Relative & Peer Valuation

Every valuation multiple is a ratio, and a ratio is a comparison waiting for its second half. When someone says a company trades at 28 times earnings, they have told you an arithmetic fact and nothing else. The number acquires meaning only when it is placed against a reference, and which reference you pick changes the conclusion completely.

Twenty-eight times might be well above what similar businesses in the same industry command, and well below what this same company itself commanded three years ago, and roughly in line with the broad market. All three statements can be true at once, and each one points somewhere different. This is not a flaw to be engineered away — it is the actual nature of relative valuation. You are not measuring value. You are measuring a difference, and a difference needs two well-chosen points.

Which is why the hardest part of comparable-company analysis is never the division. Anyone can divide a price by an earnings figure. The work, and the place where nearly all the errors live, is in deciding which companies belong in the comparison and which do not — and then in reading the resulting gap as a question to investigate rather than a verdict to act on.

There are exactly three defensible things to compare a multiple against, and each answers a different question. Using one and reporting the conclusion as though it came from another is the most common way a peer table misleads.

Sector peers answer: is the market pricing this business differently from others that face the same demand, the same input costs and the same regulatory environment? This is the sharpest of the three when the peer set is genuinely comparable, and the most misleading when it is not, because the comparison inherits every distortion in the group.

The company's own history answers: is the market pricing this business differently from how it has priced the same business in the past? This one has a real advantage — the business is held constant, so you are not comparing two different animals. Its weakness is that companies change. A five-year median multiple is only a fair reference if the product mix, growth rate, capital structure and competitive position have not materially moved over that window. After a large acquisition or a shift from project work to annuity revenue, the company's own history describes a business that no longer exists.

The broad market answers: is the market pricing this business differently from equities as a whole? This is the crudest reference and still the most useful when a whole sector has moved together, because it is the only one of the three that sits outside the sector.

A sector label is a filing convention, not an economic statement. Two companies can sit inside the same NSE sectoral index and run businesses with almost nothing in common — different customers, different margin structures, different working capital, different capital expenditure needs. Comparing their multiples produces a difference that is entirely explained by them being different businesses, which is not information.

Build the set from the revenue line upward instead. The steps below take longer than filtering a screener by sector, and they routinely end with a peer set of three or four names rather than fifteen. A small honest set beats a large dishonest one every time.

The table below uses an illustrative peer group — Bharat Cables Ltd, Konkan Conductors Ltd, Godavari Electricals Ltd and Malabar Cable Works Ltd, all illustrative and none of them real companies. What matters is not the names but the reason attached to each verdict, because that reason is the test you will apply to a real peer set.

Once the peer set is honest, the choice of multiple decides what the table is actually measuring. The valuation ratios topic covers what each one is blind to; the peer-table question is narrower — which blindness is tolerable across this specific group.

If every company in the set has broadly similar debt, P/E is fine and is the most intuitive to read. The moment debt loads diverge, P/E starts comparing financing decisions instead of operations, and the table has to move to enterprise value. If the group contains one loss-maker, no earnings-based multiple works for the whole set, so either drop that company or move everyone to EV/Sales. For banks and NBFCs, P/B read alongside return on equity is the working standard, because a lender's book value is a far more meaningful quantity than a manufacturer's.

The rule that matters: every company in the table must be on the same multiple, computed the same way, over the same period. A table mixing one company's forward P/E with another's trailing P/E is not a comparison, it is a category error — and it always flatters whichever company is quoted forward.

It is tempting. A listed Indian company and a listed overseas company make similar products, and one trades at a visibly different multiple, and the conclusion writes itself. It is almost always the wrong conclusion, because most of the gap is produced by things that have nothing to do with either business.

Cost of capital is the largest of them. The multiple a market pays is inversely related to the rate at which it discounts future cash, and that rate starts from the local risk-free rate. An economy whose government borrows at one level and an economy whose government borrows at a materially different level will not pay the same multiple for identical cash flows, ever. Tax regimes differ too, and after-tax cash is what an owner receives.

Growth runway is the second. A business with domestic demand still expanding has a longer visible growth path than the same business in a saturated market, and duration of growth is one of the biggest drivers of a justified multiple. Disclosure and accounting regimes are the third: revenue recognition, lease treatment, depreciation policy and related-party disclosure differ enough that the denominators are not measuring quite the same thing. Add index membership, currency, free float and the composition of the shareholder base, and the cross-border multiple gap becomes almost uninterpretable.

None of this makes global comparison useless. It is genuinely useful for structure — what margins a mature version of this industry earns, how consolidated it eventually becomes, what returns on capital are achievable. Use overseas peers for the shape of the business and the domestic set for the level of the multiple.

A share price can rise for two entirely different reasons: the company earned more, or the market decided to pay more for each rupee it earns. The second of those is a re-rating, and it is the mechanism behind most of the very large moves in Indian mid-caps.

Consider an illustrative case. A company earns ₹20 per share and trades at ₹300, a multiple of 15. Two years later it earns ₹28 and trades at ₹700, a multiple of 25. Earnings grew 40 per cent; the price grew 133 per cent. The remaining move came entirely from the multiple expanding, and multiple expansion has causes worth naming: a shift from cyclical project revenue toward recurring revenue, a balance sheet repaired to the point where bankruptcy risk left the conversation, entry into an index that brought passive buying, a governance issue resolved, or simply a sector coming into fashion.

De-rating is the same mechanism running backwards, and it is where relative valuation hurts most. Earnings can grow while the multiple contracts faster, so the price falls in a year the company reports its best-ever numbers. When that happens the market is usually re-pricing the durability or the risk of those earnings rather than disputing them.

The practical takeaway is that a multiple is not a constant to be treated as a law of nature. Half the return in a re-rating comes from something no earnings forecast will predict, and the same is true in reverse. That is precisely why relative valuation is a starting point rather than a conclusion.

Peer sets in India are usually small. Once you have applied a real comparability test you often have five or six names, and in a small sample one outlier destroys the average.

Look at an illustrative set of six cable and conductor makers below — all illustrative companies, none of them real. The mean multiple is 27.5 and the median is 21. Those two numbers would lead to entirely different conclusions about the same company, and the difference is created by one name whose multiple has been distorted by a temporary collapse in its earnings denominator.

Use the median for any peer set smaller than about ten names. Better still, do not reduce the set to a single number at all: report the range, and look at where the company sits inside it. A company at the low end of a tight range is a different situation from a company at the low end of a range that runs from 18 to 62, because the second range is telling you the group is not homogeneous enough to average in the first place.

Here is the structural weakness at the heart of relative valuation, and it does not have a clean fix. If every company in a sector is valued by reference to its peers, and each of those peers is valued by reference to the others, then the entire group is anchored to itself. Nothing in that loop is anchored to what the businesses are actually worth.

The consequence is that a sector can drift a long way from the cash its companies will generate while every single company inside it remains fairly priced relative to the others. Relative valuation cannot detect this, by construction. It is a tool for finding differences within a group. It is structurally incapable of finding a problem shared by the whole group, because the shared problem is embedded in the reference point.

This is the strongest argument for keeping an absolute method in the toolkit. A discounted cash flow model, covered in the DCF topic in this Academy, is anchored to the cash a business is forecast to produce and to a discount rate, neither of which is borrowed from the neighbours. It has its own well-known weaknesses — assumption sensitivity, a terminal value that dominates the answer — but its errors are independent of the peer group's errors, and independent errors are exactly what you want in a second opinion.

The workable practice is to run relative valuation as the primary lens because it is fast and grounded in observable prices, and to sanity-check the sector as a whole against an absolute anchor or against the broad market. When the peer group and the absolute method disagree sharply, that disagreement is the most valuable output of the whole exercise.

Companies inside a comparable set do not deserve identical multiples, and expecting them to is as naive as ignoring the set entirely. The question is whether the gap can be traced to something structural.

A premium is defensible when it maps onto a measurable, durable advantage. A higher return on capital means each reinvested rupee produces more, so growth is worth more. A longer growth runway means the advantage lasts more years. Lower capital intensity means more of the profit becomes cash the owners can actually receive. More predictable earnings — recurring revenue, contracted volumes, a diversified customer base — mean a lower risk premium. Cleaner governance and simpler group structures reduce the discount investors apply for the unknown. Each of these is checkable in the filings.

A premium is not defensible when the only supporting evidence is that the price has been rising, that the sector is currently in favour, or that a peer trades even higher. The last one is the circularity problem in miniature. The clean test: write one sentence explaining what this company does structurally better than the peer it trades above, then find the number in the annual report that supports the sentence. If no number exists, you have found momentum wearing the clothes of analysis.

A finished peer table looks decisive, and that is its most dangerous property. It is a grid of numbers where one company sits visibly below the others, and the mind supplies a conclusion before the analysis has begun.

The correct reading is the opposite. A gap in a peer table is the market disagreeing with you about something specific, and the market usually knows something. So the output is a question, and the question has a standard form: what would have to be true about this business for the gap to be right — and separately, what would have to be true for it to be wrong? Both versions are researchable. Neither is a conclusion.

That question then sends you back to the filings, which is where the actual work happens. Perhaps the discount reflects a customer concentration disclosed in the segment note. Perhaps it reflects a promoter pledge visible in the shareholding pattern, or a contingent liability in the notes, or a working-capital cycle that has been stretching for four consecutive quarters. Perhaps, after all that, nothing explains it — which is itself a finding worth recording, along with the date, so you can see later whether the gap closed and why.

Relative valuation done well produces a list of well-posed questions and a record of what the market appeared to be assuming at a point in time. Done badly it produces a verdict. The difference between the two is entirely in how the person reading the table chooses to treat it.