Valuation and what a price implies.

Enterprise value

What the whole business is worth, debt included. Market capitalization plus net debt.

Equity is a residual claim. It is what remains of the business after lenders have been paid, which is why enterprise value is the honest measure when comparing two companies financed differently.

It also explains why levered equity moves so violently. If enterprise value falls 30% while the debt stays fixed, the entire decline is absorbed by a smaller slice: shares worth $8.00 become $4.40. The business fell by less than a third. The equity fell by far more.

Worked through: leverage is a calendar, not a ratio.

Normalized earnings

What a business earns in an average year, with one-off items and cycle extremes stripped out. The honest denominator.

Reported earnings are an output of wherever a company happens to sit in its cycle, so any multiple computed off them is a measurement of that moment rather than of value. Normalizing means asking what margin this business earns across a full cycle, which usually takes ten years of history and about twenty minutes.

It changes answers rather than decorating them. A stock trading at a fair 15 times normal earnings shows 7.0 times at the peak and 70.6 times at the trough, looking like a bargain exactly when it is most dangerous.

The full cycle, priced: a low multiple is not a cheap stock.

Reverse DCF

Running a valuation backwards: instead of asking what a company is worth, ask what the current price already assumes.

A normal discounted cash flow requires you to forecast, which means the answer is mostly a reflection of your own assumptions. Reversing it removes that problem. Take the price as given and solve for the growth and margins required to justify it.

That converts an unanswerable question into a checkable one. You no longer have to predict the future, only judge whether the future the price implies is plausible, and how often this company has managed anything like it before.

How to run one: ask what the price assumes, and the DCF model.

Dilution

Your claim on a business shrinking because the company issued more shares.

Earnings per share can fall while profit rises, if the share count rises faster. Dilution is usually gradual, arriving through stock-based compensation, and is easy to miss because no single quarter looks dramatic.

It is most damaging when a company is forced to issue at a bad price. Raising $150.0M at $4.40 rather than $8.00 means issuing 34.1M shares, diluting existing holders 25.4%. The business can recover completely and the claim on it never does.

The gradual version: non-cash is not free. The forced version: leverage is a calendar.

Margin of safety

The gap between what you pay and what you think it is worth. Protection against being wrong.

Every valuation rests on assumptions that could be mistaken, so the discount is not a bonus. It is the allowance for error, and it is the only defense available against the fact that you cannot check your reasoning until afterwards.

It also has to be sized against how wrong you could be. A stable business with predictable cash flows needs less of one than a cyclical whose earnings can move ten times, or a levered company where a refinancing date decides the outcome.

What the price has to assume before you can size it: ask what the price assumes.

Terminal value

The single number standing in for everything after the explicit forecast ends. Usually most of the answer.

A discounted cash flow forecasts cash flows for some number of years, then needs a way to value the business continuing beyond that. Terminal value is that placeholder, most often computed as a perpetuity growing at a chosen rate and discounted back.

The uncomfortable part is how much weight it carries. On a ten year forecast it is typically around 60.2% of the total value, and on a five year forecast nearer 75.6%. The decade of careful modeling everyone argues about is the minority of the answer.

It is also the most sensitive part. Across a plausible range of perpetual growth rates and discount rates, the same company values from $1,899M to $3,300M, a spread of 74%. A useful sanity check is to divide terminal value by final year cash flow and ask whether the implied perpetual multiple is one you would actually pay.

The full argument, including why this is a criticism of our own primary method: most of a discounted cash flow is a number you made up.

Earnings yield

The price to earnings ratio turned upside down. Earnings divided by price, expressed as a percentage.

A stock at 20 times earnings has an earnings yield of 5%. The information is identical, but the framing changes what you compare it against. A multiple invites comparison with other multiples. A yield invites comparison with a bond, a savings account, or the return you need.

That comparison is the point. When government bonds yield more than the earnings yield on equities, you are being asked to accept more risk for less return, and the question of what you are being paid for that risk becomes concrete rather than theoretical.

The same warning applies as to any multiple. The denominator is an output of the cycle and of whatever one-time items landed in the period, so an earnings yield computed on flattered earnings is flattered too. Normalize first.

Why the denominator lies: a low multiple is not a cheap stock.

Cost of equity

The return a shareholder requires for owning this rather than something safe. The discount rate for equity cash flows, and the most consequential assumption in any valuation.

It is not a cost the company pays in cash. It is the hurdle: the return below which an equity holder would rather own a Treasury. Because it goes in the denominator of everything, a change of one percentage point moves a fair value estimate more than almost any operating assumption.

The textbook route is the capital asset pricing model: the risk-free rate plus beta times an equity risk premium. For McDonald's, a beta of 0.41 produces 6.05% that way. We used 8% instead, and the reason is worth stating: a low beta reflects how the share price has behaved, not how much could go wrong with the business, and accepting a six percent hurdle on an equity because the stock has been calm is how a model talks itself into a high valuation.

So treat it as a judgment, state it plainly, and test it. Every report on this site publishes the cost of equity it used and shows what the answer becomes at other values, because a fair value quoted without its discount rate is not a claim that can be checked.

Where it does the most work: justified price to tangible book, and the general case in most of a DCF is made up.

Sum of the parts

Valuing each business inside a company separately and adding them up. The right method when the pieces deserve different multiples, and an easy one to abuse.

A company running a mature cash machine and a loss-making growth venture has no single sensible multiple, because the blended earnings describe neither business. Valuing each on its own terms and adding the results is more honest than averaging.

It is also how conglomerate stories get told, and the abuse is structural: every segment gets the multiple of the best comparable anyone can find, the loss-making one is valued on revenue because it has no earnings, and the corporate costs that are real and shared go unallocated. The result is reliably higher than the market price, which is then described as a discount rather than as a disagreement.

Three disciplines make it usable. Allocate every dollar of central cost to a segment. Apply the same skepticism to the exciting segment as to the dull one, which usually means valuing it on what it earns rather than on what it might. And subtract net debt once, at the end, from the total.

Segment-level reading in practice: Uber and SoFi.