DCF modeling

The claim, and why cash

Prices wander daily; the question a DCF answers is what the business is worth. Worth, to a finance person, means: if you owned the whole company and could take out everything it earns beyond what it needs to keep running, what is that stream of withdrawals worth today? The stream is called free cash flow: the cash the business generates after paying for operations and the reinvestment needed to keep operating and growing. It is what could be handed to investors without harming the machine that produces it.

A DCF counts cash, not earnings, because earnings are an opinion shaped by accounting choices (depreciation schedules, revenue recognition, provisions) while cash entering or leaving is a fact. A company can report profits for years while consuming cash; businesses die of the second, never the first. And it counts cash in today's money, because a dollar promised in 2032 is worth less than one in hand: it must be shrunk ("discounted") for the waiting and for the risk that it never arrives. The shrinking rate is the discount rate: the return you could demand elsewhere at comparable risk, which is how risk becomes arithmetic.

The machine at a glance, and our worked example

Every DCF at every bank and fund is the same five moves:

• Forecast free cash flow over an explicit horizon, usually five to ten years.
• Choose a discount rate matching the riskiness of those flows.
• Estimate a terminal value for everything beyond the horizon.
• Discount everything to today and sum it.
• Bridge to a share price: subtract what belongs to others, divide by diluted shares.

To keep every step concrete, meet Meridian Software, an invented but realistic mid-cap: $1,000M revenue growing 12%, a 25.5% operating margin expanding slowly, a 21% tax rate, $500M of debt against $300M of cash, and 100M diluted shares trading at $86. By the end of this guide Meridian is fully valued, and the result will teach one more lesson than expected.

Step 1: Define the cash flow (FCFF vs FCFE)

There are two standard definitions, and the model must commit to one. Free cash flow to the firm (FCFF) is the cash available to everyone who financed the company, lenders and shareholders together, measured before any interest: it values the whole enterprise. Free cash flow to equity (FCFE) is what remains for shareholders alone, after interest and after debt raised or repaid: it values the equity directly. Wall Street's default is FCFF, because it values the operations independently of today's financing mix.

revenue                            1,120.0    (1,000 x 1.12)
EBIT at a 25.5% margin               285.6    operating profit
x (1 - 21% tax rate)                 225.6    = NOPAT, the unlevered profit
+ depreciation & amortization         44.8    non-cash; 4% of revenue
- capital expenditure                (56.0)   5% of revenue
- increase in net working capital    (12.0)   10% of the revenue increase
FCFF                                 202.4

Each adjustment has a reason. NOPAT (net operating profit after tax) taxes EBIT as if the company had no debt, stripping financing choices out of the operating result. Depreciation and amortization come back because that cash left when the assets were bought, not this year. Capital expenditure is subtracted because that is when cash actually leaves for new assets. And growth absorbs cash into net working capital (receivables and inventory, minus the payables that finance them): customers owe you more before they pay you more, so a growing top line quietly consumes cash on its way up.

Step 2: Forecast the horizon years

The forecast is not a guess at one growth number; it is a small model of the business, built from drivers in a fixed order: revenue from its real units (customers times price, stores times sales per store, seats times utilization), margins moving toward a defensible long-run level, and reinvestment consistent with the growth being claimed. Growth is bought, not free: a model growing revenue 15% a year on maintenance-level capex is asserting a miracle, and reviewers look for exactly that assertion first.

Note the shape: growth glides from 12% toward 5%, margins settle, and net reinvestment shrinks as growth slows, because slower growth needs less new capital. The horizon must end in a steady state (growth near the economy's, margins stable, reinvestment matched to growth) because the terminal formulas in Step 4 are only valid for a business that has stopped changing shape. Five years suits a stable company; a young or cyclical one needs ten.

Step 3: The discount rate

The discount rate is an opportunity cost: what investors could earn elsewhere at similar risk. For the equity holders that is the cost of equity, and the standard estimator is the CAPM (capital asset pricing model): begin at the risk-free rate, the yield on a long government bond in the same currency as the cash flows, then add compensation for market risk, scaled by how much of it this particular stock carries.

cost of equity = risk-free + beta x equity risk premium
               = 4.2%      + 1.10 x 5.0%              = 9.7%

after-tax cost of debt = 5.5% x (1 - 0.21)            = 4.3%
(interest is tax-deductible, so debt's true cost is net of the shield)

market values: equity = 100M sh x $86 = 8,600   debt = 500   V = 9,100

WACC = (8,600/9,100) x 9.7%  +  (500/9,100) x 4.3%    = 9.4%

Beta is the stock's sensitivity to the market: 1.10 means roughly an 11% move when the market moves 10%. The equity risk premium is the extra annual return stocks demand over government bonds, an estimate argued for decades and usually set at 4-6%. The WACC (weighted average cost of capital) blends equity and debt at market values. Every input is an estimate, so professionals treat the rate as a dial with a plausible range (roughly 7-12% for most public equities) and show the valuation across it rather than defending 9.41% to two decimals.

Step 4: Terminal value, both ways

Meridian does not end in year 5; everything beyond the horizon collapses into one number, the terminal value, and there are exactly two standard ways to compute it. Serious models run both and make each one audit the other.

Method A: the growing perpetuity (Gordon growth)

TV = FCFF(yr 5) x (1 + g) / (WACC - g)
   = 307.2 x 1.025 / (0.094 - 0.025)
   = 314.9 / 0.069                          = 4,564

g = 2.5%. Perpetual growth may not exceed long-run nominal GDP (~2-3%):
a company growing faster than the economy forever eventually IS the economy.

Method B: the exit multiple (a buyer's lens)

TV = EBITDA(yr 5) x exit multiple
   = (EBIT 411.2 + D&A 60.9) x 10.0x        = 4,721

10x taken from where comparable software companies trade today.

THE CROSS-CHECK, always: 4,721 implies a perpetual growth of ~2.8%
(solve g from method A's formula). Believable. An exit multiple that
implies 6% growth forever is announcing that it is too high.

The perpetuity method is internally consistent but hypersensitive to the spread between WACC and g: shrink the denominator from 6.9% to 5.9% and terminal value jumps 17%. The exit multiple imports the market's current mood into the model's end point, which is both its use and its danger. We carry the perpetuity figure.

Step 5: Discount everything, then bridge to a share price

Each year's flow is divided by (1 + WACC) raised to its year; the terminal value is discounted from year 5, since that is where it stands:

enterprise value = (185.0 + 192.0 + 198.3 + 201.0 + 196.0) + 2,912
                 = 972.3 + 2,912                          = 3,884

equity value    = enterprise value - net debt
                = 3,884 - (500 debt - 300 cash)           = 3,684

value per share = 3,684 / 100M diluted shares             = $36.84

...and the stock trades at $86. Do not "fix" the model to match.
Read the next section instead: the gap is the analysis.

The bridge's mechanics matter more than they look. Enterprise value belongs to all capital providers together, so shareholders receive what remains after net debt (total debt minus cash and equivalents), preferred stock, and minority interests, all measured on the same date and none allowed to silently default to zero: a missing net debt figure that becomes 0 values every levered company as if it were debt-free, an error proportional to leverage and invisible on net-cash names. Divide by diluted shares, counting the stock that options and convertibles will add, because those claims are real before they convert.

Reading the answer: a range, never a point

No professional presents a DCF as one number. The standard exhibit is a two-way sensitivity table across the two most contentious inputs, and Meridian's makes the situation plain:

The base build says $37; the friendliest corner of defensible assumptions says $60; the market says $86. That gap IS the finding. The price already assumes something beyond this build: faster growth, structurally higher margins, or cheaper capital. The honest next step is the reverse DCF: solve for what $86 implies (for Meridian, roughly holding 12%+ growth twice as long with margins reaching the low 30s) and judge whether THAT is beatable. A buy case here is a bet on those specific assumptions; naming them is what the model was for. Conviction belongs where the price sits outside the entire table, and demanding that distance before acting is what margin of safety means in practice.

Refinements professionals actually use

• Mid-year convention. Cash arrives through the year, not on December 31, so discount each flow by t − ½. Worth roughly half a year of value (4-5% at a 9% rate); most bank models switch it on.
• Stock-based compensation. The modern trap. SBC is a real expense paid in shares. Either charge it against FCF as if cash, or add the future dilution to the share count. Doing NEITHER, which is what company-adjusted figures invite, overstates value, sometimes enormously in tech.
• Operating leases. Under IFRS 16 / ASC 842 they sit on the balance sheet as debt-like liabilities. Keep treatment consistent: if lease liabilities count in net debt, lease interest must not also burden FCFF.
• Net operating losses. Accumulated tax losses mean years of below-statutory cash taxes; model the NOL runway explicitly instead of a flat 21% forever.
• Cyclicals: normalize first. A DCF launched off peak earnings embeds the peak forever; use mid-cycle margins as the base year. (The Analyzing a Stock guide shows this in the method-weighting table.)
• Scenario weighting. Run bear, base and bull builds and probability-weight them. One path hides the asymmetry; three paths reveal whether the downside is survivable and the upside real.

Where DCFs go wrong

• Growth without reinvestment. Revenue compounding at 15% on flat capex. Growth costs capital; check the implied capital efficiency every time.
• Terminal growth above the economy's. Anything past ~3% nominal claims the company eventually outgrows the world.
• Crossed cash flow and rate. FCFE at the WACC, or FCFF at the cost of equity. The matching rule, violated silently.
• A broken bridge. Stale or zeroed net debt, basic instead of diluted shares, forgotten minority interests and preferreds.
• Tuning inputs to the desired answer. The model has enough dials to hit any target; analysis fixes the assumptions from evidence FIRST and then accepts the output. The reverse order is decoration.
• Precision cosplay. A WACC quoted to two decimals does not narrow the honest range. The sensitivity table is the truthful output format, and refusing to publish one is a tell.

How Wall Street actually uses it

In a banker's deck the DCF is one panel of the football field, shown beside trading comparables and precedent transactions, leaned on hardest where market benchmarks are thin (unique assets, long-dated projects) and least where comps are liquid and current. On the buy side, the reverse DCF is often the primary weapon: extract the expectations embedded in the price, then hunt the gap between what is priced and what is likely. Both directions are the same machine. Either way, the durable value of building the model is not the output number: it is that you now know exactly which assumptions the case depends on, and those are the things to research before buying and to monitor after.