Comparable companies analysis

A DCF asks what a business is worth from first principles. Comps ask a humbler question: what is the market paying today for businesses like this one? Humble, but powerful: the market's price for similar assets is real information, and any intrinsic view that ignores it is choosing to argue with everyone at once.

The idea: price per unit of performance

A multiple standardizes price by a measure of performance so differently sized companies become comparable. EV/EBITDA of 10x says the market pays ten dollars for each dollar of annual EBITDA. The whole method is: measure what the market pays per unit for the peers, apply that rate to your company's units, and study the difference from its actual price.

The one unbreakable rule: match the numerator to the denominator

The consistency rule

ENTERPRISE VALUE goes over metrics that belong to ALL capital:
    EV / revenue, EV / EBITDA, EV / EBIT

EQUITY VALUE goes over metrics that belong to shareholders alone:
    P / E, P / B, P / FCFE

EV / earnings and price / EBITDA are category errors: they compare a
value belonging to one group with income belonging to another, and the
error scales with leverage

This is the comps counterpart of the DCF's matching rule, and violating it is the most common structural error in amateur comp sets.

Which multiple, when

Multiple	Best for	Watch out
EV / EBITDA	The default for operating companies	Ignores capex; flatters asset-heavy firms
EV / EBIT	Where depreciation is a real economic cost	Depreciation policies differ across peers
P / E	Mature, similarly levered, profitable firms	Leverage and one-offs distort it
EV / revenue	Pre-profit growth; software	A last resort: ignores profitability entirely
P / B	Banks, insurers	Meaningless where assets are intangible
Sector-specific	P/FFO (REITs), EV/subscriber, EV/tonne	Only as good as the unit's comparability

Multiples can also run forward (next twelve months' consensus estimates) rather than trailing. The street prefers forward, because prices look ahead, at the cost of importing analysts' estimate errors into the denominator.

Choosing peers: the step that decides the answer

The multiple you conclude is mostly decided when the peer set is chosen, which is why it deserves the most scrutiny and gets the least. Real peer selection matches on the drivers of multiples, not on the sector label:

Business model: who they sell to and how they earn, not the ticker classification. A payments processor is not a bank.
Growth and margin profile: a 25%-grower priced against 5%-growers produces a fake premium.
Size and liquidity: micro caps trade at structural discounts.
Geography and accounting: same rules, same currency exposure, or adjust.

The premium is information, not error

The point of comps is not to average the peers and declare the gap mispricing. A company trading above its peers is being paid a premium FOR something: growth, margin, moat, balance sheet. The analyst's job is to name the something and judge whether it is worth the premium. "Cheapest in the group" is where analysis starts, not where it ends; the cheapest stock in a comp set is often correctly the worst business in it.

Spreading: the unglamorous work that makes it real

"Spreading comps" is banker vocabulary for building the numbers by hand rather than screenshotting a data vendor, and it is where accuracy lives:

Normalize earnings: strip one-time charges, litigation, gains on sales, from every company on the same basis.
Calendarize: peers with different fiscal year-ends must be put on the same twelve months.
Build EV properly: market cap at diluted shares, plus debt, minority interest and preferred, minus cash, with leases treated consistently across the set.
Use medians: one 40x outlier drags a mean; the median resists it.

From multiple to value

Applying the peer multiple

implied EV      = peer median EV/EBITDA x subject EBITDA
implied equity  = implied EV - net debt - minorities - preferred
implied / share = implied equity / diluted shares

presented as a RANGE (25th to 75th percentile of peers), never a point

What comps cannot do

They inherit the market's mood. In a bubble, comps prove every stock is reasonably priced against other bubble stocks. Comps measure relative position, never absolute sanity; that is the DCF's job.
True one-of-a-kinds break them. The honest fallback is a wider set plus explicit adjustment, and more weight on intrinsic methods.
They tempt circularity. If everyone prices against everyone, no one has priced anything. Somewhere in the loop there has to be an anchor to cash.

Glossary for this guide

Enterprise value: The value of the whole operating business, belonging to debt and equity holders together. Market capitalization plus net debt (plus minority interests and preferred stock, when present).
Equity value: What belongs to shareholders: enterprise value minus net debt and other senior claims. Divided by diluted shares, it becomes a per-share value comparable to the stock price.
EBITDA: EBIT with depreciation and amortization added back. A rough proxy for operating cash generation, popular in comparisons because it ignores differences in asset age and financing; dangerous when treated as real cash flow, because capex is real.
EBIT: Earnings before interest and taxes, better known as operating profit. What the business earns from operations before anyone (lenders, the tax authority) takes a share.
Net debt: Total debt minus cash and equivalents. The bridge between enterprise value and equity value; getting it wrong misprices every levered company, in proportion to its leverage.
Diluted shares: Share count including the stock that options, warrants and convertibles would add. Per-share value uses diluted shares because those claims are real even before they convert.
Trading multiple: Price standardized by a unit of performance so unlike-sized companies compare: EV/EBITDA, P/E, EV/revenue. Ten times EBITDA means the market pays ten dollars per dollar of annual EBITDA.
Forward multiple: A multiple computed on the NEXT twelve months' consensus estimates rather than the last twelve reported. The street's default, because prices look ahead, at the cost of inheriting the estimates' errors.
Calendarization: Restating companies with different fiscal year-ends onto the same twelve months so their multiples compare. Skipping it quietly compares one company's January-December against another's July-June.
Normalized earnings: Earnings with one-time items stripped out on a consistent basis across a peer set: restructurings, litigation, gains on sales. The step that makes a comp set mean something.
Median vs mean: Comp sets quote medians because one 40x outlier drags a mean and the median resists it. When someone presents a mean multiple, look for the outlier doing the work.
P/TBV: Price to tangible book value: market price over book value with goodwill and intangibles stripped out. The standard bank multiple, read against ROE, since a bank earning above its cost of equity deserves more than tangible book and one earning below deserves less.
FFO: The REIT sector's earnings measure: net income with real estate depreciation added back and property-sale gains removed, because buildings on the books depreciate while often appreciating in fact.