Share price divided by earnings per share. A multiple of 20 means paying twenty years of current profit. There's also a forward multiple based on next year's earnings forecast — lower for a growing company, but it rests on analyst forecasts that are routinely wrong, especially at turning points.

40 can be cheap for a company growing 40% with rising margins, and 8 can be expensive for one whose revenue is shrinking. A proper comparison is always against two things: the stock's own history over the past five years, and genuinely comparable companies in the same industry.

An unprofitable company: a revenue multiple, because there's no profit to divide by. Companies with different debt loads: enterprise-value multiples, which neutralise capital structure and compare like with like. Banks and insurers: a price-to-book multiple, because their assets are money and the balance sheet is the business. Mature companies with steady cash: free-cash-flow yield — how much cash you receive relative to the price.

The rule that follows: a multiple suited to one kind of company is meaningless for another. Price-to-book on a software company is a number with no content, and a P/E on a company that just turned profitable jumps into the hundreds and teaches you nothing.

The theoretically correct method: a company's value is all the cash it will produce in the future, discounted to today. The catch is that it's wildly sensitive to assumptions — shifting the long-run growth rate by two points can move the answer by tens of percent. So use it as a thinking tool that forces explicit assumptions, not as a calculator that returns a "fair price".

The best practical use is inverted: instead of computing the value, ask what growth the current price already assumes. If the answer is "another twenty years at 30% a year", you know exactly what you're buying.