Estimating Intrinsic Value with the DCF Model

The number you are trying to build

The last lesson gave you ratios — P/E, P/S, P/B — and ratios are a filter. They take a list of thousands of companies and hand you back a short list worth a closer look. They do not tell you what a company is worth. That's this lesson.

Intrinsic value is your estimate of what a business is actually worth, built from the business itself: the cash it throws off, how fast that cash is likely to grow, and how much risk sits between you and it. Read that sentence again and notice the second word. Your estimate. Not the company's value. Not a fact you can look up. A number you build, out of assumptions you chose.

Figure

The estimate earns its keep by being compared to something: the stock's current market price. If your estimate lands above the market price, the stock looks undervalued — the market is asking less than you think the business is worth. If your estimate lands below the market price, it looks overvalued. That comparison is the entire point of the exercise. An intrinsic value estimate with no price next to it is just arithmetic.

Figure

The margin of safety is humility, not a discount

Knowing which side of the price your estimate falls on is not enough. You also want to know by how much, as a percentage. That gap has a name: the margin of safety — the distance between your estimate of intrinsic value and the price you actually pay.

Here's the arithmetic, using round invented numbers so the mechanics are visible. Suppose you work through a company and land on an intrinsic value of $40 a share. The market is charging $30. Your margin of safety is the $10 gap, divided by your $40 estimate: 25%.

Figure

Now the part that matters more than the formula. It is tempting to read that 25% as a bonus — a bargain, ten dollars of free upside sitting there waiting for you. That is the wrong way to hold it.

The margin of safety exists because your $40 is probably wrong. Every input that produced it was a guess about a future nobody has seen. The gap isn't profit you've located; it's room to be mistaken. If your $40 is really $34, a $30 purchase still works out. If you paid $39, it doesn't. The margin of safety is the humility built into the method — the acknowledgment, in numbers, that the method is unreliable.

Which is why the size of the gap should track how shaky your estimate is. A company with steady, boring, predictable earnings gives you inputs you can half-believe. A company whose sales and profits lurch around from year to year gives you inputs that are barely better than noise, and it should have to be much cheaper before you're interested. Less confidence, wider margin.

How a DCF works: project, then discount

The rest of this lesson builds one specific method for getting to that estimate: the discounted cash flow (DCF) model — estimating what a business is worth by projecting the cash it will produce and then converting that future cash into today's money.

Video coming soon

This lesson explains the idea in full without it.

It runs in two moves, and if you only remember two things about the DCF, remember these.

Move one: project. Take a current measure of the company's cash flow. Pick a rate you think it will grow at. Run it forward, year by year, for some number of years.

Figure

Move two: discount. Those projected dollars arrive years from now, and a dollar years from now is worth less than a dollar today — because you have to wait for it, and because it might never show up. So you shrink each future year's cash back down to what it's worth in today's money. How hard you shrink it depends on how risky the company is.

Figure

Everything else in this lesson is detail about those two moves. And before we get into the detail, learn the two directional relationships, because they are the intuition that keeps you from being fooled by your own spreadsheet:

• Growth rate up, value up. Assume the company grows faster, and every projected year gets bigger, and the estimate rises. Positive relationship.
• Discount rate up, value down. Decide the company is riskier, and every projected year gets cut back harder, and the estimate falls. Negative relationship.

Get those two solid before you touch any arithmetic. If a DCF ever spits out a number that moves the wrong way when you change an input, you've made a mistake — not a discovery.

The cash flow you project: EPS

A DCF needs a cash-flow number to start from. This course uses earnings per share (EPS) — a company's profit divided by the number of shares outstanding.

The reason to use EPS is that it's already scaled to your stake. You don't own a company; you own a share. If a company earns $2 billion and has one billion shares outstanding, EPS is $2. That $2 is what one share has a claim on. Projecting EPS forward projects what your share might claim later, which is the thing you're actually trying to value.

Be honest about the weaknesses, because they're real. EPS is an accounting output, not a bank balance, and accounting involves judgment calls — about when revenue counts, how assets wear out, what gets written off. Two companies with identical businesses can report different EPS. And some companies report negative EPS, sometimes for years. You cannot meaningfully grow a negative number forward at 8% a year, so a DCF built on EPS quietly stops working on exactly the companies whose futures are hardest to read. That's a limitation of the tool, and the tool doesn't warn you.

Growth rates: where most of your answer comes from

The growth rate is the rate at which you assume the company's earnings will grow over the projection period. It's common to project about five years out.

Here's the uncomfortable fact about the growth rate: it typically drives more of your final answer than any other input. Which means the single number you have the least ability to know is the number your estimate leans on hardest. There are three common ways to pick it, and they get progressively less naive.

Projection. Look at what earnings have actually grown at historically and draw a straight line forward. If a company has averaged 8% a year, assume 8% a year. This is the simplest approach and the easiest to defend, because at least it's anchored to something that happened. It also assumes the future resembles the past, which is exactly the assumption that fails.

Forecast. Start from the historical average, then adjust it for things you know that history doesn't. A new product that should lift earnings for the next few years argues for a rate above the historical average. A patent running out argues for less. This is more thoughtful than a projection and more dangerous, because now you're adding your own opinion to the number that dominates the model — and opinions about a company you've decided you like tend to point in one direction.

Industry average. Assume the company's growth drifts toward the average for its industry. The logic is competition: a business earning outsized growth attracts rivals, rivals compete the advantage away, and growth settles toward what everyone else in that industry manages. It's a deliberately unexciting assumption, which is roughly why it's useful — it's the one that resists your enthusiasm.

Figure

That convergence idea is worth carrying around beyond this model. A company growing much faster than its industry is, in most cases, temporarily doing so.

Terminal value: the part after the part you projected

You projected five years. But nobody buys a stock expecting the company to stop existing in year six. The price the market charges today reflects an expectation that the business keeps earning indefinitely — and "indefinitely" is not a number you can put in a spreadsheet.

The workaround is terminal value: a single estimated value standing in for everything past the end of your projection period. Instead of modeling infinity, you model five years explicitly and then say "and at the end of year five, the whole rest of the future is worth roughly this." One finite number replaces an infinite tail.

Figure

One common way to produce that number is the exit multiple — a price multiple applied to the last year of earnings you projected. Recall that a price multiple like the price-to-earnings (P/E) ratio is what buyers are paying today for each dollar of a company's earnings. Applied at the end of a projection, it becomes an assumption about what a buyer might pay for each dollar of earnings at that future point — hence "exit": it's the price you'd theoretically exit at.

Which multiple? This course uses the industry average P/E, for the same reason it leans on industry-average growth: if you believe a company's growth eventually converges toward its industry, you should believe the price people pay for that growth converges too.

Here's what nobody tells you about terminal value, so we will. In a five-year DCF, the terminal value is usually a large share of the total estimate — frequently more than half. So a model that looks like a careful year-by-year analysis of a business is, in practice, substantially one guess about a multiple somebody might pay five years from now. Keep that in view.

Discount rates and the time value of money

You have projected cash flows and a terminal value. Everything so far has been about upside. The discount rate is the reality check.

The discount rate is the rate that converts future money into today's value, and it reflects risk. It's sometimes called an expected rate of return, and both names describe the same thing from different ends: it's what you'd demand to be paid for tying up your money in this particular company rather than doing something else with it.

The idea underneath it is the time value of money: a dollar today is worth more than a dollar a year from now. Not because of a rule, but because waiting is not free. In a year, plenty can go wrong between you and that dollar — the company stumbles, the industry turns, the promise doesn't land. And the dollar you hold today could spend the year working somewhere else. The discount rate is an attempt to put a number on all of that.

Figure

The relationship runs opposite to growth, and it's worth saying both directions out loud:

• Higher discount rate → lower intrinsic value. You've judged the company riskier, so you're demanding more to hold it, so its future cash is worth less to you today.
• Lower discount rate → higher intrinsic value. Less risk, less compensation demanded, future cash worth more today.

What makes a discount rate high? Riskiness, broadly — and riskiness shows up in more than one way. A company might be fundamentally weaker than its competitors: thinner margins, more debt, a shakier position. Or its stock price might be more volatile, swinging harder day to day than the market around it. The next section takes that second kind of risk and turns it into a number.

CAPM: turning volatility into a discount rate

One standard way to produce a discount rate is the capital asset pricing model (CAPM) — a model that estimates the return an investment should have to offer, given how much market risk it carries. Its output, an expected rate of return, is what you plug into the DCF as your discount rate.

The formula:

E(Ri) = Rf + βi × (Rm − Rf)

Read it left to right and it's a story in two parts: start with what you could earn taking no risk, then add extra pay for the risk you're taking.

The risk-free rate (Rf) is the baseline — the return you could get without meaningfully risking your principal. Long-term investors commonly use the annualized yield on a 30-year Treasury, on the reasoning that a long-horizon investment deserves a long-horizon baseline. Whatever you use, it acts as a floor: no rational investor demands less than the risk-free rate to take on actual risk. Look the current yield up when you build the model — the Treasury publishes it daily. It moves constantly, and it is never the number you remember.

The market risk premium (Rm − Rf) is the second part: the extra return investors demand for holding stocks in general rather than sitting in the risk-free asset. It's the market's overall return minus the risk-free rate. The market return here means the return of a broad index like the S&P 500. This premium is not a published number you can look up — it's estimated, methods for estimating it disagree with each other, and reasonable people land in different places. So pick your estimate consciously and write down which method you used, because the last section of this lesson shows how far your answer travels when this one input moves.

Beta (βi) scales the premium to the individual stock. Beta measures how much a stock moves relative to the overall market. The benchmark index is defined as beta of 1.0 — it moves exactly with itself, by construction. A stock with beta above 1.0 has historically swung harder than the market, so CAPM enlarges the risk premium for it, raising the discount rate, lowering your estimate of its value. Below 1.0 and the reverse happens.

Beta deserves a warning the formula doesn't give it. Beta is calculated from past price movement — it's a description of history wearing the costume of a forecast. It changes depending on what window you measure it over and what benchmark you measure it against, so the "same" stock can have visibly different betas from two sources on the same day. And it treats volatility as risk, which means a stock that swings hard while the business quietly compounds gets penalized, and a stock that sits still on its way to insolvency doesn't. Beta is a convenient input, not a measurement of danger. Use it and don't believe it.

A worked CAPM example would need a risk-free rate, a market risk premium, and a beta — all three of which are live figures that would be wrong by the time you read them. The output is what matters conceptually: a discount rate customized to one company, meant to reflect what you should demand for carrying that company's particular risk.

Watch the answer move

Now the point of the whole lesson.

Take a fictional company, XYZ. Every number here is invented and round, chosen so you can follow the arithmetic — none of it is a real company, a real multiple, or a forecast.

• Current EPS: $2.00
• Assumed growth: 8% a year for five years
• Exit multiple: 15× the year-five EPS
• Market price: $30

Run it at an 8% discount rate and XYZ's intrinsic value comes out to $40.00 a share. Against a $30 price, that's the 25% margin of safety from the top of the lesson. Tidy. Convincing. Two decimal places.

Now change exactly one thing — the discount rate — and change nothing whatsoever about the business.

Video coming soon

This lesson explains the idea in full without it.

XYZ sells the same products to the same customers in every row. Nothing about it changed. But an eight-percentage-point difference of opinion about how risky it is — and eight points is well within the range that two careful analysts could disagree by — moves the answer from $43.52 to $31.42. At the top of the table you have a comfortable buy. At the bottom you have a 4% cushion, which is to say no cushion at all. The margin of safety didn't shrink because the company got worse. It shrank because you nudged an assumption.

Growth does the same thing. Hold the discount rate at 10% and move only the growth assumption:

Six percentage points of growth — the difference between "this company grows a bit" and "this company grows nicely," a distinction nobody can call in advance — swings the estimate by nearly a third.

And the loss case is the one to sit with. Suppose you'd chosen 6% and 11% because you liked the company. You'd have valued XYZ north of $45, bought at $30 feeling like a genius, and had a margin of safety that existed only in your own spreadsheet. The market price would not have cared.

So what is the model actually for, if you can't trust the number? Two things, both worth having. First, it forces you to write your assumptions down, which turns a vague feeling that a company is cheap into a specific claim — it grows at 8% and deserves a 15× multiple — that can be checked and can be proven wrong. Second, run across a range as we just did, it tells you what you'd have to believe for the stock to be worth buying. That's a more useful output than a point estimate, and it's more honest about what you know.

The number is not the deliverable. The range, and knowing where in it you're standing, is.

Key takeaways

• Intrinsic value is your estimate of what a business is worth, built from cash flow, growth, and risk. It's only useful next to the market price: estimate above price means the stock looks undervalued, below means overvalued.
• The margin of safety is the percentage gap between your estimate and what you pay. It exists because your estimate is unreliable — it's room to be wrong, not free upside. Shakier company, wider margin required.
• A DCF has two moves: project a cash-flow measure like EPS forward at an assumed growth rate, then discount those future dollars back to today at a rate reflecting risk. Higher growth rate raises the value; higher discount rate lowers it.
• Terminal value stands in for everything past your projection period, usually via an exit multiple. In a five-year model it's often more than half the total answer, so the "detailed analysis" rests heavily on one guess.
• The output looks precise and isn't. Changing a single assumption — the discount rate, the growth rate — swings the answer by a third without touching the business. The model's real value is forcing you to state what you'd have to believe, not handing you a price.