Comparable-company method: building a peer set that actually survives scrutiny

Comparable-company analysis sounds straightforward. Find listed companies in the same business as the subject company. Compute their trading multiples. Apply the median or mean to the subject's financials. The output is a valuation range.

In practice, the method fails more often than it works. The peer set is too thin, the comparables are not really comparable, the multiples being used are wrong for the subject's profile, or the range produced is so wide as to be useless. We rebuild a lot of comparable-company analyses for clients who came to us with a number from an internal exercise that they then discovered nobody believed.

What follows is the discipline we apply to peer-set construction and multiple selection. It is more work than the standard approach. It also produces an output that an acquirer's banker will engage with seriously rather than dismiss in five minutes.

The five filters that define a peer

A defensible peer set is built by applying filters in sequence. Each filter narrows the universe. The starting universe is usually a sector classification (NIC code, GICS sub-industry) of all listed companies on NSE and BSE.

Filter one: industry and business model. Same NIC code is not sufficient. Two companies with the same industry classification can have completely different unit economics. A B2B SaaS company and a B2C subscription company both sit in 'information services' but should not be in each other's peer sets. We read each candidate's last annual report to confirm the business model matches before we keep them in.

Filter two: size band. Revenue between 0.5x and 2x of the subject company is the working range. A Rs. 50 crore revenue subject should be compared against companies in the Rs. 25-100 crore revenue band. Going to Rs. 500 crore listed companies introduces scale effects that distort the multiple. Going below Rs. 25 crore introduces small-cap illiquidity premia.

Filter three: growth rate. Revenue growth within plus-or-minus 15 percentage points of the subject. A company growing at 80 percent should not be compared against listed companies growing at 12 percent, even if the businesses look similar. The market prices growth differently at different bands.

Filter four: profitability profile. Either positive EBITDA peers or negative EBITDA peers, not a mixed bag. If the subject is at 8 percent EBITDA margin, the peer set should be companies at 0-20 percent. A peer set containing both 25-percent-margin profitable peers and 5-percent-loss-making peers will produce a multiple range that does not mean anything.

Filter five: geography and listing. Indian-listed peers first. Where Indian peers are too thin, global-listed peers from comparable economies, with a country-risk adjustment applied to their multiples.

After these five filters, the peer set is usually 5-9 companies. Below 4, the analysis is not statistically meaningful. Above 12, you are diluting the set with marginal comparables.

Why Indian-listed peer sets are thin for tech and SaaS

For traditional sectors — manufacturing, FMCG, financial services, infrastructure — Indian listed comparables are abundant. The peer-set construction is mechanical.

For Indian tech, SaaS, D2C, fintech, and most modern-economy businesses, the listed peer universe is genuinely thin. Zomato, Nykaa, Paytm, PB Fintech, and a handful of others are the closest analogues, and they are large, profitable-or-near-profitable, and not particularly comparable to a Series B company.

This is where global peer sets become necessary. US-listed SaaS companies in a similar size and growth band are far more comparable than a forced match with a domestic conglomerate. The standard adjustment is to apply a country risk premium discount to the global multiple to reflect Indian execution risk and currency risk. Damodaran's country risk premium for India sits at around 2.5-3 percent on the cost of equity. Translating that into a multiple discount on EV/Sales or EV/EBITDA: typically a 15-25 percent haircut on the global peer median.

We document the country-risk adjustment explicitly in the report. Reviewers can disagree with the size of the haircut, but they cannot dismiss the analysis as unanchored.

Which multiple, and why

The four standard multiples each serve a different purpose. Using the wrong one produces an unreliable answer.

EV/Sales

Used when the subject is unprofitable or thinly profitable, and the comparables are similar. For early-stage SaaS, marketplaces, and high-growth D2C, this is the most common multiple. The trap is that EV/Sales is highly sensitive to gross-margin profile. Two companies at the same revenue but at 75 percent versus 35 percent gross margin will warrant materially different EV/Sales multiples. We restrict EV/Sales analysis to peer sets where gross margin is within a 10-percentage-point band.

EV/EBITDA

Used when the subject is EBITDA-positive and the comparables are too. The most widely-accepted multiple for mid-stage companies. The trap is non-recurring items in reported EBITDA. We work with normalized EBITDA: reported EBITDA adjusted for one-time gains, one-time losses, founder above-market salary, and unusual provisions. The normalized number is the one used in the multiple.

P/E

Used for mature, profitable businesses with stable capital structures. Less useful for growth-stage Indian companies because earnings are volatile and tax rates vary across peers. We include it as a cross-check but rarely as the primary multiple.

EV/EBIT

Used when depreciation policies vary materially across the peer set, which makes EBITDA misleading. Capex-heavy businesses (manufacturing, infrastructure) often warrant EV/EBIT over EV/EBITDA. For asset-light businesses, EV/EBITDA is fine.

The 30 percent range as a fair peer-set spread

When the analysis is done well, the peer-set multiples will spread across a range. The right way to summarize the range is to report the 25th percentile, median, and 75th percentile. A peer set where the 25-75 range is within 30 percent of the median is a clean, internally consistent set. A spread wider than that means the peers are not as comparable as the filters suggested, and the analysis should be rebuilt.

For our subject company, we typically apply the median multiple as the central estimate and the 25-75 percentile range as the valuation band. A subject company with Rs. 18 crore normalized EBITDA against a peer set median EV/EBITDA of 14x, with a 25-75 range of 12-17x, yields an enterprise value central estimate of Rs. 252 crore and a range of Rs. 216-306 crore.

Where the method legitimately disagrees with DCF

Comparable-company analysis and DCF will not produce the same number. They are not supposed to. The two methods reflect different things: DCF reflects the cash-generative capacity of the business under a specific set of forward assumptions, comparables reflect what the market is currently paying for similar businesses.

When the two methods diverge by less than 25 percent, the valuation is well-supported. When they diverge by 50 percent or more, one of them is wrong, and the right response is to investigate which. Often the DCF is wrong because the terminal value assumptions are out of line with what the market is actually paying, which is exactly what the peer-set multiples are telling you.

We present both methods in every valuation report and explain the divergence rather than averaging it away. Averaging two different methods does not produce a better answer. It produces a number that is wrong by less but for which the wrongness is no longer traceable.