P/E, P/S, EV/Sales, EV/EBITDA: When Each Multiple Matters

There is no such thing as “the” valuation multiple. Each one divides price by a different layer of the income statement, and each layer tells the truth about some companies and lies about others. This guide covers the four multiples StockResearch tracks (trailing P/E, P/S, EV/Sales, EV/EBITDA) plus forward P/E, and when each one is the right tool.

The map: two choices, one grid

Every multiple is built from two decisions. Numerator: do you use market cap (just the equity) or enterprise value (equity plus net debt: the price of the whole business, debts included)? Denominator: do you divide by revenue, EBITDA, or net earnings; that is, how far down the income statement do you go before measuring? Going further down captures more of the business’s real economics but breaks on more companies, because each subtraction (costs, depreciation, interest, tax) is another chance for the number to go negative or get distorted.

Trailing P/E: the default, and its blind spots

Price divided by the last twelve months of diluted earnings per share. It is the most information-dense multiple: earnings sit after every cost, so P/E prices what shareholders actually keep. When a company is stably profitable, P/E is the multiple to rank; its valuation percentile is usually the most meaningful one on the page.

Its blind spots are exactly its strengths inverted:

• Negative or near-zero earnings make it meaningless. A company earning $0.02 per share trades at a 4,000× P/E; one losing $0.02 has no P/E at all. Neither number says anything about value.
• One-time items pollute it. A big legal settlement, an asset sale, a tax release: trailing EPS swallows them all, and the multiple lurches without the business changing.
• It ignores the balance sheet. Two companies with identical earnings and market caps have the same P/E even if one carries heavy debt. The equity is not equally risky, but P/E can’t see that.

Forward P/E: useful context, not rankable history

Forward P/E swaps trailing earnings for the consensus estimate of next year’s. It answers a genuinely different question (“what am I paying for what analysts think is coming?”), and for cyclical or fast-changing businesses it is often closer to how the market actually prices the stock. But estimates get revised constantly and have no stable history, so there is no honest way to compute a five-year percentile on forward P/E. StockResearch shows it as a labeled context stat next to the ranked trailing multiples, and never mixes it into percentile math. A wide gap between trailing and forward P/E is itself information: the market expects earnings to change a lot.

P/S and EV/Sales: for when earnings can’t testify

Revenue is the hardest line to fake and the last to go negative, which makes sales multiples the workhorse for unprofitable or barely profitable companies: young growth businesses, cyclicals at the bottom of their cycle, companies mid-turnaround. When P/E is undefined or absurd, P/S still produces a clean, rankable series.

The cost is that sales multiples are margin-blind. A dollar of revenue at a 75%-gross-margin software company and a dollar at a 10%-margin distributor are priced as if equal. This matters most across companies (another reason self-relative ranking beats cross-sectional), but it also matters across time: if a company’s margins have structurally improved over five years, its P/S should sit high in its historical range, and a mid-range percentile may actually be conservative.

EV/Sales is P/S with the balance sheet added back: enterprise value (market cap plus net debt, minus cash) over revenue. For companies with meaningful debt or huge cash piles, EV/Sales is the more honest of the two. If a company funds itself with debt, P/S flatters it; EV/Sales doesn’t.

EV/EBITDA: the operator’s multiple

Enterprise value over earnings before interest, taxes, depreciation, and amortization. EBITDA approximates operating cash generation before financing and accounting choices, so EV/EBITDA compares businesses independent of how they’re funded and how aggressively they depreciate. It is the standard multiple in private markets and for capital-structure-heavy sectors (telecom, industrials, energy, real assets), precisely because those are the places where debt loads differ most.

Two big failure modes:

• When depreciation is a real cost, EBITDA is fiction. A semiconductor manufacturer’s fabs, an airline’s planes, a telco’s network: these assets genuinely wear out and must genuinely be replaced. Adding depreciation back treats the largest cost of the business as if it didn’t exist. For capital-intensive companies, read EV/EBITDA alongside P/E, never instead of it.
• Banks and insurers: none of the above. For financial companies, debt isn’t financing; it’s inventory. Deposits and float are the raw material of the business, so “enterprise value” and “EBITDA” stop meaning anything. Financials are priced on P/E and price-to-book; treat any EV-based multiple on a bank as a bug, not a signal.

Using them together

The multiples do their best work in combination, because divergences between them localize what changed. Each pair is a diagnostic:

• P/E down, P/S flat → margins expanded. The company is earning more per revenue dollar, and the market hasn’t re-rated the revenue.
• P/S down, P/E flat → margins compressed. Revenue got cheaper but earnings didn’t; the “cheapness” is an income-statement illusion.
• P/S low percentile, EV/Sales mid percentile → the balance sheet deteriorated. Debt grew while the equity got cheaper.
• Trailing P/E low, forward P/E high → the market expects earnings to fall. The trailing percentile is describing a past that consensus thinks is over.

This is why the screener shows percentile ranks for every multiple side by side rather than picking one: a stock that is at a low percentile on all four is a much cleaner signal than one that is only cheap on the multiple most flattering to it.