How to Build a DCF Model Step by Step: The Definitive Guide

What a DCF Model Actually Does (And Why Most Are Wrong)

A discounted cash flow model answers a deceptively simple question: what is the present value of all the cash this business will generate in the future? The logic is clean. A dollar received five years from now is worth less than a dollar received today, because today's dollar can be invested and compounded. The DCF quantifies that time-value relationship by discounting projected free cash flows at a rate that reflects the risk of actually receiving them.

In theory, a DCF is the purest form of intrinsic valuation. It does not depend on how the market prices comparable companies. It does not rely on historical multiples that may reflect bubbles or panics. It stands on its own — a direct link between the cash-generating capacity of a business and its fair value. Warren Buffett has described this approach as the only intellectually honest method of valuation, and on that narrow point, we agree.

In practice, however, most DCF models are exercises in circular reasoning. The analyst who believes a stock is undervalued builds a DCF that confirms the stock is undervalued, because the assumptions are calibrated — consciously or not — to produce the desired output. A 2023 study by NYU Stern's Aswath Damodaran found that the median DCF among equity research reports implied a terminal growth rate of 3.2%, which exceeded the long-run nominal GDP growth rate of the United States. In other words, more than half of published DCF models implicitly assume the company they are valuing will grow faster than the entire economy in perpetuity. That is not analysis. That is advocacy.

The antidote is discipline. A well-constructed DCF forces the analyst to make every assumption explicit and testable. Revenue growth cannot hide behind "management is optimistic." Margin expansion must be tied to specific operational drivers. The discount rate must be derived from observable market data, not pulled from thin air. When built with rigor, the DCF is the most powerful analytical framework available for equity valuation. This guide walks through every step. For how AI is enhancing each of these steps computationally, see our deep dive on AI-powered valuation models.

Step 1: Project Revenue and Build the Income Statement

Revenue is the foundation. Every line item below it — cost of goods sold, operating expenses, depreciation, interest, taxes — flows from the revenue assumption either directly or as a percentage relationship. Get revenue wrong and the entire model is wrong, regardless of how elegant the WACC calculation is.

Top-Down vs. Bottom-Up Revenue Forecasting

Top-down forecasting starts with the total addressable market (TAM) and applies a market share assumption. If the global cloud infrastructure market is $300 billion and you assume Microsoft Azure captures 23% share, Azure revenue is $69 billion. This approach is useful for framing the upper bound of opportunity but dangerous when used as the primary forecast driver, because TAM estimates are notoriously inflated and market share assumptions are often aspirational rather than evidence-based.

Bottom-up forecasting builds revenue from its operational components. For a SaaS company like Salesforce, that means starting with the number of paying customers, average revenue per customer (ARPU), net revenue retention rate, and new customer acquisition. For a retailer like Costco, it means store count times average revenue per store, adjusted for same-store sales growth and new store openings. Bottom-up is harder — it requires understanding the business at the unit-economics level — but it produces assumptions that are individually testable and defensible.

We believe the best DCF models use both approaches as cross-checks. If the bottom-up forecast implies a market share that is historically unprecedented for the industry, the revenue assumption is probably too aggressive. If the top-down approach suggests the company is capturing less share than its competitive position warrants, the bottom-up assumptions may be too conservative. The tension between the two methods is analytically productive.

Margin Assumptions and the Cost Structure

Once revenue is projected, build the income statement line by line. Gross margin should be projected based on the company's historical trajectory, product mix shifts, and identifiable cost pressures or efficiencies. Apple's gross margin expanded from 38.2% in fiscal 2020 to 46.6% in fiscal 2024, driven almost entirely by the mix shift toward higher-margin Services revenue (which now exceeds $96 billion annually at an estimated 70%+ gross margin). A DCF for Apple that uses a flat gross margin assumption misses the single most important structural trend in the business.

Operating expenses should be modeled as a percentage of revenue, but with attention to operating leverage. For businesses with high fixed costs and low marginal costs — software, platforms, content — operating expense as a percentage of revenue should decline as revenue scales. For businesses with proportional cost structures — retail, logistics, consulting — the ratio is more stable. Model R&D, sales & marketing, and G&A separately. Each has different dynamics and different long-term trajectories.

Step 2: Calculate Unlevered Free Cash Flow

Unlevered free cash flow (UFCF), also called free cash flow to the firm (FCFF), represents the cash generated by the business's operations that is available to all capital providers — both debt holders and equity holders — before any financing decisions. This is what you discount in a DCF-to-enterprise-value model.

The UFCF Formula

UFCF = EBIT x (1 – Tax Rate) + Depreciation & Amortization – Capital Expenditures – Change in Net Working Capital. Each component requires careful treatment.

EBIT x (1 – Tax Rate) gives you the after-tax operating income that the business would generate if it had no debt. This is called NOPAT (net operating profit after taxes). Use the company's effective tax rate as a starting point, but adjust for any known changes. The 2017 Tax Cuts and Jobs Act dropped the federal corporate rate to 21%, but many companies pay lower effective rates due to R&D credits, foreign tax holidays, and deferred tax assets. Meta Platforms paid an effective rate of approximately 17.6% in 2024. Alphabet paid roughly 13.9%. These differences are not trivial — a 5-percentage-point difference in the tax rate can shift the DCF output by 8–12%.

Depreciation & Amortization is added back because it is a non-cash expense. However, do not confuse D&A with maintenance capex. D&A reflects the accounting allocation of past capital spending, while the capex line captures actual current investment. These two numbers should be roughly similar for a mature company in steady state, but they can diverge significantly for companies ramping up or winding down investment cycles. Amazon's D&A was $48.7 billion in 2024 while capex was $83.2 billion — the gap reflects the massive data center buildout that will generate depreciation charges for years to come.

Capital Expenditures must be projected carefully. Separate maintenance capex (required to sustain current operations) from growth capex (investment in new capacity, products, or markets). Maintenance capex is a true cash cost of doing business; growth capex is a discretionary investment that should generate incremental returns. Unfortunately, companies rarely disclose this split. A useful heuristic: maintenance capex is approximately equal to D&A for a mature business. Growth capex is the excess.

Change in Net Working Capital is the most commonly neglected component. As a company grows, it typically needs more working capital — more inventory, more receivables, offset by higher payables. This investment in working capital is a real cash outflow that reduces free cash flow. For capital-light businesses like software companies, working capital changes are small. For retailers, manufacturers, and distributors, they can be enormous. Walmart's inventory alone was $56.4 billion at the end of fiscal 2025. A 5% increase in inventory to support revenue growth represents a $2.8 billion cash outflow that goes nowhere on the income statement. For a complete framework on reading balance sheets to identify working capital dynamics, see our balance sheet analysis guide.

Step 3: Estimate WACC — The Discount Rate That Makes or Breaks the Model

The weighted average cost of capital blends the cost of equity and the after-tax cost of debt, weighted by their proportions in the company's capital structure. It represents the minimum return the company must earn on its invested capital to satisfy both shareholders and creditors. It is also the rate at which you discount future free cash flows to present value.

Cost of Equity via CAPM

The Capital Asset Pricing Model calculates the cost of equity as: Risk-Free Rate + Beta x Equity Risk Premium. As of early 2026, the 10-year U.S. Treasury yield is approximately 4.2%, which serves as the risk-free rate. The equity risk premium (ERP) is the subject of perpetual academic debate. Damodaran's implied ERP estimate for January 2026 is approximately 4.6%, though many practitioners use a historical average of 5.0–6.0%. Beta measures the stock's sensitivity to market movements. Use a 2-year weekly or 5-year monthly regression against the S&P 500, but be aware that beta is unstable — it can vary by 0.2–0.5 depending on the measurement window. For a company with a beta of 1.1 and using the implied ERP of 4.6%, the cost of equity is 4.2% + 1.1 x 4.6% = 9.26%.

Cost of Debt and the Tax Shield

The pre-tax cost of debt is the yield the company pays on its outstanding borrowings. For investment-grade issuers like Johnson & Johnson (rated AAA), this might be 4.5–5.0% currently. For high-yield issuers like Carnival Corporation (rated B+), it could exceed 7.5%. The after-tax cost of debt is the pre-tax rate multiplied by (1 – tax rate), because interest payments are tax-deductible. At a 21% federal rate, J&J's after-tax cost of debt is roughly 3.6–4.0%.

Capital Structure Weights

Use market values, not book values, for the weighting. The market value of equity is market capitalization. The market value of debt is trickier — ideally use the fair value of outstanding bonds, but book value of debt is an acceptable approximation when bond prices are near par. For Apple, with a market cap around $3.5 trillion and total debt of approximately $97 billion, the equity weight is roughly 97% and the debt weight is 3%. The resulting WACC is almost entirely driven by the cost of equity. For a highly leveraged company like a typical utility or REIT, debt may represent 50–60% of total capital, making the cost of debt the dominant component.

Step 4: Terminal Value — The 60–80% of Your Model You Cannot Afford to Get Wrong

After projecting free cash flows for the explicit forecast period (typically 5–10 years), you need to capture the value of all cash flows from that point to infinity. This is the terminal value, and it is the single most influential number in any DCF model.

The Gordon Growth Model (Perpetuity Method)

Terminal Value = Final Year FCF x (1 + g) / (WACC – g), where g is the perpetual growth rate. This formula is elegant but unforgiving. A terminal growth rate of 3% with a 9% WACC produces a terminal value multiple of 17.2x the final-year FCF. Change g to 4% and the multiple jumps to 20.8x — a 21% increase in terminal value from a single percentage point change in an unknowable long-run assumption.

The terminal growth rate should not exceed the long-run nominal GDP growth rate of the economy in which the company operates. For the United States, that is roughly 4–5% nominal (2% real + 2–3% inflation). For most companies, 2–3% is the defensible range. Using a higher rate implies the company will eventually become larger than the economy. It happens occasionally — Apple grew from a $5 billion market cap in 2003 to $3.5 trillion in 2025 — but the DCF terminal value is supposed to capture the mature, steady-state phase of the business, not the growth phase.

The Exit Multiple Method

Terminal Value = Final Year EBITDA x Exit EV/EBITDA Multiple. This method is straightforward but introduces a market-based assumption into what is supposed to be an intrinsic-value framework. The exit multiple should reflect the company's expected valuation at the end of the projection period, when it has matured to steady-state growth. For most industrial and consumer companies, an exit multiple of 8–12x EBITDA is reasonable. For technology companies with durable competitive advantages, 15–20x may be justified.

The critical best practice: use both methods and compare the results. If the Gordon Growth model produces a terminal value of $200 billion and the exit multiple method produces $150 billion, the 25% gap tells you that either your growth rate is too high, your exit multiple is too low, or both. This cross-check is the single most effective quality control for terminal value assumptions. Additionally, reverse-engineer the implied terminal growth rate from the exit multiple (and vice versa) to check for consistency.

Step 5: Discount to Present Value and Derive Equity Value

With projected free cash flows and terminal value in hand, discount each cash flow back to the present using the mid-year convention. The mid-year convention assumes cash flows are received evenly throughout the year rather than as a lump sum at year-end, which is more realistic and produces a slightly higher present value (typically 2–4% higher than the year-end convention).

The sum of all discounted cash flows, including the discounted terminal value, gives you the enterprise value (EV). To get equity value, subtract net debt (total debt minus cash and equivalents), subtract minority interests, subtract preferred equity, and add the value of any non-operating assets (equity investments, excess real estate). Divide by diluted shares outstanding to get the implied share price.

A worked example: Suppose you are modeling a company with projected Year 1 UFCF of $5.0 billion, growing at 8% annually through Year 5, with a WACC of 9.5% and a terminal growth rate of 2.5%. Year 5 UFCF is $6.8 billion. Terminal value is $6.8B x 1.025 / (0.095 – 0.025) = $99.6 billion. The present value of the Year 1–5 cash flows is approximately $22.1 billion (using mid-year discounting). The present value of the terminal value is $99.6B / (1.095)^5 = $63.2 billion. Enterprise value is $85.3 billion. Subtract $10 billion in net debt and you arrive at an equity value of $75.3 billion. With 2 billion diluted shares, the implied price is $37.65 per share.

Note that terminal value accounts for 74% of total enterprise value in this example. That is not a flaw in the model — it is a feature of the math. The implication is that your terminal value assumptions are roughly three times more important than your explicit forecast assumptions. Plan your analytical effort accordingly.

DCF Components: Sensitivity and Common Errors

The following table summarizes each major DCF component, its typical sensitivity to the final valuation, and the most common error analysts make when estimating it.

The 7 Pitfalls That Destroy Otherwise Good DCF Models

Having built hundreds of DCF models and reviewed thousands more, we believe these are the errors that most frequently produce misleading valuations. Each one is avoidable with discipline.

1. Confirmation Bias in Assumptions

The analyst who is bullish on a stock builds a bullish DCF. The analyst who is bearish builds a bearish DCF. The model becomes a tool for justifying a pre-existing view rather than testing it. The solution is to build the base case first, without any view on whether the stock is cheap or expensive. Only after the model is complete should you compare the implied value to the market price.

2. Revenue Hockey Sticks

Projecting revenue growth that accelerates or remains flat at a high rate for the entire forecast period. In reality, virtually every company experiences growth deceleration as it scales. The law of large numbers applies. Alphabet grew revenue 41% in 2021, 10% in 2022, 9% in 2023, and 14% in 2024 (the 2024 reacceleration driven by AI-related cloud demand, not organic search growth normalization). A DCF that projects 20% revenue growth for Alphabet over a five-year forecast without a clear deceleration path is not credible.

3. Terminal Value Neglect

As discussed, terminal value is 60–80% of the answer. Yet in practice, it is the last assumption entered — often a round number ("3% growth, 12x exit multiple") with no analytical support. Always derive the terminal growth rate from the company's expected reinvestment rate and return on invested capital at maturity. If ROIC is 15% and the company reinvests 20% of NOPAT, the sustainable growth rate is 3% (15% x 20%). This approach ties the terminal growth assumption to the company's actual economics rather than a round number.

4. Ignoring Stock-Based Compensation

Stock-based compensation (SBC) is a real economic cost to shareholders because it dilutes their ownership. Many DCF models add SBC back as a non-cash expense, treating it like depreciation. This is incorrect. SBC is not a non-cash expense in the same way depreciation is. It transfers economic value from existing shareholders to employees. The clean approach: include SBC as an operating expense in the UFCF calculation (which it is under GAAP), and use fully diluted shares (including in-the-money options and unvested RSUs) when converting equity value to per-share value. For technology companies where SBC can represent 10–25% of revenue, this treatment materially affects the valuation. Snowflake's SBC was 40% of revenue in fiscal 2024. Treating that as non-cash would dramatically overstate free cash flow.

5. Circular WACC References

WACC depends on the market value of equity, but the market value of equity is what the DCF is trying to determine. This creates a circular reference. The standard solution is to use the current market capitalization for the weighting (since you are testing whether the current price reflects fair value) or to iterate the model until the WACC-implied equity value converges with the equity weight used in the WACC calculation. In practice, the difference between these approaches is usually small unless the stock is dramatically over- or undervalued.

6. No Sensitivity Analysis

A DCF that produces a single price target without a sensitivity table is incomplete. At minimum, present a two-way sensitivity table showing the implied price across a range of WACC and terminal growth rate assumptions. Better yet, run a Monte Carlo simulation that varies all key assumptions simultaneously to produce a probability distribution of fair value. The point is not precision — it is understanding the range of reasonable outcomes and where the current market price sits within that range.

7. Not Sanity-Checking Against Multiples

A DCF model exists in a vacuum unless you check its implied multiples against market reality. If your DCF implies a fair value of $200 per share and the stock trades at $100, ask: what forward P/E does $200 imply? If the answer is 45x for a company growing earnings at 10%, the market is unlikely to assign that multiple, regardless of what the DCF says. The implied multiple serves as a reality check on your assumptions. For more on how different valuation methods complement each other, see our guide on calculating intrinsic value using five methods.

Frequently Asked Questions About DCF Models