A practitioner's guide to DCF: where the model lies

The seduction of a clean number

Every analyst remembers the first time a discounted cash flow model spat out a single, confident figure. You feed it revenue growth, margins, a discount rate, a terminal value, and out comes a price target carried to two decimal places. It feels like science. It looks like rigour. And that feeling is exactly the problem.

DCF is the most defensible valuation method in theory and the most abused in practice. The theory is unimpeachable: a business is worth the present value of the cash it will return to its owners over its life. Nobody serious disputes that. The abuse comes from the fact that the model is almost entirely forward-looking, and the future is a place where you can put whatever you want. A multiples-based valuation at least anchors you to what other people are paying right now. A DCF anchors you to your own assumptions, and your assumptions are negotiable in a way that the market price of a comparable transaction is not.

This guide walks through the mechanics from first principles, builds a complete worked example for a fictional company, and then does the part most textbooks skip: it shows you precisely where the model bends, who tends to bend it, and how to tell an honest DCF from a reverse-engineered one. The goal is not to make you distrust the tool. The goal is to make you fluent enough that nobody can hide behind it in front of you.

Part one: the machinery

The core identity

At its heart, DCF rests on a single equation that almost everyone has seen and far fewer have internalised. The value of an asset today equals the sum of its future cash flows, each one discounted back to the present by a rate that reflects the time value of money and the riskiness of those flows: PV = Σ CFₜ ÷ (1 + r)ᵗ.

In words rather than symbols: a dollar arriving five years from now is worth less than a dollar in your hand today, both because you could have invested today's dollar in the meantime and because the future dollar might never show up. The discount rate is the machine that converts future dollars into present ones, and the further out a cash flow sits, the more aggressively the machine shrinks it.

The decay of a future dollar

The two-stage structure

Almost every real-world DCF is built in two stages, and understanding why is essential to understanding where the manipulation lives.

The explicit forecast period is the stretch of years, usually five to ten, over which you project cash flows line by line. You make explicit assumptions about revenue, margins, taxes, capital expenditure, and working capital for each individual year. This is the part that looks like hard work, and it is, but it is rarely where the value sits.

The terminal value captures everything beyond the explicit period, compressed into a single number. Because no business stops generating cash at the end of year ten, you need a way to value the indefinite tail. And here is the uncomfortable truth that the rest of this guide keeps returning to: in a typical DCF, the terminal value accounts for somewhere between 60% and 80% of the total valuation. The decade of careful, defensible, line-by-line forecasting that you sweated over is the minority of the answer. The single assumption you make about year eleven onward is the majority.

Where the value actually comes from

Which cash flow?

There are two flavours of DCF, and confusing them is one of the most common ways a model produces nonsense.

Free cash flow to the firm (FCFF), also called unlevered free cash flow, is the cash available to all providers of capital, both debt and equity holders, before any financing decisions. You discount FCFF at the weighted average cost of capital (WACC) and you arrive at enterprise value, the value of the whole business regardless of how it is financed.

Free cash flow to equity (FCFE), also called levered free cash flow, is the cash left over for shareholders after debt holders have been paid their interest and principal. You discount FCFE at the cost of equity alone and you arrive directly at equity value.

The unlevered approach is far more common in practice because it cleanly separates operating performance from financing structure, and the two should reconcile if done correctly. The cardinal sin is mixing them: discounting unlevered cash flows at the cost of equity, or levered cash flows at WACC. Each error inflates the answer, and because the mechanics look identical from a distance, the mistake survives more review meetings than it should.

The free cash flow calculator builds the unlevered line above; the two paths must never be crossed.

Part two: the worked example

Let us build a complete DCF for a fictional company so the abstractions become concrete. Meet Meridian Components, a mid-sized manufacturer of specialised industrial sensors. It is profitable, growing steadily, and the sort of unglamorous business that lends itself to a clean valuation. We will value it from the top down, then spend the second half of the guide picking the model apart. (Every figure works in unitless millions to keep the arithmetic clean.)

Step one: forecast revenue

Meridian did 500 in revenue last year. Management guides to strong near-term growth that moderates over time as the business matures and its markets saturate. We assume revenue growth of 12% in year one, stepping down by roughly two percentage points each year until it settles into a mature 4% by the end of the forecast.

The deceleration of the growth rate is itself an assumption, and arguably the single most consequential one in the entire near-term forecast.

Step two: project margins and operating profit

We assume Meridian holds an EBIT margin of 18%, with a modest expansion to 19% by year three as scale improves, then holding flat. Applying those margins to the revenue line gives us operating profit for each year.

Step three: convert EBIT to unlevered free cash flow

This is the FCFF build from the table above, applied year by year. From EBIT we subtract taxes (assume a 25% rate), add back depreciation and amortisation (assume it runs at around 5% of revenue), subtract capital expenditure (assume 8% of revenue, comfortably above the 5% D&A line because the company is still building out its asset base to fund growth), and subtract the increase in net working capital (assume it consumes around 10% of each year's incremental revenue).

The cash conversion lands in the high single digits to low teens as a share of sales, which is plausible for an industrial manufacturer. If that bottom row had read 25%, it would be a signal that the capital expenditure or working-capital assumptions were too kind.

Step four: choose the discount rate

We need a WACC for Meridian. The cost of equity comes from the capital asset pricing model: a risk-free rate (assume 4%, roughly a long-dated government bond yield), plus the company's beta (assume 1.2, slightly more volatile than the broad market) multiplied by an equity risk premium (assume 5%). We assume Meridian's after-tax cost of debt is 4.5%, and that the company is financed 70% by equity and 30% by debt.

Every single input here is an estimate, and several are genuinely contestable. Hold that 8.4% in your mind — we are going to torture it later. (You can rebuild it yourself with the WACC calculator and the cost of equity calculator.)

Step five: discount the explicit cash flows

Each year's free cash flow gets multiplied by its discount factor, which is 1 ÷ (1 + WACC)ᵗ. Year one's cash flow is divided by 1.084, year two's by 1.084 squared, and so on. Summing the five discounted figures gives the present value of the explicit-period cash flows.

The discount factors visibly shrink each year, exactly as the decay chart promised — and notice how little of the heavy lifting the explicit period does.

Step six: calculate the terminal value

Here is the part that decides the whole valuation. We use the Gordon growth (perpetuity) method: the terminal value at the end of year five equals the year-six free cash flow divided by (WACC minus the perpetual growth rate). We assume Meridian's free cash flow grows forever at 2.5%, a figure meant to approximate long-run nominal GDP growth, the idea being that no company can outgrow the economy indefinitely.

With a year-five free cash flow of about 80, growing at 2.5% to give a year-six figure of about 82, and a WACC of 8.4%, the terminal value lands at 82 divided by (0.084 minus 0.025), which is approximately 1,390. We then discount that back to today with the year-five discount factor.

The terminal value calculator runs this step in isolation, and the enterprise value calculator handles the bridge from enterprise to equity value.

The result, and the warning

When the dust settles, Meridian has an enterprise value of roughly 1,190, an equity value of roughly 1,040, and the analyst gets to write a confident number on the front page of the pitch. But notice what just happened. Of that enterprise value, around 925 came from the terminal value and only around 264 came from the five years of careful forecasting. More than three-quarters of the answer rests on two numbers we chose almost casually near the end: the perpetual growth rate and the discount rate.

That is not a flaw in our particular model. That is DCF. And it is why the next section exists.

Part three: where the model lies

A DCF does not lie on its own. It is a faithful calculator. It lies when a human points it at a conclusion and works backward, adjusting the inputs until the output matches a number that was decided before the modelling began. The dangerous part is that every individual adjustment can be defended in isolation. The manipulation lives in the accumulation, and in the direction of the thumb on the scale. Here are the pressure points, roughly in order of how much damage they can do.

Lie number one: the terminal value, where most of the body is buried

We have already established that the terminal value is usually 60% to 80% of the answer. That alone should make it the first place you look. But the specific vulnerability is the perpetual growth rate, because of the cruel mathematics of the perpetuity formula. The denominator is WACC minus the growth rate, and as the growth rate creeps closer to WACC, that denominator shrinks toward zero and the terminal value explodes toward infinity.

In our Meridian model, moving the perpetual growth rate from 2.5% to 3.5% does not change the answer by a modest amount. The denominator falls from 0.059 to 0.049, the terminal value jumps from roughly 1,385 to roughly 1,670, and a full chunk of enterprise value materialises out of a one-percentage-point change to a single assumption about the indefinite future. An analyst who wants a higher number does not need to touch the operating forecast at all. They simply argue, with a straight face, that the company's long-run growth is "closer to GDP-plus" and let the perpetuity formula do the rest.

The terminal value cliff — enterprise value vs perpetual growth

The honest discipline here is a hard ceiling: the perpetual growth rate must never exceed long-run nominal GDP growth, because a company that grew faster than the economy forever would eventually become the entire economy. Most careful practitioners cap it well below GDP, often in the 2% to 3% range, and treat anything higher as a red flag requiring extraordinary justification.

There is also a sanity check that the perpetuity formula begs for and rarely gets: the implied exit multiple. Take your terminal value and divide it by the terminal-year EBITDA. If your perpetuity assumption implies the business will be worth 30 times earnings in year ten when comparable mature businesses trade at 12 times, your growth assumption is doing the lying for you.

Here the perpetuity method implies a conservative 7.9× while a market-based exit multiple of 11× would imply 1,936 — a 40% gap. When the two roads disagree this far, one of your assumptions is fantasy, and the discipline of forcing them to converge is one of the best defences against terminal-value abuse. The EBITDA multiple calculator is the cross-check.

Lie number two: the discount rate, the quiet lever

If the terminal value is where the body is buried, the discount rate is the shovel. It compounds against every cash flow in the model, and because it sits in the denominator of both the explicit-period discounting and the terminal-value formula, small changes ripple through everything at once.

Recall our Meridian WACC of 8.4%. Watch what a seemingly reasonable disagreement does. Argue the beta down from 1.2 to 1.0 because "the business has become less cyclical," and the cost of equity falls, the WACC falls, and both the explicit cash flows and the terminal value get discounted less harshly. Trim the equity risk premium from 5% to 4.5% because "current market conditions warrant it," and the WACC falls again. Each adjustment is individually arguable. Together they might move the WACC from 8.4% to 7.4%, and a single percentage point off the discount rate, applied across the model with a fat terminal value, can lift the valuation by 15% to 25% with no change whatsoever to the operating story.

Valuation vs discount rate

The components most open to manipulation, roughly in order, are the equity risk premium (a genuinely contested number where reasonable estimates range across several percentage points), beta (which depends entirely on the lookback period and the choice of comparables), and the capital structure weights (where using a "target" capital structure rather than the actual one quietly changes the blend). None of these is fraudulent on its own. All of them are soft, and softness in the discount rate is leverage on the whole model. The discount rate calculator lets you see how far each lever moves the answer.

Lie number three: hockey-stick revenue and margin creep

The explicit forecast period may be the minority of the valuation, but it is the part that looks most rigorous, and that respectability makes its distortions easy to wave through. The classic tell is the hockey stick: a revenue line that grows modestly for a year or two (matching recent actuals, so it looks grounded) and then accelerates into the back half of the forecast, conveniently in the years that are far enough away that nobody can falsify them yet. The same trick appears in margins, where a company currently earning an 18% EBIT margin is forecast to expand steadily to 25% by year five on the strength of "operating leverage" and "mix shift" that have not yet shown up in a single quarter of results.

The hockey stick, exposed — revenue growth rate

The defence is to compare the forecast against two things: the company's own history and the realistic ceiling of its industry. If a business has never grown faster than 10% and the model assumes 20% in year four, someone owes you an explanation. If the entire industry operates at a 15% margin and the model marches to 25%, the burden of proof sits squarely on the optimist. Growth and margins both face gravity, in the form of competition, market saturation, and reversion to the mean, and a forecast that ignores that gravity is selling you a story, not a valuation.

Lie number four: capital expenditure that forgets to grow

This one is subtler and rarely caught. To generate the growth in the revenue forecast, a company has to invest: in plants, equipment, technology, working capital. An honest model ties capital expenditure to the growth it is supposed to fund. A manipulated model lets revenue sprint ahead while capital expenditure quietly falls as a percentage of sales, which inflates free cash flow in exactly the years that feed the terminal value. The company is, in effect, modelled to grow faster while investing relatively less, which is the corporate equivalent of running a marathon faster while eating less. It happens occasionally in reality through genuine efficiency gains. It does not happen routinely, and a model that assumes it routinely is borrowing free cash flow from a future that will not deliver it.

The check is to look at the relationship between capital expenditure and depreciation, and between capital expenditure and revenue growth, across the forecast. For a growing company, capital expenditure should generally run at or above depreciation. If it dips below depreciation while revenue is still climbing, the model is implicitly assuming the asset base shrinks even as the business expands, and that is usually a sign that free cash flow has been flattered.

Lie number five: the false precision of the output

The final and most pervasive lie is not in any single input. It is in the presentation. A model fed with a dozen contestable assumptions, several of which could each move the answer by 20%, produces a point estimate carried to the decimal. That decimal is a lie of confidence. The genuine output of any DCF is not a number but a range, and the width of that range is itself important information about how much you actually know.

This is why a single-point DCF should always be treated with suspicion and a sensitivity analysis should always be demanded. The standard tool is a two-variable data table that flexes the two assumptions with the most leverage — the WACC and the perpetual growth rate — and shows the resulting valuation across a grid.

Look at the corners. The same business, valued with assumptions that are each individually defensible, is worth 910 in the bottom-left and 1,752 in the top-right — nearly double. The bold centre cell is our base case, but the entire grid is the real answer. Any single number pulled from it without the context of the grid is, at best, a convenient simplification and, at worst, a deliberate one.

Tornado chart — ranking the lies by impact

Part four: how to read a DCF like a sceptic

Having built one and dismantled it, here is the practical checklist. When someone hands you a DCF, do not start at the top with the revenue forecast. Start at the bottom, with the assumptions, because that is where the answer was actually decided.

First, find the terminal value as a percentage of total value. If it is north of 75%, you are not really looking at a cash flow model, you are looking at a single bet on the perpetuity assumption dressed up as a decade of analysis. Second, check the perpetual growth rate against nominal GDP. Third, cross-check the terminal value against an implied exit multiple. Fourth, interrogate the discount rate component by component. Fifth, lay the revenue and margin forecasts against history and industry reality. Sixth, confirm that capital expenditure grows with the business it is meant to fund. Finally, refuse to accept a single number — ask for the sensitivity table, because the width of the range is the truest thing the model has to tell you.

The honest verdict on an honest tool

None of this means DCF is worthless. Quite the opposite. The discipline of building one forces you to articulate, explicitly and in writing, exactly what you believe about a company's future: how fast it grows, how profitable it becomes, how much it must invest, and how risky the whole proposition is. That articulation is enormously valuable, arguably more valuable than the number it produces. A DCF is a structured argument about the future, and a well-built one makes its assumptions visible enough to be challenged.

The trouble only begins when people forget that the visible number is downstream of invisible choices, and when those choices are made to serve a predetermined answer. The model does not lie. People lie, and the model, being an honest calculator, faithfully reports the consequences of whatever lies it is fed. The terminal value will dutifully balloon if you tell it growth is perpetual. The discount rate will obediently lift the valuation if you talk the beta down. The free cash flow will inflate if you let revenue outrun capital expenditure. At every step the arithmetic is impeccable and the inputs are negotiable, and that combination is what makes DCF simultaneously the most teachable and the most manipulable method in valuation.

So learn the machinery until it is second nature, then spend the rest of your career watching the assumptions rather than the answer. Build your own from first principles with the DCF calculator, and pressure-test every figure that feeds it. The number on the front page was never the point. The argument behind it always was.