Valuation multiples — P/E, P/B, EV/EBITDA and more
The answer
A valuation multiple divides what you pay by something the company produces — profit, book value, sales or cash earnings. Each multiple asks a different question, and using the wrong one gives you a confident wrong answer.
The skill is not calculating them. It is knowing which multiple a business can be judged by, and which 1 number inside it can be manipulated.
Why this costs you money
You compare 2 companies in the same industry. One trades at 14 times earnings, the other at 31. You buy the cheaper one.
Six months later the cheaper one has fallen and the expensive one has risen. Nothing about your comparison was arithmetically wrong. It was structurally wrong, and here are the 3 ways that usually happens.
You compared across capital structures. The P/E ratio ignores debt. A company with no borrowings and a company financed half by borrowings can show the same P/E while carrying completely different risk. If you are comparing 2 businesses with different levels of debt, the P/E is not comparing the same thing twice. It is comparing 2 different things.
You used a multiple the business cannot support. Price to book is a strong tool for a bank and close to useless for a software company, because the software company's real assets — its code, its customers, its people — are not on the balance sheet. A software firm at 12 times book is not expensive. Book value simply is not where its value lives.
You trusted the denominator. EBITDA means earnings before interest, tax, depreciation and amortisation. It is not defined by any accounting standard . Companies calculate it themselves, and many of them also publish "adjusted EBITDA" with further items removed. Every removal makes the multiple look cheaper. You are not comparing 2 companies. You are comparing 2 marketing decisions.
The specific loss looks like this. A reader screens for companies under 8 times EV/EBITDA, buys 5 of them, and finds 2 years later that 3 were cyclical businesses at the top of a cycle and 1 had large lease obligations that never entered the debt figure. The screen was not broken. The screen was answering a question the reader had not asked.
How it works
Every multiple has the same shape.
What you pay ÷ what the company produces
The choices sit in both halves, and the top half has 2 versions that people constantly confuse.
Price means the market value of the equity alone. Share price multiplied by share count. It is what a buyer of the shares pays.
Enterprise value, written EV, means the cost of buying the whole business free of its financing. The standard calculation is market value of equity, plus total debt, plus minority interest and preference shares, minus cash and cash equivalents. If you bought every share of a company that holds ₹100 crore of cash and owes ₹300 crore, you would also inherit the debt and gain the cash. Enterprise value says so out loud.
That single distinction sorts the multiples into 2 families.
Equity multiples put price on top. They must therefore have a number on the bottom that belongs to shareholders only — profit after interest and tax, or book value of equity.
- P/E — price ÷ earnings per share. Covered in full in the P/E article. Use it for a profitable, reasonably stable business.
- PEG — P/E ÷ the expected annual growth rate in earnings, written as a plain number. A P/E of 30 with 30% expected growth gives a PEG of 1. The idea is that a fast grower deserves a higher multiple. The weakness is that the growth rate is a forecast, and PEG hides the forecast inside a single number so you stop questioning it.
- P/B — price ÷ book value per share. Book value is assets minus liabilities. This works where the balance sheet is the business, which mostly means lenders and insurers.
- Dividend yield — dividend per share ÷ price, as a percentage. This is a multiple turned upside down. It is the only one that measures cash actually paid to you.
Enterprise value multiples put EV on top. The bottom must therefore be a number produced before interest is paid, because interest belongs to the lenders who are included in EV.
- EV/EBITDA — the most used multiple in professional work. It strips out financing choices, tax rates and depreciation policy, so 2 companies in different countries with different debt loads become comparable.
- EV/EBIT — the same idea but with depreciation left in. Slower to say, harder to manipulate, usually more honest. Depreciation is a real cost. A factory wears out.
- EV/Sales — used when there is no profit yet. It makes almost no assumption, which is its strength and its weakness.
P/S, price ÷ sales, is the equity version of EV/Sales. It is common in screeners and inferior to EV/Sales, because sales are produced by the whole business including the part funded by debt.
Here is the rule that makes all of this usable. Match the top half to the bottom half. If the bottom number is calculated before interest, the top must be enterprise value. If the bottom is calculated after interest, the top must be price. P/EBITDA is a broken ratio and it appears in screeners more often than it should.
What it tells you, and what it does not
A multiple tells you what the market currently expects, expressed as a number. Nothing more. It is a price tag with the size divided out.
It does not tell you whether that expectation is correct. It does not tell you whether the company is good. And it does not travel between industries. A company that runs restaurants and a company that writes software will never carry the same EV/EBITDA, and neither is mispriced because of it.
Four traps break multiples reliably.
The cyclical trap. Steel, cement, sugar, chemicals, shipping, airlines. At the peak of a cycle profits are enormous, so every earnings multiple looks tiny. At the bottom profits vanish, so every multiple looks absurd or turns negative. For a cyclical business, a low multiple is a late-cycle warning and a high multiple is often an early-cycle opportunity. This is the exact opposite of the instinct.
The accounting trap. Book value depends on how assets were recorded and whether they have been written down. Goodwill — the amount paid above the fair value of assets in an acquisition — sits inside book value until somebody decides to impair it. A company that has made many acquisitions can carry a large book value that represents optimism, not assets.
The lease trap. Since the lease accounting standards changed, most operating leases appear on the balance sheet as a right-of-use asset and a lease liability. In India this is Ind AS 116, applicable from 1 April 2019; in the United States it is ASC 842, effective for public companies from financial years beginning after 15 December 2018. Before those dates, a retailer with 500 rented stores showed almost no debt. Comparing an EV/EBITDA figure from 2017 with one from 2021 is comparing 2 different definitions.
The adjustment trap. Adjusted EBITDA, adjusted earnings, cash earnings per share. Every adjustment is a management judgement about what is not representative. Some are fair. Restructuring costs that recur every year for 6 years are not one-off, whatever they are called.
The decision rule
Pick the multiple from the business, not from the screener. Then ask what the denominator would look like in a normal year.
Step 1. Ask what kind of business this is.
- Lender, bank or insurer → P/B read against return on equity. Never EV/EBITDA; a bank's debt is its raw material, so enterprise value is meaningless.
- Stable, profitable, moderately geared → P/E and EV/EBIT.
- Capital-heavy with large depreciation, or heavily indebted → EV/EBITDA and EV/EBIT together. If the 2 tell different stories, depreciation is the story.
- Cyclical → EV/EBITDA on mid-cycle profit, or price against replacement cost of assets. Never the reported trailing multiple at a profit record.
- Loss-making but growing → EV/Sales, and only as a placeholder until profit exists.
- Asset-light with recurring revenue → EV/Sales alongside gross margin and the rate of customer retention. Book value tells you nothing.
Step 2. Ask what the denominator looks like in a normal year. Take the 5-year average, or the median. If the current year is 60% above the 5-year average, you are looking at a peak, and the multiple is lying to you in the comfortable direction.
Step 3. Compare only 2 things. The company against its own multiple range over 7 to 10 years, and the company against direct competitors doing the same work. Comparing a multiple against the index is not a comparison.
Step 4. Say the assumption out loud. "At this multiple the market expects profits to roughly double in 5 years." Now you can disagree with a sentence rather than with a number.
Try this now
Five minutes, 1 holding you already own.
- Open your broker app or a financial data site and find your holding's market capitalisation. Write it down.
- Open the latest balance sheet — the annual report, or the results filing on the NSE, BSE or SEC website. Find total borrowings (long term plus short term, plus lease liabilities) and cash and cash equivalents plus current investments.
- Calculate enterprise value: market capitalisation + total borrowings − cash and investments.
- Find EBITDA. If the company does not report it, take operating profit from the income statement and add back depreciation and amortisation, which is a separate line. Divide enterprise value by EBITDA.
- Now compare your answer to the EV/EBITDA your screener or app displays for the same company.
What you should see. Three things.
Your number and the app's number will probably differ, sometimes by a lot. The usual reasons are lease liabilities, a different treatment of investments held in subsidiaries, and whether the app used adjusted EBITDA. Once you have seen that gap, you will never again quote a multiple from a screener as if it were a fact.
If your holding carries meaningful debt, enterprise value will be noticeably larger than market capitalisation. That difference is what a P/E ratio was hiding from you.
And if your holding is a bank or a non-banking finance company, step 3 will produce a strange, enormous number. That is correct. It is the article on valuing banks telling you that this tool does not belong there.
Three real cases
1. Valeant Pharmaceuticals, 2015 to 2016 (United States and Canada) — the adjusted denominator Valeant grew by buying other drug companies and cutting their research spending. It reported "cash earnings per share", which excluded the amortisation of the intangible assets created by those acquisitions. On that measure the shares looked reasonably priced. On reported earnings under generally accepted accounting principles they did not. The share price peaked near $262 in early August 2015. After questions about its relationship with the mail-order pharmacy Philidor, a delayed annual filing and a restatement, the shares traded below $30 by mid-2016 — a fall of more than 85%. The multiple that looked cheap was cheap only because of what had been removed from the bottom half.
2. Carillion plc, 2016 to January 2018 (United Kingdom) — the enterprise value that was wrong Carillion was a large construction and facilities management contractor. Its reported year-end net debt looked manageable, so its enterprise value looked modest and its EV/EBITDA looked ordinary. Two things were not in that figure. It used reverse factoring, a supply-chain finance arrangement under which a bank paid its suppliers early and Carillion repaid the bank later; this was not classified as borrowing. And its average net debt through the year was far higher than the figure reported on the last day of the year. It issued profit warnings from July 2017 and went into compulsory liquidation on 15 January 2018 with liabilities including a large pension deficit. If the top half of a multiple is understated, the multiple is understated.
3. Luckin Coffee, April 2020 (China, listed on Nasdaq) — the sales multiple with no sales Luckin listed on Nasdaq in May 2019 and grew store count faster than Starbucks had in China. It was loss-making, so investors valued it on a multiple of revenue. On 2 April 2020 the company announced that an internal investigation had found that roughly 2.2 billion renminbi of 2019 sales had been fabricated . The shares fell around 75% to 80% in a single session and the company was delisted from Nasdaq in June 2020. An EV/Sales multiple makes almost no assumption about profitability, which is why it is used for young companies. It makes 1 enormous assumption instead: that the sales are real.
The question that resolves it
A novice looks at 2 multiples and asks: which number is lower?
An expert looks at the same 2 and asks: is the denominator in each one measuring the same thing, in a normal year, for a comparable business?
Almost every serious mistake with multiples is a comparison mistake. Very few are arithmetic mistakes.
What would make this wrong
If multiples reliably identified good investments, buying the cheapest decile of the market on EV/EBITDA every year would beat the market every year. It does not, and there have been long periods — much of 2010 to 2020 in the United States — when cheap multiples underperformed expensive ones badly enough to end careers.
The honest limits are these. Multiples are relative tools; they tell you the price of this business against that one, never whether the whole group is sensibly priced. They break entirely for loss-making companies, for companies whose profit is near zero, and for anything at a cycle extreme. They are shaped by accounting policy, which differs by country and changes over time. And a multiple that has been low for 10 years is usually not a discovery. It is usually a description of the business.
If you find a company at half the multiple of its 3 closest competitors and you cannot explain why in 1 sentence, the most likely explanation is that you have misunderstood the business, not that 3 sets of analysts have.
In India
Indian listed companies file quarterly results with the exchanges under SEBI's listing regulations, generally within 45 days of the quarter end and 60 days for the fourth quarter. Accounts follow Indian Accounting Standards, known as Ind AS, which are largely converged with international standards.
Consolidated against standalone changes the answer, not the decimal. Indian groups commonly hold operating businesses inside subsidiaries. A holding company's standalone profit can be almost entirely dividends received from those subsidiaries. Using standalone figures for a group holding company will produce a multiple that is wrong by a multiple, not by a rounding error.
EBITDA is not a reported line in the Indian income statement format. Under Schedule III of the Companies Act, 2013, the profit and loss statement runs from revenue through expenses to profit before tax. There is no EBITDA subtotal. Every EBITDA figure you see for an Indian company was assembled by somebody, and different sources assemble it differently — particularly on whether "other income" is included.
Related-party complexity reduces the value of enterprise value. Where a listed company transacts with unlisted entities controlled by the same promoter family, part of the economics may sit outside the entity whose EV you calculated. Enterprise value assumes the enterprise you are buying is the enterprise that produces the cash.
Price to book has a specific job here. For Indian banks and non-banking finance companies it is the primary multiple, read against return on equity. For most other Indian companies it is a weak tool, because asset values on the balance sheet may reflect historical cost from decades ago.
In the United States
US companies file form 10-Q quarterly and form 10-K annually with the Securities and Exchange Commission, under US Generally Accepted Accounting Principles.
Non-GAAP measures are regulated but permitted. Regulation G requires a company presenting a non-GAAP figure to also present the most directly comparable GAAP figure and reconcile the 2. That reconciliation table is one of the most useful pages in any US filing. It lists, line by line, exactly what management removed to get to the flattering number. Read the table, not the headline.
Stock-based compensation is the largest single adjustment at many technology companies. Shares issued to employees as pay are a genuine cost to existing shareholders because they increase the share count. Adjusted EBITDA that adds back stock-based compensation will always produce a cheaper-looking multiple than the economics deserve.
Segment disclosure is far richer. US filers must report by operating segment, with revenue, profit and assets for each. That allows a sum-of-the-parts valuation: value each segment at the multiple appropriate to its industry, add them, subtract net debt. For a conglomerate this is often the only honest approach.
Buybacks change the arithmetic. Large US companies buy back shares continuously, which reduces the share count and lifts earnings per share even when total profit is flat. Any per-share multiple compared across 5 years must be read alongside the share count over the same 5 years.
Where they differ, and what that tells you
EV/EBITDA is a weaker tool in India than in the United States, for 2 concrete reasons.
First, there is no reported EBITDA line in the Indian statutory format, so the number is always reconstructed and sources disagree. Second, where a listed company sits inside a promoter-controlled group with significant related-party dealings, enterprise value may not capture the whole enterprise. Cash may be lent to a related entity. Revenue may be booked at a price set within the group. Enterprise value is a clean concept that assumes clean boundaries.
The practical consequence: in India, weight EV/EBIT and cash-flow-based checks more heavily, and always read the related-party transactions note before trusting an enterprise value. In the United States, where segment detail is richer and non-GAAP reconciliations are mandatory, EV/EBITDA does more work — but read the reconciliation table first.
Segment detail differs sharply. A US conglomerate can usually be valued piece by piece from its own filing. An Indian conglomerate often cannot, because segment reporting is thinner. That pushes Indian analysis back toward the consolidated whole, which means you need a stronger view on the group's weakest business, not just its best one.
Buybacks have historically mattered less to Indian per-share figures. Indian companies have used buybacks far less than US companies, and for most of the last decade an Indian buyback carried a distribution tax paid by the company under section 115QA. From 1 October 2024 buyback proceeds became taxable in the shareholder's hands as a deemed dividend instead. The habit that follows is asymmetric: for a US holding, check the share count over 5 years before believing per-share growth. For an Indian holding, check for the opposite — preferential allotments, convertible instruments and employee stock options that increase the count.
Carry this
- Match the halves. Enterprise value goes with numbers before interest. Price goes with numbers after interest.
- Pick the multiple from the business. Banks get P/B. Capital-heavy businesses get EV/EBIT. Loss-makers get EV/Sales and a warning.
- Before believing any multiple, ask what the denominator looks like in a normal year, and what management removed to produce it.