The valuation spread

Put five companies that do the same thing side by side - five listed makers of the same product, or five lenders, or five consumer brands - and read the multiple the market pays for each. They will not be the same. One trades at twenty-five times earnings, another at twelve; one at three times book, another at one. That range is the valuation spread, and it is the single most information-dense thing a peer set has to tell you, because it is not noise and it is not an accident. Every gap in that spread is a judgement the market has already made and priced: this business deserves to be paid up for, that one does not. illustrative

The mistake almost everyone makes is to treat the spread as either a menu or a verdict. As a menu, you scan for the lowest number and call it the bargain. As a verdict, you assume the market is right and pay the premium the leader commands without asking why. Both skip the only work that matters. The spread is neither a menu nor a verdict - it is a set of claims, one per peer, each of the form "this is worth more (or less) than that, by this much." And a claim is a thing you interrogate. For every premium in the set you ask: what fundamental justifies it - faster growth, higher returns on capital, a more durable moat, cleaner governance, less capital hunger? For every discount you ask the mirror question: is this cheapness deserved, or has the market mispriced a business whose fundamentals do not warrant the markdown?

This module is about reading that spread the way a forensic reader reads a set of accounts - assuming nothing, tying every number to the fundamental that should move with it, and treating the gap between the multiple and its justification as the object of interest. A peer priced at a premium it has earned is not an opportunity and not a warning; it is simply correct, and you move on. A peer priced at a discount its fundamentals do not explain is the one thing in the whole set worth more of your time. Learning to tell those apart - and to notice when the entire set has been re-rated together so that relative cheapness hides absolute expense - is what turns a screen full of ratios into a judgement.

A spread is only as good as the set it is drawn from, so the first discipline is normalisation: the peers must be genuinely comparable and the multiples computed on a like-for-like basis, or the ranking is comparing things that were never the same. Three things have to be lined up before the numbers mean anything.

First, the set itself. Peers must share the economics you are trying to price - the same product, the same customer, the same capital structure of the business, broadly the same stage. A specialty chemicals maker and a commodity chemicals maker both sit under "chemicals" and belong in different sets; a business growing at 8% and one growing at 30% are not peers even in the same industry, because the multiple that is fair for one is absurd for the other. A loose set produces a spread full of false gaps.

Second, the multiples, and each asks a different question. Price-to-earnings - price over per-share profit - asks what the market pays for a rupee of the bottom line, and it is the most quoted and the most easily distorted, because profit after tax sits below depreciation, interest and tax, all of which differ across companies and can be managed. EV/EBITDA - enterprise value over operating profit before those items - strips out capital structure and the accounting of assets, which is why it is preferred for capital-heavy and differently-leveraged businesses. Price-to-book - price over accounting net worth - asks what the market pays for a rupee of book value, and it comes into its own where the balance sheet is the business. And EV/sales - enterprise value over revenue - is the multiple of last resort, used when there is no profit yet to divide by, as in an early-growth business deliberately spending its margin away. You rank the set on more than one, because a business can look cheap on one multiple and dear on another, and the disagreement between them is itself a signal.

Third, the normalisation of the numbers you feed in. A multiple computed on a cyclical's peak-year earnings looks deceptively low precisely when the stock is most dangerous, and high or meaningless at the trough when it may be cheapest - so cyclicals are ranked on mid-cycle or normalised earnings, never the reported figure of a single good year. One-off gains, a low tax year, a merger that inflated book value with goodwill - each has to be pulled out before the peer's multiple is honest. The point of all this care is not precision for its own sake; it is that a spread built on un-normalised numbers manufactures gaps that are artefacts of accounting rather than judgements about business, and you will spend your interrogation explaining differences that were never real.

Only once the set is clean, the multiple fits the business, and the inputs are normalised does the spread become what it should be: a row of comparable market judgements, ready to be matched against the fundamentals that are supposed to justify them.

Here is the core move. Set the spread beside the fundamentals that are supposed to explain it, and read across. Where a peer's premium is matched by superior fundamentals, the market has ranked it correctly and there is nothing to do. Where a premium sits on top of ordinary fundamentals, the market may be paying for a story. And where a discount sits on top of strong fundamentals, you have found the one genuinely interesting cell in the table. illustrative

Read the table as a set of tests, not a ranking to memorise. Peer A's premium is the easy case: the highest ROCE and growth in the set, a moat you can point to, governance you can trust - the market is paying for quality that exists, and paying up for real quality is not a mistake. Peer C's discount is the other easy case: the lower multiple is the correct price of slower growth and weaker returns, and a reader who "buys the cheap one" here has simply bought a worse business at the price a worse business deserves. The two cases that carry information are D and E, and they are mirror images. E is a premium the fundamentals do not earn - a middling business priced like a great one, which is where the market's story has run ahead of its returns. D is the reverse and the prize: returns and growth close to the leader's, but a multiple down among the laggards. That is the shape of a possible mispricing.

The scatter makes the same point geometrically. Plot each peer's multiple against the fundamental that should justify it, draw the line the market would price if it rewarded quality consistently, and the read stops being the height of any dot and becomes its distance from the line.

The fundamentals worth lining the spread against are always the same short list, because they are the things a multiple is supposed to capture: growth (faster-growing earnings deserve a higher multiple), returns on capital (a business compounding at high ROCE is worth more per rupee of profit than one that must reinvest heavily to stand still), moat durability (a premium is only safe if the returns will last), capital intensity (a business that swallows capital to grow converts less of its profit to cash and deserves less for each rupee of it), and governance (a promoter you cannot trust caps the multiple however good the numbers look, and often should). When a premium or a discount is matched by the right fundamental, the market has done your work for you. When it is not, the mismatch is the whole return on the exercise.

Every gap in the spread resolves, on inspection, into one of three verdicts, and keeping them distinct is what stops relative valuation from misleading you.

The first is the deserved discount. The cheap peer is cheap for a reason the fundamentals confirm - slower growth, structurally lower returns, more capital hunger, a governance flag, a weaker balance sheet. Here the market is right, and the low multiple is not an opportunity but an accurate price. Most low multiples in most peer sets are this. A reader who cannot bring himself to conclude "cheap and deserves to be" will spend his life buying the worst business in every set.

The second is the genuine mispricing - peer D. The discount is real but the fundamentals do not explain it: comparable returns and growth to better-rated peers, sound accounts, yet a multiple among the laggards. This is rare, and it is the only case where the spread points to a real opportunity rather than merely describing what the market thinks. Even here the discount is a candidate, not a conclusion: a market that prices a good business cheaply usually knows something - a pending regulatory overhang, a customer concentration, a succession vacuum - that the headline fundamentals miss, and the work is to find what that something is and decide whether it justifies the gap after all. Only a discount that survives that search is a mispricing worth acting on. This is where a genuine margin of safety can live, and equally where a value trap hides wearing the same clothes.

The third verdict is the one relative valuation is structurally blind to: the whole set is expensive together. Rank five peers against each other and one will always be the cheapest - but if all five trade well above their own long-run multiples, "cheapest of the set" is still expensive in absolute terms. A pure relative reading cannot see this, because it only ever compares peers to each other, never to what the businesses are actually worth. The defence is to anchor each peer to its own history and to some absolute sense of value, not only to the set. When every name in a beloved sector has been re-rated together, the spread will still helpfully tell you who is least loved - and that reading will walk you into an overvalued stock with complete confidence.

There is no universal "right multiple." The one that carries the signal is set by the economics of the business, and a multiple that is central in one sector is meaningless or misleading in another. This is the inversion at the heart of relative value: run the same instruction - "rank the peers on their multiple" - across sectors and you must reach for a different multiple each time, because the line that means the most is a different line in each business.

The inversion is sharpest at the lender, and it is worth dwelling on because it catches readers trained on ordinary companies. Tell someone to rank a set of banks and, if they learned valuation on consumer names, they reach for P/E - and a bank will happily show them a low one. But a lender's profit is a residual left after the single most discretionary line in its accounts, the provision, so its reported earnings are close to an opinion, and a P/E built on them ranks opinions. The multiple that carries the real signal is price-to-book, and even that is only as honest as the book, which is why credit cost and provision coverage against the non-performing asset stock have to be read alongside it. A capital-heavy cyclical inverts differently: its danger is not the choice of multiple but the earnings you feed it, because EV/EBITDA on a peak-cycle number is a screaming buy at exactly the wrong moment. The consumer business inverts toward cash conversion - the same P/E is worth more where more of the profit is real cash - and the early-growth business inverts all the way to revenue, because it has chosen to have no profit at all. A reader who carries one sector's multiple into another will rank the wrong line and come away confident and wrong: challenging a bank on its P/E while ignoring its provisioning, or calling a cyclical cheap on the very earnings the cycle is about to take away.

You will not build the raw spread from scratch - the multiples for a set of listed peers are on every screener, and pulling P/E, EV/EBITDA and P/B for five comparable companies takes minutes. The skill this module teaches begins after that, and it is mostly refusal: refusing to accept the screener's numbers as normalised, and refusing to read any multiple without its column of fundamentals beside it. illustrative

Start by fixing the set. The screener's "peers" list is generous and often wrong - it groups by broad industry, not by economics - so trim it to businesses that genuinely share a customer, a product and a stage, and throw out the ones that merely share a sector heading. Then normalise: for a cyclical, replace the reported EBITDA with a mid-cycle estimate from several years of the annual report; strip a one-off gain or a merger's goodwill out of the book; note where a low tax year has flattered a P/E. This is where the annual report earns its place - the screener gives you the multiple, but only the filings give you the normalised number underneath it and the fundamentals to rank it against. None of this touches any judgement about whether a real company is well or badly run; it is the structural work of making five numbers comparable.

Now build the one table that matters: the multiples in the left columns, and to their right the fundamentals - three-year growth, ROCE, a sense of moat durability, capital intensity read from the capex-to-cash-flow relationship, and a governance read from the related-party and pledge disclosures. Read across each row, not down each column. The peer whose premium is matched by superior fundamentals needs no more thought. The peer whose discount is matched by weaker fundamentals is correctly cheap. And the row where the multiple and the fundamentals disagree - a low multiple beside strong fundamentals, or a high one beside ordinary numbers - is the only row that has told you something the market has not already priced. That row is where the next module's work, deciding whether the discount is a margin of safety or a trap, actually begins.

The spread locates the mispricing candidate; it does not confirm it. A discount the headline fundamentals do not explain is a prompt to look harder, and the market is right often enough that the base case for any unexplained discount is that the market knows something you have not yet found - a contingent liability, a customer about to leave, a regulatory change, a promoter problem. Reconciliation applies here exactly as it does to accounts: the gap is the question, never the verdict, and a discount that survives a genuine search for its cause is worth acting on precisely because most do not.

Nor can relative value tell you the set is worth owning. It is silent, by construction, on the level - it ranks peers against each other and never against absolute worth, so it will hand you the cheapest name in a sector that is expensive top to bottom and call it a buy. The multiple is a relative instrument, and a reader who lets it substitute for an absolute anchor - the business's own history, its through-cycle cash generation, what a private buyer would pay for the whole company - has mistaken a scalpel for a compass.

And the spread cannot make a business comparable that is not. Force two businesses with different economics into one set and every gap in the resulting spread is an artefact - a fast grower will always look "expensive" beside a slow one, an asset-light firm always "dear" on book beside an asset-heavy one, and reading those manufactured gaps as judgements is worse than not looking at all, because it comes dressed as rigour. The quality of the spread is capped by the quality of the set, and no amount of careful ranking rescues a set that was never comparable.

The first and most common way is buying the lowest multiple. The cheapest name in the set is the one the reader's eye goes to, and the reflex is to read low as opportunity. But the lowest multiple is the market's strongest statement that a business is worst, and it is right about that far more often than it is wrong. Buying the cheapest peer without asking whether the cheapness is deserved is the classic route into a value trap - a business that is cheap because it is quietly deteriorating, and that gets cheaper for exactly as long as you hold it. The discipline is to treat every low multiple as a deserved discount until the fundamentals prove otherwise, not the reverse.

The second is comparing across sectors, or across non-comparable businesses within one. A P/E of 15 is not "cheaper" than a P/E of 30 when the first is a slow cyclical and the second a fast compounder; a bank at 1.5x book is not "dearer" than a software firm at 8x, because the two multiples are measuring different things about different economics. The number strips away the context that made it meaningful, and a reader who ranks multiples without ranking the businesses behind them is comparing prices of things that were never the same good.

The third is anchoring to the peer average and forgetting the set can be wrong together. "Trades below the sector average" feels like a reason, but the sector average is just the middle of a spread that can be entirely too high. In a re-rated, crowded, beloved sector, below-average is still expensive, and the reader who buys "the cheap one relative to peers" has anchored to a number that carries no information about absolute worth. The average is a description of the crowd's mood, not a valuation, and treating it as a valuation is how disciplined-feeling readers buy the top of a whole sector at once.