesswork; they do not eliminate it.

The models assume the drivers are the right ones and the inputs are honest. For early-stage companies with little history, both are uncertain, so every output inherits that uncertainty.

The benchmarks used here, including 3:1 LTV to CAC, sub-18-month CAC payback, net revenue retention of 110 to 130%, 18 to 24 months of runway, a 1.25 to 1.4 times fully loaded cost, and a 60 to 90 day ramp, are practitioner heuristics that vary by sector, stage, and cycle.

The eleven guidelines, the sales and MRR models, and the MARCS workflow are practitioner frameworks. Lishego's guide is a self-published resource by an investment associate, useful as consensus practice rather than peer-reviewed evidence.

This module builds and stress-tests the forecast. Discounting these cash flows into a valuation is the work of Week 5, using the Gordon Growth Model.