Stock Valuation Guides: DCF, P/E, and Multiples Explained

No single method works for every company. Experienced analysts triangulate across three approaches:

• Discounted cash flow (DCF). The theoretically correct method — price equals the present value of future free cash flows. Best for mature, FCF-positive businesses where you can reasonably model the next 5–10 years. Useless for pre-revenue companies or those with highly cyclical or lumpy FCF.
• Relative multiples. Compare the stock to peers on P/E, EV/EBITDA, EV/Revenue, or P/FCF. Fast and practical, but dangerous in isolation — it tells you whether the stock is cheap or expensive relative to peers, not whether the whole sector is mispriced. Use multiples to calibrate DCF outputs, not replace them.
• Asset-based / sum-of-the-parts (SOTP). Value each business segment or asset separately, then sum and subtract net debt. Best for conglomerates, holding companies, or asset-heavy businesses (real estate, mining) where the operating model is secondary to balance sheet composition.

The most robust valuation uses all three. If DCF says $80, EV/EBITDA says $95, and SOTP says $72, you have a range with $75–85 as the credible zone — not a single price target that implies false precision. Try the Intrinsic Value Calculator to get a quick per-share estimate with margin of safety built in.

A DCF has many inputs but three dominate the output:

• Free cash flow growth rate (years 1–5). The near-term forecast. Anchor this to actual business mechanics: revenue growth, margin trajectory, capex intensity. Do not extrapolate recent growth linearly — check for mean reversion signals in margins and competition.
• Terminal growth rate. The perpetuity growth rate applied after your forecast period — typically 2–4% for a stable US business (roughly in line with nominal GDP). Every basis point matters here because it applies to a perpetuity. A 3% vs. 4% terminal rate can swing intrinsic value by 15–25% on a high-multiple stock.
• Discount rate (WACC). The weighted average cost of capital — blends cost of equity and after-tax cost of debt. For large-cap US equities, 8–10% is typical. For high-growth or unprofitable companies, 12–15% is more defensible. A lower discount rate inflates every future cash flow — analysts gaming DCFs tend to shave discount rates, not change growth assumptions.

Run sensitivity tables on these three inputs before trusting any DCF output. If your investment thesis only works at a 7% discount rate and 5% terminal growth, you should know that before buying.

Multiples are context-dependent. Using the wrong one for the wrong business type is a common and costly error:

• P/E. Simplest and most widely cited — but distorted by leverage, non-cash charges, and tax timing differences. Useful for comparing similar businesses with similar balance sheets. A P/E of 25x is cheap for a company growing EPS at 30%; it's expensive for a company growing at 5%. Never look at P/E without looking at the earnings growth rate alongside it.
• EV/EBITDA. The standard for M&A and LBO analysis because it's capital-structure neutral (enterprise value captures both equity and debt). Better than P/E when comparing companies with different leverage ratios or depreciation policies. Sector medians: software 20–30x, industrials 8–12x, banks N/A (use P/B or P/TBV instead).
• EV/FCF (or P/FCF). The most honest multiple for FCF-heavy businesses because it doesn't rely on EBITDA adjustments or accounting choices. If a company trades at 40x EBITDA but 20x FCF, it's a lower-capex, lower-maintenance business than the EBITDA multiple suggests — and may be cheaper than it appears.
• P/S (price-to-sales). Used for unprofitable growth companies as a proxy for future earnings potential. Low P/S only matters if the business can reach meaningful margins at scale — a P/S of 3x on a 5% gross margin business is not cheap. Compare P/S to gross margin percentages, not standalone.

When you buy a share of Microsoft, you're not buying a ticker symbol. You're buying a fractional claim on a business that generated $74 billion in free cash flow in fiscal 2024 — money left over after paying every bill, every employee, and every capital expenditure. Intrinsic value is a simple idea: it's the sum of all that future cash, discounted back to what it's worth in today's dollars. Nothing more.

Market price is something else entirely. It's what the next buyer will pay you right now. Sometimes that buyer is rational. Often they're extrapolating last week's headlines. The gap between what a business is worth and what the market charges for it is where investing actually happens.

Microsoft trades near $450 a share — all-time highs. That feels expensive until you think through what you're buying. Azure is growing roughly 30% a year and compounding into a business that didn't meaningfully exist a decade ago. Office 365 and LinkedIn throw off cash with almost no incremental cost. The company earned about $11.80 per share in FY2024. At 38 times earnings, you're paying a premium. But if Azure sustains double-digit growth and margins stay near 45%, the cash flows you collect over the next 20 years could justify that price — or exceed it. A stock at an all-time high can still be cheap if the underlying earnings power is growing faster than the price.

The reverse is also true. A stock down 60% from its peak is not automatically cheap. If a company is burning cash, losing customers, and facing a secular headwind, the intrinsic value may be falling faster than the price. Cheap-looking is not the same as undervalued.

This is the discipline intrinsic value imposes: it forces you to think like an owner, not a trader. An owner asks what this business will produce over its lifetime. A trader asks what someone will pay for it tomorrow. Intrinsic value analysis won't tell you when the stock moves — it tells you whether you're being paid fairly for the risk you're taking. That's the only question worth answering before you buy.

A discounted cash flow analysis answers one question: what should you pay today for every dollar this business will ever generate? Strip away the jargon and that's it. You're estimating all future cash flows, then discounting them back to present value because a dollar three years from now is worth less than a dollar today.

Start with free cash flow — operating cash flow minus capital expenditures. Alphabet generated roughly $72 billion in FCF in 2024. That's the engine. Next, project how fast that engine grows. Alphabet's revenue has compounded around 12–15% annually over the past five years, but most analysts model a gradual slowdown — call it 12% declining to 8% over a ten-year projection window. The exact number matters enormously. A 10% growth assumption versus 8% can swing a valuation by 20% or more.

Then you need a discount rate — typically the weighted average cost of capital, or WACC. For a large-cap like Alphabet, analysts often use 9–10%. This reflects what equity investors require as a return, blended with any debt costs. The higher the discount rate, the lower the present value of future cash flows. If you think markets are riskier than average, you raise it. If you're modeling a near-monopoly with predictable cash flows, you might shade it lower.

1. Estimate free cash flow — Alphabet's ~$72B starting point
2. Project growth — model annual FCF growth over 10 years, stepping down from 12% to 8%
3. Apply a discount rate — 9–10% WACC converts future dollars to today's dollars
4. Calculate terminal value — assume the business grows at ~3% in perpetuity beyond year 10

That last step is where most DCFs live or die. Terminal value — the present value of all cash flows beyond your projection window — typically represents 70% or more of the total estimated value. For Alphabet, using a 3% long-term growth rate and a 9.5% discount rate, the terminal value alone runs into the trillions. Which is the point: you're not really modeling ten years of cash flows. You're modeling a perpetuity with a ten-year warm-up. Get the terminal assumptions wrong and the whole analysis is off.

This is why two analysts using the same starting FCF can reach valuations 40% apart. DCF isn't a calculator that produces the right answer. It's a structured way to make your assumptions explicit — and then argue about them.

Every valuation multiple is a shortcut. Each one answers a slightly different question, and each one can lie to you in a different way. Knowing which multiple to reach for — and what it's hiding — separates useful analysis from noise.

P/E is the most widely quoted multiple and the easiest to manipulate. It divides share price by GAAP earnings per share. Apple trades around 28x trailing earnings. That sounds clean. But GAAP earnings absorb depreciation, amortization, stock-based compensation, and one-time charges. A company can depress earnings by accelerating write-downs, or inflate them by releasing reserves. When you see a P/E that looks suspiciously low or high, the first question is whether the "E" is clean.

EV/EBITDA strips out the capital structure so you can compare companies regardless of how they're financed. Enterprise value — market cap plus net debt — is divided by earnings before interest, taxes, depreciation, and amortization. A company with $10B in debt and one with none can both post similar EBITDA; the EV/EBITDA multiple reflects what you'd actually pay to own the whole business. The catch: it ignores capex entirely. For capital-intensive businesses like airlines or semiconductor fabs, EBITDA can massively overstate economic earnings.

P/FCF — price to free cash flow — gets closest to what the business actually produces. Free cash flow is operating cash minus capex. It's harder to engineer than GAAP earnings. Netflix is the textbook case: in 2023 it reported approximately $5.1B in GAAP net income but generated roughly $6.9B in free cash flow. The P/E implied one valuation. The P/FCF implied a materially cheaper stock. The difference came from non-cash charges that compressed reported earnings without touching actual cash generation.

EV/Sales matters most when a company earns nothing — or nearly nothing. High-growth SaaS and biotech firms often trade on revenue multiples because earnings are years away. A 10x EV/Sales multiple isn't inherently expensive if margins are expanding toward 30%; it can be catastrophic if they're not. Revenue multiples require a thesis about future profitability. Without one, you're guessing.

No single multiple is right. Use at least two that approach the business from different angles. If P/E and P/FCF tell conflicting stories, the gap usually explains exactly what management wants you to overlook.

Valuation looks mechanical until you realize how many ways you can get the inputs wrong. These aren't obscure edge cases. They're the mistakes that cost investors real money, repeatedly.

1. Anchoring to a historical multiple that no longer applies. Intel traded at 15–20x earnings for years when it dominated PC processors and printed cash. Investors bought the 2022 dip at 12x assuming mean reversion. The problem: Intel's competitive position had collapsed, margins were under structural pressure, and the earnings base itself was impaired. A cheap multiple on a broken business isn't value — it's a trap. Always ask why the stock is at its historical discount before assuming the market is wrong.
2. Ignoring the denominator. The P/E ratio is a fraction. Most investors fixate on the price; few interrogate the earnings. Meta traded at roughly 20x in late 2022, which looked reasonable until you noticed analysts were cutting estimates quarter after quarter. Twenty times declining earnings is more expensive than 25x growing earnings — the forward multiple expands as the E collapses under you. When Peloton was still posting triple-digit revenue growth, its absurd earnings multiple was at least tracking something real. When growth stalled and losses mounted, the multiple meant nothing.
3. Confusing a great business with a great investment. Visa processes $14 trillion in payment volume annually, earns 50%+ operating margins, and requires almost no capital. It is one of the finest businesses ever built. That doesn't mean you make money buying it at any price. At 35x earnings with mid-single-digit revenue growth, you need everything to go right just to match the market. The same trap caught investors in Starbucks through 2023 — premium multiple, premium brand, mediocre returns as sentiment cooled. Quality earns a premium. It doesn't justify an unlimited one.

Every valuation is a set of assumptions dressed up as a number. The problem is that most investors treat the output — a $180 price target, say — as if it were a fact rather than a forecast built on inputs that could easily be wrong. Stress-testing forces you to make those inputs visible, then deliberately break them. A model that falls apart at 8% growth when you assumed 15% isn't conservative analysis — it's wishful thinking with a spreadsheet attached.

1. Write every assumption down explicitly. Revenue growth rate. Operating margin at maturity. Terminal growth rate. Discount rate. Don't leave them buried in cells. List them in plain English: "I'm assuming 15% annual revenue growth for five years, expanding to a 22% free cash flow margin by year five, discounted at 10%." Now you can attack them.
2. Identify which assumption drives the most value. In a high-growth tech stock, it's almost always the growth rate. Run a simple sensitivity table: hold everything else fixed, change only growth. For a company like a mid-cap SaaS trading at 30x forward earnings, the difference between 20% growth (bull), 12% (base), and 5% (bear) might produce intrinsic values of $95, $58, and $31 respectively. That $64 spread is your real risk — not volatility, not macro, not management commentary.
3. Assign honest probabilities to each scenario. Most analysts weight base at 60%, bull at 20%, bear at 20%. That's a reasonable starting point, but tech stocks mean-revert. If the sector is decelerating — cloud spend slowing, ad markets softening — the bear case deserves more weight than habit suggests.
4. Ask whether you can hold at the bear-case price. If the bear case implies a 45% drawdown from current levels and you'd panic-sell at minus 20%, the position is too large regardless of what the base case shows. Stress-testing isn't just analytical — it's a position-sizing tool.