There is no universally best multiple
A multiple compresses a business and a price into one number. Compression is useful only when the omitted details are understood. The right starting multiple is the one whose denominator is meaningful and reasonably repeatable for the business in front of you.
A profitable consumer business can start with P/E. A bank's equity capital and return on it can make P/B more coherent. A pre-profit business may require P/S or EV/Sales. A non-financial operator with material debt is usually clearer on EV/EBITDA than on an equity-only ratio. An established distributor of cash may justify a dividend-yield view.
Those are starting points, not sector laws. A company's economics, accounting and present stage can make a usual sector shortcut unsuitable. Name why the denominator represents the business before calculating anything.
Step one: match the claimholder
Ask whose price sits in the numerator. Price, market capitalisation, P/E and P/B are equity measures, and their denominators belong to ordinary shareholders. P/S also starts with an equity numerator, but revenue is produced before lenders are paid; that mismatch is why EV/Sales is usually cleaner when leverage differs.
Enterprise value describes the operating business financed by lenders and shareholders together. Pair it with measures before interest allocates the result between them: EBITDA or revenue. Pair earnings after interest with equity value.
This single check prevents common errors such as P/EBITDA or EV/EPS. It also tells you when debt cannot be treated as detachable financing. For a bank, deposits and interest are operating inputs, so conventional EV/EBITDA is the wrong language.
Start with the business, then choose the lens
Select a business. Each route names the first multiple, the variable that explains it and the check aimed at what it misses.
The bank and infrastructure cases are unnamed hypotheticals. The routes demonstrate fit; they are not automatic rules for every company in those industries.
Step two: pair every multiple with its driver
A ratio is explained by business fundamentals, not by its label. P/E travels with growth, payout, risk and returns on new investment. P/B travels with return on equity and the credibility of book value. P/S travels with the margin and cash path.
EV/EBITDA travels with operating growth, capital intensity, cash conversion and risk. EV/Sales travels with the operating margin it may eventually earn. Dividend yield travels with payout sustainability, growth and the return earned on cash retained inside.
The companion variable turns a screen into a question. Two companies at 2 times sales are not alike if one can sustain a 20% margin and the other loses money on every order. Two banks at 1.5 times book are not alike if one earns 18% on sound assets and the other earns 7% before a new provision.
- Profitable non-financial business
- P/E → normalised earnings, growth, returns, debt and cash conversion
- Bank or similar lender
- P/B → sustainable ROE, asset quality, provisions and capital
- Pre-profit, revenue-producing business
- P/S or EV/Sales → margin path, unit economics, cash burn and funding
- Debt-heavy non-financial operator
- EV/EBITDA → maintenance capex, working capital, leverage and cash
- Cyclical producer
- P/E or EV/EBITDA on normalised mid-cycle results → cycle position, commodity sensitivity and balance-sheet survival
- Mature established payer
- Dividend yield → cash coverage, payout durability, retained cash and total return
Step three: keep the comparison honest
Use the same definition, period and accounting treatment. Trailing P/E is not directly comparable with forward P/E. EBITDA before leases is not the same as EBITDA after leases. Gross marketplace billings are not the same as revenue recognised after paying sellers.
Compare companies with similar economics, not merely the same industry label. Growth, margins, returns, reinvestment, debt, accounting and risk explain why a multiple deserves to differ. If those drivers differ, a lower number is not evidence of a cheaper business.
Then reconcile the chosen denominator towards owner cash. That route exposes the main costs the multiple omitted. If the conclusion disappears after normalising one-off earnings, adding debt or allowing for required capex, the original ratio never carried it.
What no multiple can answer
A multiple tells you what is being paid per unit of a chosen accounting or cash measure. It does not prove the unit is durable. It does not forecast the business, select the correct peers or show that the current market price is below intrinsic value.
One multiple can support two opposite stories. A low P/E can reflect a bargain or peak cyclical profit. A high P/B can reflect productive capital or an unjustified premium. A high yield can reflect dependable cash or an expected cut. The number states the relationship; the business evidence chooses between the stories.
Use at least one check aimed at the primary ratio's blind spot. That check should be independent enough to fail: sales need margins, EBITDA needs cash after capex, book needs returns and asset quality, yield needs coverage, and earnings need reinvestment and balance-sheet context.