Guide

Comparable company analysis explained

Harbor Analytics, a vertical data-software vendor, traded at 8.1× EV/EBITDA on trailing numbers — a headline discount to “software peers.” The first comps table pulled twelve names from a broad GICS screen: three were loss-making growth stories with negative EBITDA, two were IT resellers with 4% gross margins, and one carried a one-time litigation gain that inflated earnings. The naive median implied fair value at 18.2× — more than double the spot multiple — and bulls filed it as proof of deep undervaluation. After rebuilding the peer set with business-model filters, normalizing stock-based compensation and restructuring add-backs, and switching to forward consensus EBITDA, the comp band tightened to 10.8×–12.1× (median 11.4×). Harbor was roughly fair, not a steal; the stock rerated modestly when guidance confirmed margin expansion rather than a multiple catch-up.

Comparable company analysis (CCA, or trading comps) values a target by applying valuation multiples observed on similar publicly traded companies. It is relative valuation: the market’s price for peers becomes a ruler for the subject. Comps complement discounted cash flow models and anchor negotiation in M&A. This guide covers peer selection, financial normalization, enterprise value multiples, statistical aggregation, applying the comp range to a target, the Harbor Analytics refactor, a technique decision table, pitfalls, and an investor checklist.

What trading comps measure (and what they do not)

Comps answer: what is the market willing to pay today for businesses like this one? They do not forecast intrinsic cash flows — that is DCF territory. They also do not capture control premiums or synergy value the way acquisition comps might. Trading comps reflect minority, liquid stakes in public companies, usually at a discount to what a strategic buyer might pay for 100% control.

Common comp multiples

EV/EBITDA — the workhorse for operating businesses with positive EBITDA; capital-structure neutral when EV is built correctly.
EV/Revenue — used when EBITDA is negative or immaterial; requires gross-margin context.
P/E (price-to-earnings) — equity-level multiple; breaks on negative earnings and distorts with leverage differences.
EV/FCF or P/FCF — ties valuation to cash after capex; harder to standardize across capex-heavy sectors.

Each multiple embeds assumptions about growth, margins, and risk. A high EV/EBITDA on a peer set usually means the market expects faster growth or lower risk — not that every name deserves the same sticker price.

Building a defensible peer set

Garbage peers produce garbage ranges. Screen in layers rather than dumping every company in the same industry code into a spreadsheet.

Screening dimensions

Business model — subscription vs license vs services mix; asset-light vs capital-intensive; same revenue recognition pattern.
End market — enterprise vs SMB; geographic exposure; cyclical vs defensive end demand.
Scale — revenue and EBITDA within roughly 0.3×–3× of the target; micro-caps and mega-caps rarely belong in the same row.
Growth and profitability — similar revenue CAGR bands and EBITDA margin profiles; a 40% grower with −10% margins is not a comp for a 12% grower at 25% margins.
Balance sheet — extreme leverage or net-cash outliers distort equity multiples; note them even if you keep EV-based metrics.

Aim for 6–12 peers after filters. Fewer than five makes the median fragile; more than fifteen usually means the screen is too loose. Document why each name is in or out — future you (and skeptical investors) will ask.

Normalizing financials before you multiply

Reported GAAP numbers rarely line up across peers. Normalization is where most amateur comp tables fail.

EBITDA adjustments

Start from operating income and add back D&A, or use company-reported adjusted EBITDA with skepticism.
Add back one-time restructuring, litigation, and M&A costs that will not repeat — but do not add back recurring stock-based compensation unless you apply the same rule to every peer.
For cyclical industries, consider mid-cycle EBITDA rather than trough or peak trailing twelve months.

Enterprise value construction

EV = equity market cap + total debt + preferred stock + minority interest − cash and equivalents. Use the same share count (basic vs fully diluted) consistently. For recent financings, use pro forma share count. Mis-stated EV is the silent killer of comp tables — see enterprise value explained for the full bridge.

Calendarization and LTM vs NTM

Peers with mismatched fiscal year-ends need calendarized revenue and EBITDA. Trailing twelve months (LTM) reflects history; next-twelve months (NTM) from consensus estimates reflects expectations. Growth stories often trade on NTM; mature cyclicals on LTM mid-cycle. Mixing LTM numerators with NTM denominators (or vice versa) without labeling it is a common error.

From peer multiples to a valuation range

Calculate the chosen multiple for each peer (e.g., EV/LTM EBITDA).
Drop peers with negative or near-zero denominators unless EV/Revenue is the explicit metric.
Aggregate with median as the default — means are pulled by one outlier acquisition premium or distressed name.
Report 25th and 75th percentiles (or high/low) to show the band, not a false precision point.
Apply the median (or selected percentile) to the target’s normalized metric to get implied enterprise value.
Bridge from EV to equity value: subtract net debt, divide by diluted shares for implied price per share.

Worked intuition (Harbor Analytics)

After the peer rebuild, seven vertical-software comps traded at 10.8×–13.2× NTM EV/EBITDA (median 11.4×). Harbor’s normalized NTM EBITDA was $142 million. Implied EV = 11.4 × $142M ≈ $1.62B. Net debt of $180 million implied equity value ≈ $1.44B; at 52 million diluted shares, ≈ $27.70/share — within 8% of the then-market price, not the 120% upside the first sloppy table suggested.

Harbor Analytics refactor: what changed

Peer count 12 → 7 — removed resellers, negative-EBITDA growth names, and a conglomerate with <20% segment exposure.
LTM → NTM EBITDA — aligned with how growth software trades; consensus matched management guidance within 3%.
SBC policy — deducted stock-based comp uniformly; one peer’s “adjusted EBITDA” add-back had inflated the old median by 1.4 turns.
Statistics — reported median and interquartile range; dropped mean after one peer spiked on a short-squeeze week.
Secondary check — EV/Revenue on the same peers corroborated the band when margin differences were noted.

The refactor did not change Harbor’s operations — it changed the question from “are we cheaper than a random software basket?” to “are we priced like similar vertical SaaS at our growth and margin profile?”

Technique decision table

Approach	Best for	Weak when
Trading comps (CCA)	Public companies, sector with liquid peers, quick relative checks, fairness opinions	No true peers, illiquid niche, one-off assets
DCF	Unique cash flows, high growth with path to profitability, LBO returns	Terminal value dominates, no visibility, extreme rate uncertainty
Precedent transactions	Control M&A, synergy-rich deals, private targets	Stale deals, different credit cycle, sparse data
P/E multiples	Mature, profitable, low-leverage consumer and industrial names	Negative earnings, wide leverage dispersion within peer set
EV/Revenue	Pre-profit growth, early SaaS, biotech pre-commercial	Margin structure differs wildly across peers

Common pitfalls

Industry-code screening only — GICS buckets lump resellers with software and banks with fintech.
Mixing LTM and NTM — doubles or halves implied value without a footnote.
Ignoring SBC in software comps — adjusted EBITDA that adds back recurring equity pay is not comparable.
Mean instead of median — one M&A rumor or meme squeeze poisons the average.
Wrong EV — forgetting minority interest, preferred, or pension deficits; using basic shares when options are deep in the money.
Cyclical peak EBITDA — cheap multiple at the top of the cycle is a value trap, not a bargain.
Applying peer growth to a slower target — the multiple is a package deal of growth, risk, and margins; don’t paste the highest peer multiple on the lowest-growth name.
Comps as a single point — reporting one median without a range hides uncertainty and false precision.

Investor and analyst checklist

Define the target’s business model, scale, growth, and margin profile in writing.
Screen peers on model and scale, not industry code alone; target 6–12 names.
Normalize EBITDA (or revenue) with consistent SBC and one-time rules.
Build EV identically for every peer; verify net debt and share count.
Choose LTM, NTM, or mid-cycle denominators and use the same basis across the set.
Calculate EV/EBITDA (or chosen multiple) per peer; drop invalid denominators.
Report median and interquartile range; note outliers and why they stay or go.
Apply the band to target metrics; bridge EV to equity value per share.
Cross-check with a second multiple (e.g., EV/Revenue if EV/EBITDA is primary).
Compare comp-implied value to DCF and spot price; triangulate, do not average blindly.
Re-run when earnings season moves peer multiples or guidance shifts the target.
Document exclusions — audit trail beats a pretty table with no footnotes.

Key takeaways

Comps are relative, not intrinsic — they reflect what similar public stocks trade for today, not guaranteed fair value.
Peer quality beats peer quantity — seven matched names beat twenty industry codes.
Normalization is the work — EBITDA, EV, and LTM/NTM alignment matter more than spreadsheet formatting.
Harbor’s implied multiple fell 18.2× → 11.4× after peer and SBC fixes — the stock was fair, not deeply mispriced.
Triangulate with DCF and transaction data when available; no single method wins every situation.