Comps (Comparable Company Analysis)

Key takeaways

Comps prices a business relative to its peers, not on intrinsic cash flows. Quick to build, easy to communicate, but only as good as the peer set.
The standard multiples: EV/EBITDA, EV/Revenue, P/E. EV/EBITDA is most common because it's capital-structure-neutral.
A peer set should match on industry, scale, growth profile, geography, and margin structure — not industry alone.
Output a fan: median, 25th percentile, 75th percentile multiples applied to your target's metric. The range is the answer; the single number is a fiction.
Comps complement DCF — neither is the truth. Triangulate.

The method

Comps says: this target should trade roughly the way comparable peers trade today. If peers are at 12× LTM EBITDA, the target's implied EV is 12× the target's LTM EBITDA. Bridge to equity value through the standard EV-to-equity adjustment (subtract net debt), divide by diluted shares, get an implied price per share.

The math is trivial. The judgment is in selecting the peer set, picking the right multiple, and choosing whether to use median, mean, or a specific percentile. A defensible comps analysis is mostly about defending those choices.

Building the peer set

A peer set is good when its peers really are comparable to the target. Match on:

Industry — same end markets, same business model
Scale — within an order of magnitude on revenue and market cap
Growth profile — similar revenue growth band; high-growth and stable peers should not be averaged
Geography — same major markets, similar regulatory and tax regimes
Margin structure — similar EBITDA margins; otherwise EV/EBITDA isn't measuring the same thing
Capital intensity — similar capex/D&A patterns
Trading liquidity — exclude micro-caps where prices may be stale

Typical peer set size: 5–10 names. Fewer than 5 and the median is noisy; more than 10 and the set probably includes weak comps that should be culled. Document why each peer is in (or out) of the set — this is the part that comes up in committee.

Standard multiples and when to use each

Multiple	When to use	Watch out for
EV/EBITDA	Default for most industries	Hides capex differences
EV/EBIT	Capital-intensive industries (penalises higher D&A appropriately)	Sensitive to depreciation accounting
EV/Revenue	Pre-EBITDA companies (early SaaS, biotech)	Says nothing about profitability
EV/(EBITDA − CapEx)	Cross-industry capital-intensity comparisons	Less standardised data
P/E	Mature, financially-stable, similar-leverage businesses	Capital-structure-dependent, tax-jurisdiction-dependent
P/B	Banks, insurance, asset-heavy financials	Useless for asset-light businesses
EV/Customer or EV/Subscriber	Cable, telco, SaaS — when units are stable and quality-comparable	Doesn't reflect monetisation differences
EV/MW (utilities), EV/Reserves (E&P)	Industry-specific operating multiples	Only meaningful within the industry

Worked example

Target: a mid-cap consumer brand. LTM EBITDA $80M, LTM Revenue $400M, Net Debt $50M, Diluted Shares 50M.

Peer	LTM EV/EBITDA	LTM EV/Revenue
Peer A	11.5×	2.1×
Peer B	13.0×	2.5×
Peer C	10.5×	1.9×
Peer D	14.2×	2.8×
Peer E	12.0×	2.3×
Median	12.0×	2.3×
25th pct	11.0×	2.0×
75th pct	13.5×	2.6×

Applying median EV/EBITDA: EV = 12.0 × $80M = $960M. Less net debt $50M → equity $910M → implied price $18.20/share. The 25th–75th percentile range gives you a band: implied price between roughly $16.40 and $20.40. The wider EV/Revenue check (2.3× × $400M = $920M EV) cross-validates.

Median, mean, or quartile?

Use the median as the central tendency — it's robust to outliers in the way a mean is not. With a small peer set (5–10), one outlier can swing the mean meaningfully. Quote median in committee, but show the full distribution: 25th, median, 75th, max, min. Anyone reading the table can pick their own central tendency.

Where mean might be appropriate: very large peer sets (20+) where outliers wash out, or when the distribution is well-behaved (close to normal).

The size and growth premium / discount

Two adjustments often applied to raw comps multiples:

Size discount — smaller companies typically trade at lower multiples than larger peers due to liquidity and risk premia. If your target is meaningfully smaller than the peer median, apply a 10–25% discount to the implied multiple. Damodaran's size premium tables are the standard reference.
Growth adjustment — if the target is growing meaningfully faster (or slower) than peers, the multiple should reflect it. PEG-style adjustment: divide multiple by growth rate and compare across peers.

Both are judgment calls. State them explicitly in the model so the committee can challenge or accept.

Comps vs DCF — different lenses

	Comps	DCF
What it measures	Relative value (vs peers)	Intrinsic value
Time required	Hours	Days
Inputs	Peer multiples, target metric	Five-year projection, WACC, terminal
Strength	Market-based, defensible, communicable	Builds the operational thesis
Weakness	Wrong if peers are mispriced (bubbles)	Sensitive to long-dated assumptions

Standard practice: do both, present them as a "football field" chart that shows the implied valuation range from each method (comps, precedent transactions, DCF, LBO bid math). The point isn't a single answer — it's where the methods converge.

Common comps errors

Bad peer selection. Picking peers by industry alone, ignoring scale or growth differences. The most common silent error.
Mixing LTM and NTM within the same comps table. Pick one and apply consistently.
Wrong EV calculation. Forgetting preferred stock, minority interest, or capitalised leases produces understated EVs and inflated multiples.
Stale data. Multiples computed on share prices from a month ago in a moving market are wrong. Use current prices and recent financials.
Median of three. Three peers is too few — you're effectively looking at the middle one. Aim for 5–10.

How Smalt AI builds it

Smalt AI's comps tab pulls market data and financials for the peer set, computes EV consistently (including preferred and minority interest), reports LTM and NTM multiples for the standard set (EV/EBITDA, EV/Revenue, EV/EBIT, P/E), and outputs the median, mean, 25th, and 75th percentile by multiple. Peer selection logic is documented per row (industry, size band, growth band, geography). The output bridges from peer median to the target's implied valuation range.