In comparable company analysis there is no single "correct" multiple — the analyst's first real decision is which valuation multiple to apply, and that choice can swing the implied value by tens of percent. The right multiple is the one whose denominator (1) is positive and meaningful for the company and its peers, (2) is least distorted by accounting and capital-structure noise, and (3) best captures the economic value driver that the market actually pays for in that industry. The core tension is that the most available multiple (trailing P/E) is rarely the most comparable one, and the most comparable one (a forward, capital-structure-neutral measure) often requires data or normalization the analyst does not have.

The selection framework

Three filters narrow the field, applied in order.

1. Numerator–denominator consistency (the non-negotiable rule). The claim represented in the numerator must match the claim represented in the denominator. Enterprise value (EV) belongs to all capital providers, so it pairs only with pre-interest flows available to all of them: EBITDA, EBIT, revenue, unlevered cash flow. Equity value (price / market cap) belongs only to shareholders, so it pairs only with post-interest flows: net income, EPS, book equity. Mixing them — EV/Net Income or Price/EBITDA — produces a meaningless ratio because numerator and denominator accrue to different groups (Macabacus; Wall Street Prep). This single rule eliminates most invalid multiples before any judgment is needed.

2. Profitability and life stage. The denominator must be positive and reasonably stable, or the multiple is uninterpretable. A negative-earnings company has no usable P/E or EV/EBIT, forcing the analyst up the income statement toward EBITDA, gross profit, or revenue. The conventional ladder (CFA Institute; Valutico):

• Mature, profitable, comparable leverage → P/E (and PEG when growth differs).
• Differing leverage, tax rates, or D&A → EV/EBITDA or EV/EBIT (capital-structure neutral).
• Capital-intensive (telecom, cable, utilities, industrials) → EV/EBIT or EV/(EBITDA − capex), because EBITDA flatters businesses with heavy depreciation; Damodaran notes EV/(EBITDA − capex) suits cable but is wrong for an asset-light consulting firm.
• Unprofitable / high-growth / cyclical trough → EV/Sales or P/S (revenue is positive and harder to manipulate).
• Financials (banks, insurers) → P/B and P/E; EV is ill-defined when debt is the raw material, and book value tracks regulatory capital.
• Sector-specific operating metrics where GAAP is uninformative → EV/EBITDAR (retail/airlines, pre-rent), EV/subscriber, EV/daily production (E&P), price/AUM, price/FFO (REITs).

3. The value driver the market prices. The best multiple's denominator should correlate tightly with what investors actually capitalize in that industry. Where margins are unstable, an earnings multiple is noisy and a revenue multiple may be cleaner; where the asset base drives returns, book-based multiples dominate.

How it's used in practice

Analysts almost never rely on one multiple. Standard practice is to build a comps table showing several (commonly EV/EBITDA, EV/EBIT, P/E, and EV/Sales) side by side, take the peer-set median or interquartile mean for each, apply them to the target's corresponding metric, and triangulate — treating wide divergence between multiples as a signal to investigate rather than average away (Street of Walls; Morgan Stanley Counterpoint Global).

Two refinements matter. First, forward over trailing: applying multiples to next-twelve-month or next-fiscal-year estimates (forward P/E, forward EV/EBITDA) is standard on the buy- and sell-side because price discounts the future. Second, clean comparability: many analysts prefer EV/EBIT or EV/NOPAT over EV/EBITDA precisely because EBITDA ignores the real cost of capital consumption (D&A) and taxes, letting capital-intensive firms look artificially cheap (Footnotes Analyst).

Adoption, debate & evidence

The framework above is the consensus taught by the CFA program and in every investment-banking comps model — it is mainstream, not contested. What is debated is which multiple is most accurate, and here the folklore (that EV/EBITDA is the professional's gold standard) diverges from the measured evidence.

The most-cited study, Liu, Nissim & Thomas (2002, Journal of Accounting Research), tested multiples on thousands of US firms and found a consistent accuracy ranking across nearly all industries: (1) forward-earnings multiples performed best, then trailing earnings, then cash-flow and book-value measures (roughly tied), and sales multiples worst. Forward-P/E pricing errors fell within ~15% of price for about half the sample. Notably, EBITDA-based multiples did not outperform simple earnings multiples in their sampling. Their headline lesson: the choice of flow timing (forward vs trailing) matters more than the choice of denominator type, and revenue multiples are a last resort, justified mainly when earnings are negative. European replications (e.g. Schreiner) broadly echo the primacy of forward earnings while noting knowledge-intensive industries reward equity-value multiples.

The honest caveat: these are explanatory-power studies (how well a multiple matches contemporaneous prices), not proof that any multiple predicts forward returns. A multiple tells you what peers trade at, not whether peers are correctly priced — if the whole sector is in a bubble, every multiple inherits the bubble.

Strengths & limitations

Choosing well makes comps fast, market-anchored, and transparent. The method's weaknesses all trace to multiple selection: garbage-denominator risk (negative, near-zero, or one-off-distorted denominators give nonsense ratios), false comparability (peers differing in growth, margin, or leverage are not actually comparable on a raw multiple — the fix is regression or fundamentals-adjustment, per Damodaran), and sector mis-fit (using EV/EBITDA on a capital-intensive firm, or P/E across firms with very different leverage). The single most common misuse is defaulting to trailing P/E because it is easy, ignoring that leverage differences mechanically distort P/E (more debt → lower P/E with no change in value) and that earnings may be negative, cyclical, or low-quality. The second most common is cherry-picking the multiple that yields the desired answer — the discipline of showing all of them in one table exists precisely to constrain this.

Sources

• CFA Institute — Market-Based Valuation: Price and Enterprise Value Multiples (refresher reading): rationale per multiple, numerator-denominator matching, fundamentals drivers.
• Liu, Nissim & Thomas (2002), "Equity Valuation Using Multiples," Journal of Accounting Research 40(1) — accuracy ranking (forward earnings > trailing > cash flow/book > sales).
• Aswath Damodaran (NYU Stern), Relative Valuation lecture notes — consistency, peer-group definition, EV/(EBITDA − capex) for capital-intensive firms, "abuse not use."
• Macabacus & Wall Street Prep — enterprise-vs-equity numerator/denominator consistency rule and invalid combinations.
• The Footnotes Analyst — Relative valuation conflicts: EV/EBITDA versus P/E — leverage-driven divergence; case for EV/NOPAT over EV/EBITDA.
• Street of Walls; Valutico; Morgan Stanley Counterpoint Global — practitioner comps-table construction and life-stage selection.

Dispute flagged: "EV/EBITDA is the most accurate professional multiple" is folklore; measured evidence (Liu et al.) ranks forward earnings highest and finds EBITDA multiples do not beat earnings multiples. Accuracy studies measure explanatory power vs current prices, not predictive edge.